UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl) UvA-DARE (Digital Academic Repository) An experimental study of the van Elswijk plan: value added taxation instead of wage taxation as a means to finance unemployment benefits van Winden, F.A.A.M.; Riedl, A.; Wit, J.; van Dijk, F. Link to publication Citation for published version (APA): van Winden, F. A. A. M., Riedl, A., Wit, J., & van Dijk, F. (2001). An experimental study of the van Elswijk plan: value added taxation instead of wage taxation as a means to finance unemployment benefits. Unknown Publisher. General rights It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons). Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible. Download date: 10 Jun 2020
198
Embed
UvA-DARE (Digital Academic Repository) An …an Elswijk Plan, he nev ertheless decided to con tract out the researc hto CREED, since the aforemen tioned request w as explicitly laid
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)
UvA-DARE (Digital Academic Repository)
An experimental study of the van Elswijk plan: value added taxation instead of wage taxationas a means to finance unemployment benefits
van Winden, F.A.A.M.; Riedl, A.; Wit, J.; van Dijk, F.
Link to publication
Citation for published version (APA):van Winden, F. A. A. M., Riedl, A., Wit, J., & van Dijk, F. (2001). An experimental study of the van Elswijk plan:value added taxation instead of wage taxation as a means to finance unemployment benefits. UnknownPublisher.
General rightsIt is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s),other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, statingyour reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Askthe Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam,The Netherlands. You will be contacted as soon as possible.
Faculty of Economics and EconometricsUniversity of Amsterdam
AN EXPERIMENTAL STUDY OF THE
VAN ELSWIJK PLAN
Value added taxation instead of wage taxation as a means to
�nance unemployment bene�ts
Prof. dr. F.A.A.M. van WindenDr. A. RiedlDr. J. WitDr. F. van Dijk
This is a translation and slightly revised version of the Dutch report "Experimenteel economischonderzoek naar het Plan Van Elswijk (CREED, August, 1999), commissioned by the Dutch Min-istry of Social A�airs and Employment. Translation by Miriam Hall.
CENTER FORRESEARCH IN EXPERIMENTAL ECONOMICS AND POLITICAL DECISION-MAKING, UNIVERSITY OF AMSTERDAM, AMSTERDAM, FEBRUARY 2001.
Preface
Starting point for this study is a motion adopted by the Second Chamber of the Dutch Parliamentin 1996.1 In this motion, a proposal to substitute net value added (value of production) for wagesas a tax base for social insurance contributions, the so-called Van Elswijk Plan, plays a centralrole as an instrument for creating employment. The Dutch Cabinet is requested, among otherthings, to commission the research centre for experimental economics CREED of the Universityof Amsterdam, to \test the e�ects of the Van Elswijk Plan in an international perspective". As aresult of this, CREED developed a project proposal that led to a commission from the Minister ofSocial A�airs and Employment.2 The ministry made it known that there was a need to quantifythe e�ects of the plan in areas which had not yet been investigated by the economic advisoryinstitute CPB (such as the e�ects on technological development, cross-border investments, thestructure of the sectors etc.). In addition, the advantages and disadvantages of the plan in thecontext of the Dutch economy should be weighed up.3 Although the Minister had some doubtsabout whether a laboratory experiment could provide more clarity about the e�ects of actuallyputting into practise the Van Elswijk Plan, he nevertheless decided to contract out the research toCREED, since the aforementioned request was explicitly laid down in the parliamentary motion.Moreover, he indicated that innovative research into strengthening employment opportunitiescould have a more general positive e�ect on acquiring relevant knowledge and insight. Forthe time being, the contract was limited to the �rst two parts of the project proposal. Thismeans that in the present study, attention is paid to the e�ects of the Van Elswijk Plan onemployment, capital ight, net wages, shifts between production sectors, and welfare. Dynamicfactors such as capital accumulation, technological development and economic growth couldnot be investigated. Consequently, there were limits to the extent to which the preconditionsset by the Minister could be ful�lled.4 The researchers believe that the conditions have been
1The motion 'Kalsbeek-Jasperse et al.' (Tweede Kamer, 1996-1997, 25.000 XV, no. 23)2The assignment is based on F.A.A.M. van Winden and F. Van Dijk, 'Experimenteel economisch onderzoek
naar het Plan Van Elswijk, projectvoorstel', CREED, University of Amsterdam, 10 October 1997.3Letter of the Minister to the Chairperson of the Second Chamber, 5 November 1997.4See the letter of the Minister of Social A�airs and Employment to the Second Chamber, 3 July 1997 (Tweede
Kamer, 1996-1997, 25 000 XV, no. 63). According to the Minister, in order to avoid a partial assessment,the following conditions should be adhered to. \(1) It must be expected that the research results provide newinformation that reduces the substantial uncertainty on the e�ects of the plan. Priority is given to informationon the dynamic e�ects of the plan on capital accumulation and technological development, which in turn willa�ect employment and the relative level of welfare; (2) It should be possible to model all the essential elementsof the Van Elswijk Plan on a realistic scale. These include the introduction of a labour subsidy, a tax on valueadded and its di�erentiated nature; (3) As much as possible, institutions and actors that together form the Dutcheconomy and would seem to be important for the e�ects of the Van Elswijk Plan should be included in the designof the experiment." As has already been mentioned, the nature of the assignment makes it impossible to payattention to capital accumulation and technological development (part of condition 1). As far as condition 3 isconcerned, the project proposal emphasises that it is not possible to completely replicate the Dutch economy viaany research method. It is impossible to avoid a certain level of abstraction. Complexity should be avoided in
i
ii PREFACE
adequately ful�lled: (1) The results of the research provide new information, (2) all the essentialelements of the Van Elswijk Plan have been modelled on a realistic scale, and (3) The design ofthe experiment takes into account the nature of the Dutch economy in so far as deemed to beimportant and (also �nancially) practically feasible.
For the �rst time, qualitative and quantitative information has been acquired about the fullimplementation of this plan in a real functioning 'international' economy in which the countrywith the Van Elswijk System is a small open economy with important characteristics of theDutch economy.
The researchers received advice and comments on the study from a Supervisory Committeeconsisting of the following members: Prof. A.L. Bovenberg, Prof. F.A.G. den Butter (Chairper-son), Prof. E.E.C. van Damme, Prof. K.P. Goudswaard, Prof. S.K. Kuipers and Prof. J.C. vanOurs. The Committee was involved in the de�nitive design of the experiment.5 Furthermore,the results and analyses were discussed with them, which led to an additional theoretical sen-sitivity analysis being conducted relating to the economic model which was at the basis of theexperiment.P. van Elswijk, ME, who devised the Plan, was also given the opportunity to comment on theexperimental design. The Supervisory Committee as well as Van Elswijk, formulated hypotheseswhich could be tested using the experimental results. These hypotheses expressed strongly di-verging expectations of the economic e�ects of the Van Elswijk Plan, so that a more structuredanalysis of the economic performance of the Plan in comparison to the present wage tax systemcould be achieved.
The report is structured as follows. Part 1 presents the de�nition of the issue and themost important research questions as they were formulated together with the Ministry and theSupervisory Committee. Part 2 discusses the design of the experiment. The economic modelthat is at the basis of the experiment and its implementation in the experiment are discussed.Then the parameters that were used and the reasons why they were chosen are presented. Part3 presents the results. First, the theoretical equilibrium results of the economic model arediscussed for both the wage tax system and the Van Elswijk system as are the results of thesensitivity analysis. Then the experimental results of both taxation systems are presented anddiscussed, followed by the results of the hypotheses tested and a discussion of these results. Bothparts end with conclusions. The `Executive Summary', which opens the report, summarises themost important conclusions of the research.
The research has been carried out by Dr. F. van Dijk, Dr. A. Riedl, Prof. F.A.A.M. vanWinden (project leader) and Dr. J. Wit.6 The researchers are very grateful to Ir. J. Theelen forhis excellent computer programming for the experiment. They thank the students G. Cotteleer,J.H.H. Notmijer and M. Smits for their help in carrying out the experiment and G. Cotteleeris also thanked for her help with analysing the results of the experiment. The researchers arefurther grateful to Dr. R.A. de Mooij from the CPB for data concerning the MIMIC model.Furthermore, helpful discussions with Prof. K. Hildenbrand (University of Bonn), Prof. C.R.Plott (California Institute of Technology), Prof. R. Selten (University of Bonn), Prof. S. Sunder
order to be able to trace important economic mechanisms. The nature of the Dutch economy has been taken intoconsideration in so far as this was deemed to be relevant and practically feasible (given the budgetary constraint).
5Prof. F.A.A.M. van Winden, dr. F. van Dijk, dr. A. Riedl and dr. J. Wit, \Experimenteel economischonderzoek naar het Plan Van Elswijk, Ontwerpfase 2", CREED, University of Amsterdam, 8 july 1998. Thisde�nitive design is based on the 'Ontwerpvoorstelfase 2' 16 February 1998, by the same authors.
6Dr. F. van Dijk now works for the Ministry of Justice and dr. J. Wit works for ATT in the USA. Of course,these organisations cannot be held responsible for this study in any way.
iii
(Carnegie-Mellon University) and Prof. C. Weddepohl (University of Amsterdam) are gratefullyacknowledged.Finally, the researchers are also grateful to M. Hall for doing the translation work and C. vanden Bos for proof-reading of the translation.
iv PREFACE
Executive summary
Van Elswijk Plan
Experts believe that unemployment in The Netherlands is partly due to the way in whichthe social insurance system functions. The present method of �nancing this system, to a greatextent via a tax on labour, is considered to play an important role in this respect. The resultingdi�erence between the labour costs of the employer and the net wage of the employee arethought to be damaging for employment (the \wedge" issue). It is argued that there is a viciouscircle between unemployment and higher labour costs.7 According to the Minister, plans for analternative, more employment-friendly �scal structure, therefore merit serious attention.8
The Van Elswijk Plan - named after its devisor - is such an alternative.9 In brief, the planconsists of three important parts. Firstly, the tax base for social insurance contributions isextended from the wage sum to net value added (against factor costs), by which not only labourbut also capital is taxed. Secondly, an employment or labour subsidy is introduced as a bonusfor the employer for the money saved on not having to pay out an unemployment bene�t toa person in employment. As this subsidy is lost if the employee is dismissed, the employer isdirectly confronted with the �nancial costs of unemployment to society. This is in contrast tothe present method of �nancing the social insurance system, whereby others also carry thesecosts via an adjustment in the contributions, through which, in fact, dismissal is subsidised.10
Thirdly, the tax on net value added is di�erentiated according to production sectors, such thaton introduction of the plan, the total social insurance contributions would equal the amount thatwas paid before the implementation of the plan. This would avoid a shift in the �nancial burdenfrom labour intensive to capital intensive �rms directly after implementation. According to theVan Elswijk Plan, the social insurance contributions are therefore determined as a percentageof the net value added speci�c to the production sector minus the employment subsidy (thestandardised unemployment bene�t equivalent) for those employed.Van Elswijk expects considerable positive economic consequences as a result of his plan, espe-cially in relation to its e�ects on employment.
Earlier evaluation of the Van Elswijk Plan
In 1995, the Dutch economic advisory institute CPB calculated the macro-economic e�ects ofthe Van Elswijk Plan using the so-called FKSEC model, the operational macro-economic model
7See, for example, CPB, \Alternatieve �nanciering van de sociale zekerheid:Plan Van Elswijk", Werkdocumentno. 79, Den Haag, 1995.
8Letter from the Minister of Social A�airs and Employment to the Second Chamber, 3 July 1997 (TweedeKamer, 1996-1997, 25 000 XV, no. 63).
9P. van Elswijk, \De markteconomie sociaal ingevuld", Van Gorcum, 1996. The essence of the plan is givenbrie y in P. van Elswijk, \Homeostatische �nanciering van de sociale zekerheid", ESB, 4-5-1994, pp. 408-411.
10See R.W. Boadway en D.E. Wildasin, Public Sector Economics, Toronto, 1984, p. 484.
v
vi EXECUTIVE SUMMARY
of this research institution.11 According to the CPB, fully implementing this plan would yield a45% tax rate on value added in order to �nance the removal of the social insurance contributionsand the labour subsidy (60% of the gross wage after deducting the employee contributions).The model was not considered suitable to calculate such a major operation: \These amountsare so large that they fall far outside the range in which calculations are normally made withthe FKSEC model. The FKSEC model is not considered suitable for such major changes, asthe equations are predominantly locally valid. Therefore a more modest size of the impulsesis chosen (about 10% of the original plan)".12 The results of this limited implementation, inwhich the di�erentiation of the tax rates according to production sectors was left out (thethird part of the Van Elswijk Plan), showed a clearly bene�cial e�ect on employment, evenin the long term (8 years), while the e�ects on production and investments were rather slight.The CPB emphasised that the manner in which wages are adjusted is an important source ofuncertainty when determining the e�ects on employment, regardless of the issues not studied,concerning capital ight and practical feasibility (valuation problems,13 sensitivity to fraud,avoidance of tax by international businesses, the technical feasibility of di�erentiating the taxrate according to production sectors as well as it possibly being in con ict with the principle ofequal treatment14)15 All things considered, the CPB concluded that the risks of implementingthis plan were too large. Incidentally, it is noted that positive e�ects on employment of the VanElswijk Plan were also found in an earlier study referred to by the CPB.16
A problem with these models is that they are based on existing behavioural reactions, thatis, behavioural reactions to the present way of �nancing the system. In this context, one of thetwo economic experts consulted by the Ministry warned that policy measures can cause suchlarge changes in behavioural reactions that it makes no sense to calculate the e�ects of the policymeasures with models which are based on existing behavioural reactions.17 General equilibriummodels, such as the MIMIC model of the CPB, meet this criticism. According to the CPB study,additional calculations with the MIMIC model support the results of the FKSEC model.18
The experts previously mentioned also point at the aforementioned implementation problemsin their comments, and they indicate the importance of the way in which wages will be adjusted.It is expected that capital ight will take place because of the (implicit) tax on capital in theVan Elswijk Plan. The net (marginal) return on capital will decrease, which may induce Dutch�rms to invest more in foreign countries, and foreign �rms to invest less in the Netherlands.Moreover, as the more pro�table �rms within a production sector will be hit relatively thehardest, the competitive position of the Netherlands could be damaged because the incentive toinnovate and modernise will be negatively in uenced. Entrepreneurship will be rewarded less. Itis also expected that the extra taxation of the production factor capital, will have a discouraginge�ect on the introduction of labour saving technology. This, combined with the labour subsidy(through which the costs of dismissal will increase), will inhibit the functioning of the labour
11See note 7.12Ibid., p.12.13'Net value added ' is, for example, not a current business economic concept.14It must be noted, though, that the di�erentiation is meant to keep the tax burden on implementation equal
to the tax burden under the present method of �nancing.15Also see CPB, MEV 1997, p. 107.16See CPB, ibid., p. 7. The CPB also presents a summary of other studies into the e�ects of broadening the tax
base of social insurance in the Netherlands. It concludes that: \Macro-calculations by the NEI and SEO showedfavourable e�ects on employment, as did the CPB calculations in work document 11. The e�ects on productionand investments varied from practically none to negative" (ibid., p. 6).
17See the Minister's letter of 3 July 1997, p. 5.18Ibid., p. 15. In this model, wages are determined on the basis of negotiations between employers and unions.
vii
market. All things considered, these experts predict large adverse e�ects on the competitiveposition and welfare of the Netherlands, and they do not consider the Van Elswijk Plan as asolution to the issue of employment.19
Finally, we would like to mention the evaluation of a �eld experiment in Rotterdam, inwhich the Van Elswijk Plan was implemented on a limited scale (twelve �rms �nally took part)and for a limited experimental period (1996-1997).20 The experiment was particularly designedto investigate implementation problems. The results suggest that the Plan is feasible and canwork alongside other existing forms of taxation. Moreover, it will further the creation as wellas the maintenance of employment. As far as the latter is concerned, it was concluded that theemployers considered the labour subsidy as a sort of investment premium that they could alsouse for recruiting graduates, for training courses and for accelerating investments (in which case�nancial barriers and the risks attached to investments seem to play a role).
In economic science, �eld experiments are generally regarded as a useful instrument forobtaining empirical data about the e�ects of radical changes in government policy.21 However,for a number of reasons, the signi�cance of the results obtained in this case is questionable.22
Firstly, the reliability of the results is doubtful because of the short experimental period and thelimited number of �rms taking part (initially 7, later 12). Owing to this, chance may have playedan important role. For example, one of the �rms was responsible for 70% of the increase in thenumber of labour years. Furthermore, participation was voluntary and the �rms were given theguarantee that they would not lose out (\no pay guarantee"), so they could only bene�t fromthe experiment. Moreover, the small scale of the experiment meant that possible in uences onthe economic structure (such as shifts between labour intensive and capital intensive sectors)could not be observed. For the same reason, it could also not be clearly established in howfar the results were determined by other factors (for example, speci�c regional developments).The small-scale implementation also implied the absence of economic feedback mechanisms, likevia wages, and the presence of unclear distortions through the fact that the di�erent ways of�nancing social insurance occurred simultaneously.
The experimental economic study of the Van Elswijk Plan
Experimental economics
Besides the econometric method, as applied by the CPB, and the survey method, conductingexperiments is seen as an important additional method of obtaining empirical data about theeconomic e�ects of government policy.23 A distinction is made between the aforementioned�eld experiments and laboratory experiments. In particular, the latter are associated with therapidly developing new research area of experimental economics. As a laboratory experiment isused in this study of the Van Elswijk Plan, this research method will �rst be brie y explained.
19R. de Mooij comes to a similar conclusion on the basis of his review of the book by Van Elswijk in ESB,9-4-1997.
20See NEI, Evaluatie experiment Prohef, Rotterdam, May 1998.21A much cited example concerns the �eld experiments in the USA looking into the labour supply and other
e�ects of a negative income tax (see, for example, R. Ferber andW.Z. Hirsch, Social experimentation and economicpolicy: a survey, Journal of Economic Literature, 16, 1978, pp. 1379-1414; J.A. Hausman and D.A. Wise, eds.,Social Experimentation, University of Chicago Press, Chicago, 1985).
22The majority of the reasons are also given in the letter from the Minister, dated 3 July 1997.23See, for example, H.S. Rosen, Public Finance, Irwin, Homewood/Boston, 1992, chapter 3.
viii EXECUTIVE SUMMARY
In an economic laboratory experiment, economic issues are studied using subjects in a con-trolled environment (a laboratory). In the laboratory, the subjects are faced with an economicsituation in which they have to make decisions that have �nancial consequences for them. Thegreat strength of this method lies in the possibility of systematically investigating changes inan economic situation, such as a change in the taxation system (controllability of the circum-
stances). Behavioural reactions can be closely observed. Moreover, this method enables theexperiment to be accurately repeated (possibility of replication). Other research methods thatempirically analyse behavioural reactions are not able to control the economic circumstances insuch a way, so that all sorts of noise can occur. Furthermore, in an experiment, the researchcan be concentrated on that aspect of a theory, hypothesis or claim (such as in relation to thesupposed e�ects of the Van Elswijk Plan) considered the most important. The economic situa-tion can be simpli�ed as much as possible. Assuming that the claim or theory should also holdtrue in such simple circumstances, the tenability of the claim or theory is highly implausibleif the result is negative, whereas with a positive result, the burden of proof is shifted on tothose who contest the result.24 They will then have to indicate where the experiment has failed,which could lead to further research. For these reasons, laboratory experiments are particularlysuitable for investigating the e�ects of institutional changes, such as policy innovations. In thisway, information can be obtained about the possible e�ects, without the �nancial or other risksthat would occur if the policy innovation was actually carried out.
Nowadays, laboratory experiments are conducted in practically all areas of economic science.Much is published on this topic in eminent economic journals and various textbooks have beenwritten.25 An important area of research is related to the performance of markets for productsand production factors (labour and capital). Policy-oriented research, in which aspects charac-teristic of certain markets are implemented in laboratory experiments to investigate the e�ectsof government regulations, increasingly fall under this category.26 Studies have also appearedwhich investigate the e�ects of taxation and international trade using complete systems of mar-kets for products and production factors (economies).27 These were used when designing theexperimental research into the Van Elswijk Plan.
24See J.H. Kagel and A.E. Roth (eds.), The Handbook of Experimental Economics, Princeton University Press,Princeton, 1995, p. 58.
25See for example D.D. Davis and C.A. Holt, Experimental Economics, Princeton University Press, Prince-ton, 1993; J.H. Kagel and A.E. Roth, The Handbook of Experimental Economics, Princeton University Press,Princeton, 1995. There is also a journal Experimental Economics, published by Kluwer.
26See Kagel and Roth, ibid., pp. 55-58. Other policy-directed experiments that have taken place include, forexample, markets for tradable emission rights and the auction of radio frequencies. CREED has, for example, doneresearch for the ower auction in Aalsmeer (VBA) and the Ministry of Transport and Public Works (auctioningof radio frequencies and privatisation of the railways). For indications that policymakers take these studiesinto consideration see, for example, D.M. Grether, and C. Plott, The e�ects of market practices in ologopolisticmarkets: an experimental examination of the ethyl case, Economic Inquiry, 22, 1984, pp. 479-507, C. Plott,Laboratory experiments in economics: the implications of posted-price institutions, Science, 232, pp. 732-738,M. Bycovsky, M. Olson, and A. Schram, Veiling van etherfrequenties, ESB, 1-3-1995, pp. 201-205, J.O. Ledyard,C. Noussair, and D. Porter, The allocation of a shared resource within an organisation, Economic Design, 2,1996, pp. 163-192, and C. Plott, Laboratory experimental testbeds: application to the PCS auction, Journal ofEconomics and Management Strategy, 6, 1997, pp. 605-638.
27See, for example, H.C. Quirmbach, C.W. Swenson and C.C. Vines, An experimental examination of generalequilibrium tax incidence, Journal of Public Economics, 61, 1996, pp. 337-358; C.N. Noussair, C.R. Plott andR.G. Riezman, An experimental investigation of the patterns of international trade, American Economic Review,85, 1995, pp. 462-491.
ix
Strengths and limitations of the research
The most important goal of the experimental research into the Van Elswijk Plan is to gaininsight into the behavioural reactions to a change in the method of �nancing social insurance andthe economic e�ects of this change. Employment plays a central role, but e�ects on productionstructure, wages, capital ight and welfare are also taken into consideration. The hypothesesformulated by the Supervisory Committee and Van Elswijk concerning these e�ects are testedin the study. As mentioned above, a laboratory experiment has the relative advantages, in com-parison to econometric and �eld experiments, that these e�ects and the hypotheses formulatedcan be systematically investigated under controlled circumstances and can be replicated. As allthe essential elements of the Van Elswijk Plan are realistically incorporated in the experiment,both qualitative and quantitative data are obtained for the �rst time on the e�ects of a complete
implementation of this plan in an actual operating international economy in which the country
with the Van Elswijk system approximates a small open economy with important characteristics
of the Dutch economy.
Technical and institutional aspects of the Dutch economy are taken into account in thedesign of the experiment. Data are used from the applied general equilibrium model, MIMIC,of the CPB.28As mentioned, the CPB made additional calculations with this model regardingthe economic e�ects of the Van Elswijk Plan. In contrast to the econometric FKSEC model,in which the e�ects are determined under the assumption that the behavioural reactions arethe same as with the present way of �nancing social insurance, in a general equilibrium model,the behaviour of economic agents can adjust to the policy innovation (the new way of �nancingsocial insurance). However, in comparison to a laboratory experiment, this type of model hasthe disadvantage that certain behavioural assumptions still need to be made based on micro-economic theories (for example, utility maximisation). In an experiment, behaviour is freeto change within the given institutions. Moreover, general equilibrium models usually have anumber of solutions (equilibria), which means that it is unclear in which direction the economice�ects will occur. In contrast, in an experiment, the direction in which the economy developscan be observed. The results of the research will demonstrate the advantage of this.
The research also has limitations. Firstly, it should be noted that the research assigned toCREED is limited to the �rst two parts of the project proposal. Therefore, no attention ispaid in this study to dynamic factors such as capital accumulation, technological developmentand economic growth. Secondly, it should be realised that, as with all economic models, certaineconomic aspects have to be omitted from this study for reasons of manageability and tractabilityof the e�ects.29 Thus, no attention is paid to implementation problems, for example.
Furthermore, there is the issue of the external validity of the experimental results. To whatextent can it be expected that the observed e�ects will also take place in a natural environment(in practice)? We would like to make the following remarks. It has already been stated that
28A possibly important di�erence with the MIMIC model concerns the design of the labour market, whichshows a fully decentralised determination of wages, instead of negotiations between unions and employers. Eventhough technical considerations played a role here, this choice can be justi�ed by the fact that wage negotiationsare strongly decentralised nowadays and that they do not occur at the level of the sectors di�erentiated in thestudy (the exposed and sheltered sector). Furthermore wages are not only determined by negotiations, but alsoby the incidental wage component, or \wage drift". Finally, there is no consensus on modelling the labour market,among economists.Other possibly important di�erences with the MIMIC model are the fact that the heterogeneity of labour isignored and that attention is focussed on voluntary unemployment.
29If one is not satis�ed with such a procedure, information can only be obtained by actually carrying out thepolicy innovation.
x EXECUTIVE SUMMARY
when no support is found for a theory or claim in a relatively simple environment, it is unlikelythat this support will exist in a complex real life situation. If support is found, the burden ofproof can be shifted onto the critics, who should then indicate where the experiment has failed,which could lead to altering the design of the experiment. Furthermore, it has been found thatexperimental markets appear to be able to reproduce phenomena that are observed on actualmarkets.30 In addition, there is the general �nding that competitive markets can be createdand studied in the laboratory. Con�dence in the practical signi�cance of this research methodis also suggested by the aforementioned fact that governmental institutions as well as privateinstitutions are increasingly relying on experimental research to obtain information about thepossible e�ects of important policy changes and are found to adjust policy in accordance withthis information.31 Just as with, for example, the testing of prototypes of new aeroplanes in awind tunnel or testing the design of dams in a hydrology laboratory, policy-oriented laboratoryexperiments derive their strength from the systematic and controlled investigation of an issuewithin a simpli�ed environment, in which the focus is on those mechanisms considered essential(the \wind tunnel" function of an experiment). Finally, we would like to refer to the followingstatement from a leading economist in the �eld of experimental economy: \While laboratoryprocesses are simple in comparison to naturally occurring processes, they are real processes inthe sense that real people participate for real and substantial pro�ts and follow real rules indoing so."32
It should be emphasised that experimental economics is not regarded as a substitute forother forms of empirical research in which, for example, an attempt is made to approach thecomplexity of reality as far as possible (such as in the macro-economic policy models of the CPB)or for theoretical research. It is a complementary research method with relative advantages anddisadvantages compared to other methods.
Macro-economic experiments are on the increase and have already achieved interesting re-sults. In general, there is already ample experience with the experimental investigation ofmarket systems.33 However, the application of this research method is new concerning thepolicy-oriented study of the economic e�ects of macro-economic policy innovations. Also forthis reason, a caveat holds for this study. Finally, due to the abstractions necessary in any the-oretical and empirical research, \common sense" remains an important resource when makingde�nitive policy decisions.
Summary of the research results
The most important conclusions that can be drawn from this experimental study into the VanElswijk Plan are summarised below. First, there is a comparison of the economic e�ects observed
30An example is the phenomenon of the \winner's curse" (see Kagel and Roth, ibid. p. 60). This phenomenonconcerns the claim - �rst expressed concerning the auction of rights to extract oil- that the actual yield (value) ofthe object attained is often disappointing in some auctions. Experiments have aÆrmed the idea that this is dueto the fact that the winning bidder tends to be the one with the highest expectation on the value of the objectbeing auctioned and that this bidder does not suÆciently keep in mind that his or her expectation may be toooptimistic.
31See note 26.32C.R. Plott, Industrial organisation theory and experimental economics, Journal of Economic Literature, 20,
1982, p. 1486.33See note 27, and, in addition, J. Davis and C. Swenson, Experimental evidence on tax incentives and demand
for capital investments, The Accounting Review, 3, 1993, p. 482-514; J. Goodfellow and C. Plott, An experimentalexamination of the simultaneous determination of input prices and output prices, Southern Economic Journal,1990, pp. 969-983.
xi
in the experiment of the present wage tax system with the alternative Van Elswijk systemas means of �nancing unemployment costs. In addition, some conclusions are presented of atheoretical general equilibrium analysis concerning the economic model at the basis of the smallopen economy implemented in the experiment. Finally, a number of additional observationsregarding the study will be made.
Summary
The most important experimental results concern the testing of the hypotheses regardingcapital ight, employment, net wages, the relative size of the labour intensive (sheltered) sectorand welfare. The results of these tests are favourable for the Van Elswijk system. In comparisonto the wage tax system, under the Van Elswijk system there is no capital ight and employmentdoes not decrease. Ultimately, a substantial capital import and increase in employment areobserved. The greater production leading from this does not appear to be more labour intensive.In this regard, the internationally operating (exposed) sector is not displaced. When tax ratesare adjusted to the budget de�cit, there are also positive e�ects on net wages and on the variouswelfare indicators. Few negative e�ects (decreases), in the net wages and in one of the fourwelfare indicators, only occur if tax rates are not adjusted to the budget de�cit, that is, if theyare kept constant. However, this is coupled with a substantial budget de�cit for the wage taxsystem and a substantial budget surplus for the Van Elswijk system.
Furthermore, it appears that with the wage tax system, the (domestic) wage tax rate alwayshas a negative e�ect on the variables that are central to the hypotheses and this e�ect is usuallysigni�cant. With the Van Elswijk system the e�ect of the product tax rate is in the samedirection, but it is always less negative and usually not signi�cant.
These economic e�ects of the Van Elswijk system can be explained by two factors. In the�rst place, the assumed presence of some immobile capital leads to a shift in the tax burdenfrom labour to capital under this tax system, in which not only labour but also capital is taxed(implicitly). This makes possible an increase in eÆciency (a well-known result of the theory ofoptimal taxation). In the second place, it appears that the shift in the moment of taxation playsan important role. With the wage tax system, the producer is already confronted with taxationat the moment of acquiring production factors (labour). At this moment, it is still uncertain towhat extent the production costs that are incurred will lead to pro�ts. On the contrary, withthe Van Elswijk system, the costs of the production factor labour are no longer taxed but are infact subsidised. Moreover, taxation takes place at a later stage and is in proportion to the pricedevelopment on the product markets. In contrast to the wage tax system, with this system thegovernment therefore shares the sales risk of the producer. The relatively better performance ofthe Van Elswijk system can be explained by also taking into account these uncertainty e�ects.
The additional theoretical analysis, in which these uncertainty e�ects are ignored, shows thatthe Van Elswijk system can lead to better economic performances than the wage tax system dueto the mere possibility of shifting the tax burden to capital. We say, \can lead" because besidesa \favourable" equilibrium for this taxation system in comparison to the wage tax system, an\unfavourable" equilibrium is also found. In the \favourable" equilibrium, there is a considerableincrease in employment with some capital ight. However, in the \unfavourable" equilibriumthere is a considerable decrease in employment and substantial capital ight. The \favourable"equilibrium, which receives the most empirical support in the experiment, does not disappearwhen a sensitivity analysis is conducted, involving changes in the unemployment bene�t levelor a decrease in the labour subsidy. The theoretically favourable e�ects of this taxation system
xii EXECUTIVE SUMMARY
also appear to have less to do with the labour subsidy than with the increase in eÆciency whichoccurs through the implicit taxation of the �xed factor (immobile capital). This is because ofthe higher tax rates accompanying a subsidy. However, even if this �xed factor is considerablyreduced, the levels of employment and welfare in the \favourable" equilibrium are still clearlyhigher than with the wage tax system. The same holds for when the ratio of the unemploymentbene�t rate and wage rate is held constant instead of there being a �xed bene�t.
Additional remarks
All things considered, the results of this experimental study demonstrate noticeable positivee�ects for the Van Elswijk system compared with the present wage tax system. From an experi-mental point of view, there appears to be no reason to doubt the reliability of the results. If theexperiment were repeated, it is expected that the results would be replicated. The theoreticalanalysis performed in addition to the study further supports the �ndings.
Given that the Dutch situation was taken into account as much as possible when designingthe experiment, it seems justi�ed to conclude that the economic e�ects of the Van Elswijk systemon the Dutch economy could be substantial and positive. An increase in employment of morethan 10% could be possible. For the �rst time, there are now indications for this based on anactual working economy in which the Van Elswijk system has been fully implemented. At thevery least, no indications have been found that run contrary to this conclusion. With a view tothe desirability of a possible debate at the European level, we would like to point out that thetheoretical analysis demonstrated that it would be more favourable for the Van Elswijk systemto be introduced simultaneously at home and abroad. However, this possibility has not beeninvestigated experimentally. Therefore, given the fact that the experimental results di�er fromthe theoretical results, empirically nothing can be said about this issue at this stage.
Although it is expected on the basis of the strength and the systematic nature of the observede�ects that replicating the experiment would produce similar results for the Van Elswijk system,it should be emphasised that the number of experimental sessions in this study were the minimumnumber considered necessary. If the Van Elswijk Plan is to be considered further, it is importantfrom an experimental economic point of view that the robustness of the results is studied byreplicating the experiment. An important reason for this is the theoretical �nding that thereis also a (very) unfavourable equilibrium for the Van Elswijk system. The economic process inthe experimental economy does not appear to be attracted by this equilibrium. Repeating theexperiment can determine whether this is a robust result. Furthermore, a sensitivity analysiscould then be carried out. The possibility of investigating the e�ect of substituting the �xedunemployment bene�t with a bene�t adjusted to the wage level (a constant replacement rate)is important in this regard. Another important option would be to analyse the e�ect of the sizeof the �xed factor (immobile capital).
In our opinion, this study demonstrates that laboratory experiments can be a useful meansof studying the e�ects of macro-economic policy innovations. From a broader perspective, thereis the valuable prospect of linking up experiments, through which empirical information canbe obtained regarding presumably important behavioural reactions to policy innovations, andtraditional macro-economic policy models.
D Convergence analysis for the wage tax system 163
E Detailed results of hypotheses testing 167
F Production eÆciency and FGLS-estimates 177
Part I
De�nition of the Issue
3
Chapter 1
The research questions
In the motion adopted by the Second Chamber of the Dutch Parliament concerning the VanElswijk Plan, the e�ects on employment played a central role. In the motion, CREED wasrequested to investigate these e�ects experimentally in an international perspective. During dis-cussions with the Ministry of Social A�airs and Employment, it became apparent that there wasa need for further information on the quantitative e�ects of the Plan, leading to a more generalassessment. For example, implementation of the plan could lead to changes in the internationalcompetitiveness of the Dutch economy, especially with respect to the most important tradingpartners in the European Union. The fact that The Netherlands is a "small open economy" isof importance in this context. Since capital, rather than labour, is generally regarded as beinginternationally (very) mobile, the occurrence of "capital ight" should be seriously considered ifcapital is to be taxed. This would, at least implicitly, be the case if the Van Elswijk Plan wereto be implemented, because a value-added tax would a�ect both wages and pro�ts. If capitalwere to "leak" abroad this would have various economic consequences. It could lead to thesubstitution of capital by labour and to a decrease in production capacity, which would in turna�ect employment. Since some production sectors are relatively more dependent on the inputof capital than others, shifts could also occur in the relative size of sectors. This could leadto Dutch production becoming more labour intensive. Another important characteristic of theDutch economy, which also holds for many other economies, is that it is possible to distinguishroughly two sectors. Firstly, there is an exposed sector, which produces products that are tradedon the international market and is therefore directly exposed to international competition. Pro-duction in this sector is relatively capital intensive. Secondly, there is a sheltered sector thatproduces products for the internal market and is relatively labour intensive.34
These considerations have led to the following de�nition of the research issue:
De�nition of the issue
What would be the economic performance of the Van Elswijk Plan compared to the present way
of �nancing social insurance, in a small open economy with a relatively labour intensive sheltered
sector and a relatively capital intensive exposed sector, when the present way of �nancing social
insurance is maintained abroad?
In order to achieve a more structured evaluation of the economic performance of the VanElswijk Plan compared to the present �nancing system (the wage tax system), the researchers
34A similar di�erentiation is also made in the CPB's applied general equilibrium model MIMIC.
5
6 CHAPTER 1. THE RESEARCH QUESTIONS
asked Van Elswijk and the Supervisory Committee to formulate a number of hypotheses thatcould be tested using the results of the experiment. This is because there are no reasons whyopinions concerning these methods of �nancing the system could not be applied to the realeconomic process in the experiment, of which the design had their approval. Five hypotheseswere formulated, concerning the following variables which were judged as important:
� capital ight
� employment
� the net wage
� the relative size of the production sectors
� welfare
Point of departure is the \null hypothesis" that no (signi�cant) changes will be found betweenthe two methods of �nancing the system - the wage tax system and the Van Elswijk system. Inthe listing of the hypotheses below this hypothesis is named the 'Null hypothesis'. The alter-native hypotheses formulated by Van Elswijk and the Supervisory Committee concerning theexperimental results are brie y denoted as 'VE hypothesis' and 'SC hypothesis', respectively.35
Hypotheses
Hypothesis 1 Capital flight
Null hypothesis: Implementing the Van Elswijk system will not lead to capital ight.
VE hypothesis: The same as the null hypothesis.
SC hypothesis: Capital ight will take place.
Hypothesis 2 Employment
Null hypothesis: Implementing the Van Elswijk system will not lead to a change in total employ-
ment.
VE hypothesis: Employment will increase.
SC hypothesis: Employment will decrease.
Hypothesis 3 Net wage
Null hypothesis: Implementing the Van Elswijk system will not lead to a change in the net wages
of employees.
VE hypothesis: Net wages will not decrease.
SC hypothesis: Net wages will decrease.
Hypothesis 4 Shifts in production
Null hypothesis: Implementing the Van Elswijk system will not lead to a shift in the relative
production level of the production sectors.
VE hypothesis: Conditional.36
SC hypothesis: The capital intensive sector will become smaller compared to the labour intensive
sector.35It is noted that the hypotheses formulated by the Supervisory Committee do not necessarily concur with the
personal opinion of the individual members.36Van Elswijk's opinion did not become totally clear to the researchers. It was concluded (but not con�rmed)
that if a shift should take place, according to Van Elswijk this would be inversely related to the relative marketprices in the production sectors. This is the reason why his hypothesis has been phrased 'conditional'.
7
Hypothesis 5 Welfare
Null hypothesis: Implementing the Van Elswijk system will not lead to changes in welfare.
VE hypothesis: Implementing will lead to an increase in welfare.
SC hypothesis: Implementing will lead to a decline in welfare.
In chapter 4, in which the hypotheses are tested, a precise de�nition of the main variables willbe given. Important macro-economic variables such as the budget de�cit and the gross nationalproduct (GNP) will also be considered when assessing the level of welfare.
Besides testing the hypotheses, which we believe to clearly re ect the contesting opinions inthe debate concerning the Van Elswijk Plan, attention will also be paid to more dynamic aspectsof both tax systems. The development of the economic process under the alternative methods of�nancing social insurance and the reactions of producers and consumers will be focussed upon.A theoretical analysis will be presented in addition to the experimental study requested, whichcan be used as an additional source of information.
8 CHAPTER 1. THE RESEARCH QUESTIONS
Part II
Research Approach
9
Chapter 2
Experimental Design
2.1 Introduction
In this chapter the economic model forming the basis of the experiment as well as its implemen-tation will be described.
The experimental design used should comply with several di�erent criteria. Firstly, thedesign should include a social insurance system that can be �nanced in di�erent ways. In thecase of the Van Elswijk Plan this means that it must be possible to levy di�erent tax rates onlabour and capital. Secondly, the design should include elements that can be interpreted withinthe context of the Dutch economy. Here it is possible to di�erentiate between several elements.The experimental economy should include at least two production sectors which make use ofdi�erent technologies (that is, labour-capital ratios) to produce their products. There should bea sheltered sector in which production is directed to the domestic market and the productionis relatively labour intensive, and an exposed sector, in which production is directed to theinternational market and is relatively capital intensive. There must also be a labour market inwhich unemployment can occur as an endogenous variable determined by the economic process.The labour market will have a domestic nature, given the relatively limited international mobilityof labour. E.g., a Dutch worker will not look for employment on the German labour marketand vice versa. In contrast to the labour market, there must be an international capital marketin order to study the e�ects of internationally mobile capital. Finally, the domestic (Dutch)economy should be quite small relative to that abroad, in order to comply with the criterion ofa small open economy. Thirdly, the economic incentives given to the subjects must be as clearas possible to avoid unnecessary complexity. Fourthly, it should preferably be possible to applyrelevant economic theories and �ndings, in order to be able to obtain a benchmark and to beable to �nd relevant parameter values.
Speci�c choices had to be made concerning the experimental economy, given that there is nogenerally accepted economic model of a macro-economy nor of the Dutch economy, in particular.In doing so the aforementioned criteria have been taken into account in order to reach a goodexperimental design.An important decision had to be made concerning the design of the labour market and theoccurrence of unemployment. In the chosen design, unemployment occurs because employeesare not willing to o�er their labour on the labour market at any wage. This behaviour is inducedas a result of two factors. Firstly, employees derive utility from labour units that are not soldon the labour market. Secondly, the existence of social insurance bene�ts inhibits the supply of
11
12 CHAPTER 2. EXPERIMENTAL DESIGN
labour. A second important decision had to be made concerning the design of the productionsectors. By di�erentiating between a relatively labour intensive sheltered sector and a relativelycapital intensive exposed sector, as well as through the speci�cation of the production function,the design will be qualitatively similar to the MIMIC model of the CPB.37
The rest of this section is organized as follows. Producers, consumers, and markets of theeconomy are described in 2.2. The social insurance system and the role of government will bedealt with in 2.3. In 2.4 the foreign economy will be discussed after which the implementationof the economic model in the experiment and the design of the experimental sessions will bediscussed in 2.5. In 2.6 the choice of the di�erent parameter values will be presented.
2.2 Producers, consumers, and markets
The economy encompasses three di�erent agents: producers, consumers (who also own produc-tion factors) and a government. The characteristics and incentives of producers and consumerswill �rstly be discussed below. This will be followed by a description of the implemented markets.The role of the government will be discussed thereafter.
Producers
In the experimental economy there are two production sectors present in both the home andthe foreign country. One production sector produces good X, which production is relativelycapital intensive, and the other sector produces good Y, which production is relatively labourintensive. In total, nxk and nyk, producers participate in these production sectors, where thehome country is denoted by k = h and the foreign country is denoted by k = f . A producer i incountry k from the production sector Z, where Z = X;Y , produces according to the followingproduction function:
Zki = Azk(�1� zz L z
zki + (1� �z)1� zK z
zki)�= z with Z = X;Y; z = x; y; and k = h; f
where Lzki and Kzki denote the number of units of labour and capital, respectively, used by theproducer to produce Zki. Azk represents a scaling factor and z, �z, and � represent productionparameters (see 2.6).
For both goods X and Y there is a market on which the producers can sell these products.Good X can be traded internationally. This means that a domestic X producer can sell hisproducts in both the home and the foreign country. The same applies to a foreign producer.Product Y , however, can only be traded on the domestic market. This means that a domesticproducer of Y can only sell his products on the domestic market and that a foreign producer ofY can only sell his products on the foreign market. Therefore, the production sector X is calledthe exposed sector and the production sector Y is called the sheltered sector.
The price of product X is denoted px and the price of Y in country k is denoted pyk.Additional costs for producers form their contributions to the social insurance system. Thetotal contribution of producer i in the production sector Z in country k is denoted by Pzki. Wewill specify how this costs are calculated later on. The producer may possibly receive a subsidy
37See J.J. Graa and and R.A. Mooij (1994), Analysing Fiscal Policy in the Netherlands: Simulations with arevised MIMIC, Research memorandum No. 140, Central Planning Bureau, The Hague.
2.3. GOVERNMENT AND SOCIAL INSURANCE SYSTEM 13
for employing labour. The size of this subsidy is equal to Szki. The pro�t made by producer iin the production sector X in country k is equal to:
�xki = pxXki � wkLxki � rKxki � Pxki + Sxki;
The pro�t made by producer i in the production sector Y in country k is equal to:
�yki = pykYki � wkLyki � rKyki � Pyki + Syki;
where wk represents the market price of labour (the market wage) in country k and r representsthe market price of capital (the capital price).
Consumers
Consumers are endowed with both labour and capital. Both are production inputs requiredfor production. There are nch consumers in the home country and ncf consumers in the foreigncountry. In total there are nc = nch + ncf consumers.
Any consumer j in country k possesses a number of labour units �Lkj, which can be sold onthe labour market. Each consumer also receives a bene�t w0k for each unit that is not sold onthe labour market. At the same time, j also possesses a number of capital units �Kkj, which canbe sold on the capital market. Furthermore, consumer j has a portfolio of shares of the �rms inhis country, and therefore receives part of the pro�ts made by these �rms. Suppose that j sellsLkj labour units on the labour market for an average wage of wk and Kkj capital units on thecapital market for an average price of r. Furthermore, assume that each consumer in country khas 1=nck shares in each �rm in that country. Then the consumer's budget is given by
Bkj = wkLkj + w0k(�Lkj � Lkj) + rKkj +�k=nck;
where �k is the total pro�t of all �rms in country k.
Consumer j derives utility Ukj from consuming the goods X and Y , and leisure �Lkj � Lkj.We assume that each consumer in country k has the following utility function:
Ukj = XakjY
bkj(
�Lkj � Lkj)c; a > 0; b > 0; c > 0;
where Xkj and Ykj are the number of units of X and Y bought by the consumer and a, b and care preference intensity parameters (see 2.6).
Markets
From the above it follows that the economy consists of a total of six markets. Three ofthese markets are factor markets for the production factors: a domestic labour market, a foreignlabour market, and an international capital market. There are also three output markets: adomestic market for product Y (Yh), a foreign market for product Y (Yf ), and an internationalmarket for product X which is relatively capital intensive.
2.3 Government and social insurance system
In each country there is a social insurance budget out of which unemployment bene�ts arepaid. They are �nanced by the producers' social insurance contributions. Both the national and
14 CHAPTER 2. EXPERIMENTAL DESIGN
foreign government carries sole responsibility for managing the social insurance system. Theyprovide bene�ts for unemployed labour units and receive social insurance contributions from theproducers. Furthermore, they are supposed to select the tax rates that will balance the budget.
Suppose that the total number of labour units in country k is equal to �Lk :=P
j�Lkj and
that Lk =P
j Lkj is actually used in the production process. Then the unemployment rate Wk
will be equal toWk := 1� Lk=�Lk:
The government in country k provides a bene�t w0k for each labour unit that is not usedin the production process. These bene�ts are �nanced by the social insurance system. Twodi�erent types of �nancing this system can be distinguished: the wage tax system and the VanElswijk system.
Wage tax system
Under the wage tax system (system A), which is a simpli�ed version of the actual taxationsystem, producers pay a �xed proportion over the wage sum. In other words, a certain percentageof the wage of each unit of labour employed goes towards �nancing unemployment bene�ts.Assume that this percentage is equal to �wk � 100%. Then the total contribution of producer iin production sector Z in country k will be equal to:
PAzki = �wkw
Ak L
Azki;
where A denotes the wage tax system. Consequently, the government will receive �wkwAk L
Ak ,
where LAk is total employment in country k.
The government's expenses in country k are equal to w0k(�Lk � LAk ). For a balanced budget
the following condition has to be satis�ed:
�wkwAk L
Ak = w0k(�Lk � LA
k ):
The Van Elswijk system
Under the Van Elswijk system (system B) producers pay a certain percentage of the priceof each sold unit of output to the government to �nance unemployment bene�ts. Furthermore,they receive an employment subsidy w0k for each unit of labour employed, irrespective of theproduction sector. The tax rate di�ers across the production sectors. According to the VanElswijk Plan the percentage per production sector is determined by the actual level of the taxburden at the time when the change from the old system (A) to the new system (B) occurs. Inthe following this is explained in detail:
Assume that the tax rate for a producer in production sector X in country k is equal to �pxkand that the tax rate for a producer in production sector Y is �pyk. Then the total tax burdenand the subsidy received for producer i in production sector X (Y ) in country k will be equalto:
PBxki = �pxkp
BxX
Bki and Sxki = w0kL
Bxki (P
Byki = �pykp
BykY
Bki and Syki = w0kL
Byki):
Note that the government's expenditure will be independent of the size of unemployment inthis system. This is because in the Van Elswijk system the government pays w0k, both for eachunemployed unit of labour and each unit of labour employed (as an unemployment bene�t anda subsidy on employment, respectively). The total expenditure will therefore remain constant
2.4. FOREIGN COUNTRY 15
since it will always be equal to w0k�Lk. In order to have a balanced budget under this system,
the following condition must be satis�ed.
�pxkpBxX
Bk + �pykp
BykY
Bk = w0k
�Lk:
As has been mentioned above, the relationship between �pxk and �pyk in the Van Elswijk Planis determined by the tax burdens of the production sectors under the wage tax system. Thisrelationship has therefore to be determined using the prices and quantities in the wage taxsystem. This is carried out as follows: under the wage tax system, the producers in productionsector X in country k must pay the government a total of �wkw
Ak L
Axk, and those in production
sector Y must pay a total of �wkwAk L
Ayk, where L
Axk and L
Ayk denote employment in the respective
sector. Furthermore, the product prices are equal to pAx and pAyk and the quantities sold are
equal to XAk and Y A
k . Under the Van Elswijk system the production sectors would have topay �pxkp
AxX
Ak and �pykp
AykY
Ak given these prices and quantities and they would receive w0kL
Axk
and w0kLAyk subsidies, respectively. According to the Van Elswijk Plan, the tax burden for the
production sectors should remain the same at the time of the transition to the new tax system. Inother words they should be equal to: �opxkp
AxX
Ak �w0kL
Axk = �wkw
Ak L
Axk and �
opykp
AykY
Ak �w0kL
Ayk =
�wkwAk L
Ayk, where �
opxk and �
opyk are the resulting tax rates. The relationship which, according to
Van Elswijk should be maintained between the rates �pxk and �pyk is determined by the condition�pxk=�pyk = �opxk=�
opyk.
2.4 Foreign country
As has been indicated above, there is a home and a foreign economy, which are linked with eachother. Both economies comprise consumers and producers with the same characteristics. Theonly di�erence is formed by the fact that the foreign consumers possess a multiple of the numberof labour and capital units of domestic consumers, in order to approximate a small open domesticeconomy. The scaling factor for the production functions of the foreign production sectors(Azf ) is such that, theoretically, a symmetric equilibrium is possible. The foreign economy isthen merely a larger version of the domestic economy, if the same system of social insurance(tax system) operates in both countries.38 Furthermore, it must be emphasised that only theproducts of the relatively capital intensive production sector X are traded internationally. Inother words, the production sector X represents an open (exposed) sector, while the relativelylabour intensive production sector Y is a closed (sheltered) sector. Moreover, there is a di�erencein the mobility of the production factors. It is assumed that consumers can o�er their labourin their own country only, whereas they can sell their capital endowments to producers in boththeir own country as well as in the other one. Furthermore, it is assumed that the wage taxsystem is always operative in the foreign country.
2.5 Implementation and design of the experiment
In order to study the performance of the two di�erent systems of �nancing social insurance(the wage tax system and the Van Elswijk system) in a small open economy, a computerizedexperiment was designed in which the economy described above was implemented. A totalof six experimental sessions was run. The wage tax system was implemented in the home
38In the experiment the foreign economy is seven times as large, see 2.6.
16 CHAPTER 2. EXPERIMENTAL DESIGN
Table 2.1: Summary of the experimental sessions
date place # of subjects experience tax system # of trading periods(practice rounds)
081098 University of 16 training session wage tax 16Amsterdam closed economy (2)
081098 University of 16 training session wage tax 16Amsterdam closed economy (2)
091098 University of 16 training session wage tax 16Amsterdam closed economy (2)
221098 University of 16 training session Van Elswijk 16Amsterdam closed economy (2)
221098 University of 16 training session Van Elswijk 16Amsterdam closed economy (2)
231098 University of 16 training session Van Elswijk 16Amsterdam closed economy (2)
Note: In the foreign country always the wage tax system is in operation.
country in three of the sessions while the Van Elswijk System was implemented in the otherthree. In the foreign country the wage tax system was always applied. Each session tookplace in the CREED laboratory for experimental economic research, University of Amsterdam.Sixteen subjects participated in each session. Each subject only took part in the experimentonce. Therefore, no subjects took part in more then one experimental session. All subjectswere trained extensively on the required use of the computer and the trading rules before theyparticipated in the experiment. Moreover, they �rst participated in an experiment with a closedeconomy. The subjects were selected for the actual (open economy) experiment on the basisof their performance in the closed economy experiment. All earnings were paid out in cash(Dutch guilders), at the end of the experiment. A detailed description of the training, the closedeconomy experiment and the method of payment is given in Appendix A. The experimentalsessions carried out are shown in table 2.1 (an explanation of the last column is given below).At the beginning of each experimental session, the participants received the instructions aboutthe experiment. These consisted of a general part relevant to all subjects, irrespective of theirrole in the experiment (consumer or producer) and a part relevant only to either consumers orproducers. The general part was read aloud. All participants then read the second part on theirown. While reading, several questions had to be answered to ensure that the procedures, thecalculation of the earnings, and so forth had been understood. Furthermore, an explanationwas given of how to �ll in the personal registration form. The top of the form summarisedinformation relevant to the participant, such as the tax rate on the markets where they werealoud to trade and (for the consumers) the endowment of labour and capital. The participantshad to record their transactions and earnings on this form. The aim of this was to make themaware of the e�ects of their decisions. The information could also be found using the computer.Appendix B contains a copy of the instructions, including the forms used and the computerscreens seen by the participants.
Each experimental session consisted of 16 trading periods, preceded by 2 practice periods tofamiliarise the participants with the experiment and to check whether the personal registrationforms were �lled in properly. The earnings made during the practice periods were not paid out.
At the beginning of each trading period, the consumers were endowed with labour andcapital along with a certain amount of cash, while the producers received a certain amount of
2.5. IMPLEMENTATION AND DESIGN OF THE EXPERIMENT 17
cash as working capital. By using their working capital the producers could buy units of labourand capital from the consumers, determining their production of goods X and Y . Besides theearnings from selling labour and capital, the consumers also received a �xed bene�t for eachunsold unit of labour. Along with the cash, this income determined the total amount of moneythat the consumers had to buy products from the producers with. The earnings of consumerswere determined by the number of unsold labour units (leisure) and the number of bought unitsof the products X and Y via the utility function applicable to them.39The earnings of producerswere determined by their pro�t at the end of the trading period. Cash and working capital notspent, along with units of capital and products not sold, could not be transferred to the nexttrading period, and therefore did not yield any earnings for the participants.
This ensured that all trading periods were identical, which was an important factor in orderto be able to determine whether the economic process would stabilise. The only exceptionwas the tax rate. From trading period 9 onwards, the tax rate - imposed on producers - wasno longer kept constant. It was determined by the government's balance during the previoustrading period in such a way that the same economic behavior would lead to a balanced budget(see 2.6). This was introduced to prevent distortions of the performance of the two tax systemsas a result of a di�erent score for the government's balance. All participants knew that the taxrate could change from trading period 9 onwards.
Each trading period consisted of two phases. During the �rst phase the factor markets forlabour and capital were open. During this phase producers could buy units of labour and capitalfrom the consumers. At the end of this phase, \production" took place using these productionfactors. The number of products (X or Y ) which could be produced using the units of labour andcapital bought were determined by the computer using the relevant production function. Thenthe second phase started in which the producers could sell their products to the consumers onthe markets for the products X and Y . There are two reasons why this sequential procedure wasselected. Firstly, because this is similar to real life. Products cannot be produced until produc-tion factors have been acquired. Because of this sequentiality of decisions, which is inherent tothe production process, the precise conditions on the product market will not yet be known whentrading the production factors. This a�ects both the producers and the consumers (see 3.4).The second reason for this sequential process is that it prevents participants from having totrade on more than two markets at the same time. If all the markets were open simultaneously(simultaneous markets) the consumers would have to trade on four markets (for labour, capital,X and Y ) and the producers would have to trade on three markets (for labour, capital, X orY ). This would increase the complexity of the experimental economy considerably, with thepossible consequence of bad economic performance because of the complexity.40 Experimentalresearch has shown that experimental sequential markets do not necessarily operate worse thansimultaneous markets.41 Furthermore, it should be emphasised that this experimental study isaimed at comparing the economic e�ects of di�erent tax systems and not at how the economyoperates under a speci�c tax system.
Ten of the participants were producers and 6 were consumers in each experimental session.
39In the experiment the franc was used as currency. The earnings were converted from francs into guildersusing an exchange rate known to the participant. As has already been mentioned, the total earnings were paidin cash in guilders at the end of the experiment.
40The use of simultaneous markets means that producers and consumers receive information about the situationon the product markets sooner but, nevertheless, also means that production factors need to be traded �rst beforeproducts become available. Section 4.5 will deal with the uncertainty that is created as a result of this.
41See H.C. Quirmbach, C.W. Swenson and C.C. Vines (1996), An experimental examination of general equilib-rium tax incidence, Journal of Public Economics, 61, pp. 337-358.
18 CHAPTER 2. EXPERIMENTAL DESIGN
Table 2.2: Number of buyers and sellers on the markets
market buyers sellers
labor market home country 5 3
labor market foreign country 5 3
capital market (international) 10 6
market for good Y home country 3 3
market for good Y foreign country 3 3
market for good X (international) 6 4
Note: Consumers are sellers of labor and capital on the
factor markets and buyers of X and Y on the product
markets; producers are buyers of labor and capital
on the factor markets and sellers of X and Y on the
product markets.
There were 2 producers of X, 3 producers of Y and 3 consumers both in the home and theforeign country. Table 2.2 gives a summary of this. These numbers of agents are based onthe experimental �nding that it is suÆcient to have at least 3 economic agents on both sidesof a market to ensure competitiveness.42 Furthermore, limiting the number of participants asmuch as possible enhances experimental control.43 The market mechanism used was an adaptedcomputerised version of the so-called \multiple units double auction" procedure.44 The basicprinciple of this procedure is that buyers and sellers can propose bids and asks simultaneously,whereby only the highest bid and the lowest ask is shown on the screen in the experiment. Atransaction takes place when the bid is higher than or equal to the ask. Details concerning thetrading rules used can be found in Appendix A.The computer screen for the factor markets (phase 1 of the trading session) showed all therelevant information needed to buy and sell labour and capital. Consumers and producers sawthe current bids and asks on the capital market and the (domestic or foreign) labour marketrelevant to them. In a similar way, the computer screen for the product markets (phase 2 of thetrading session) showed the participants the going bids and asks on the markets on which theycould deal. The going bids and asks were given both as market prices (that is, not includingtaxes or subsidies) and inclusive prices (that is, including taxes or subsidies). Furthermore, therelevant price for the participant was highlighted. The market prices were always highlightedfor the consumers, because they did not have to pay taxes and received no subsidies (except forunemployment bene�ts). The inclusive wage (the market price for labour plus the wage tax) washighlighted for the producers operating within the wage tax system. For those operating in theVan Elswijk system the inclusive wage (the market price for labour minus the subsidy) and theinclusive product price (the market price for the product plus the product tax) were highlighted.Both computer screens also gave the tax rates relevant to the participants, their own inventoriesand sales, the cash available to them and the total earnings at the time. Between phases 1
42See C.N. Noussair, C.R. Plott and R.G. Riezman (1995), An experimental investigation of the patterns ofinternational trade, American Economic Review, 85, pp.462-291.
43The choice of creating a small domestic open economy with an equal number of economic agents in bothcountries meant that the actors abroad had to be considerably 'bigger', in other words many more units had to betraded. The advantage of this is that one is not confronted with the technical diÆculty to use a large number ofparticipants (there would have been 64 in this experiment). A disadvantage is that a foreign producer or consumeroperating badly can have a considerable economic e�ect.
44See C.R. Plott and Peter Gray (1990), The multiple unit double action, Journal of Economic Behaviour andOrganisation, 13, pp. 245-258.
2.6. PARAMETER CHOICE 19
and 2 the consumers received information about the total number of units of labour unsold andthe bene�ts involved, and the producers were informed about the number of products producedon the basis of the units of labour and capital bought. Furthermore, statistics concerning theamount of labour and capital traded were given along with average market prices and the averageprice that the participant received/paid for the units sold/bought. After the product marketshad closed the participants were shown a screen with similar market statistics for the productmarkets relevant to them. Finally, each participant could receive a summary of all the previoustransactions on the markets on which they could trade at any time during the experiment.Copies of the screens that the participants could view are given in Appendix A.
The time schedule for the activities in a trading period was as follows. Firstly, the factormarkets were opened for a period of 4 minutes and 30 seconds. Then the �rst informationscreen was shown for 20 seconds. Then the product markets were opened for 3 minutes andthirty seconds after which the trading session was closed and the second information screenwas shown for 2 minutes. This time schedule was based on experience gained during the twopilot sessions and the closed economy experiments. It appeared that this gave the participantssuÆcient time to trade all the units they wanted to trade and to �ll in their personal registrationforms.
2.6 Parameter choice
Choices had to be made concerning the values of the (exogenous) parameters for the experimentaleconomy. On an individual level, the choices concerned the production possibilities (productionfunctions), the consumer preferences (utility functions) the total number of units of labour andcapital available, the amount of cash of the consumers and the working capital of the producers.On a macro-economic level, the choices concerned the unemployment bene�t, the wage tax ratefor the wage tax system, and the subsidy of labour and the product tax rates for the VanElswijk system. Table 2.3 gives a summary of the production function parameters (see 2.2), ashort description of these parameters and the size of the working capital.
The parameter values of the production functions are based on the CPB's applied generalequilibrium model MIMIC for the Dutch economy, which includes the same type of productionfunction. The substitution elasticity in the relatively capital intensive sector X is equal to1=3, namely 1=(1 � x), and in the labour intensive sector Y it is equal to 1=7, namely 1=(1 � y). These values are somewhat higher than those used in MIMIC, in which, for example, thesubstitution elasticity between labour and capital in the sheltered sector (here Y ) is equal to0. In the experiment however, attention is mainly focused on long-run e�ects, therefore somesubstitution possibility must be taken into account. A similar argument can be applied tosector X. Qualitatively, the chosen values are in line with MIMIC. The substitution elasticityis small in both sectors and it is larger in the capital intensive exposed sector X, than in thelabour intensive sheltered sector Y . The labour intensity parameters (�) also have slightly highervalues than those used in MIMIC. This is mainly for methodological reasons. Otherwise, thedi�erence between the market wage and the capital price would be very large in a theoreticalequilibrium situation. Since typing large numbers is more time-consuming and can lead toerrors more easily, this should be avoided. Furthermore, large price di�erences could be morediÆcult to accept for the participants, which could slow down a possible convergence towards thetheoretical equilibrium. Therefore the choice was made to slightly increase the labour intensityin both sectors proportionally (from 3=4 to 9=16 in sector X and from 9=10 to 27=40 in sector
20 CHAPTER 2. EXPERIMENTAL DESIGN
Table 2.3: Parameter values concerning the producers
parameters description parameter values
production functions home country (k=h) foreign country (k=f)
Axk scaling factor sector X 1.21 1
�x labor intensity parameter 0.5625 0.5625sector X
x substitution elasticity -2 -2parameter sector X
Ayk scaling factor sector Y 1.21 1
�y labor intensity parameter 0.675 0.675sector Y
y substitution elasticity -6 -6parameter sector Y
� returns-to-scale parameter 0.9 0.9
working capital - X cash for X-producers 1223 8557for purchase oflabor and capital
working capital - Y cash for Y-producers 815 5705for purchase oflabor and capital
Y ) maintaining the same relative intensity as in MIMIC. As in MIMIC, both sectors are morelabour intensive than capital intensive (�x; �y > 1=2) and the sheltered sector Y is more labourintensive than the exposed sector X. In contrast to MIMIC, a \�xed factor" has been includedin the experimental economy. This is given as 1�� and is assumed to be equal to 0:1. A possibleinterpretation of this factor is that there always exists some immobile capital in the economy,for example, physical or social infrastructure or land as input in the production.45 As a resultof this the producers make a pro�t in a theoretical equilibrium. The presence of a (small) �xedfactor appears to be more realistic. Besides this, this factor has the methodological advantagethat, in the experiment, the producers can be given �nancial incentives in a natural way via thepro�ts made. As has already been mentioned, the experimental economy has been designed insuch a way that in a symmetric equilibrium (with the same taxation system both in the homeand the foreign country), the foreign economy will be a 'copy' of the home economy which isseven times as large. In order to be able to achieve this, the returns-to-scale parameter in theproduction functions for the foreign country (Azf ; z = x; y) has been set to 1:21 and that ofthe home country (Azh; z = x; y) has been set to 1.The size of the working capital which the producers in the experimental economy can use to buyproduction factors has been determined using the symmetrical theoretical equilibrium which willbe discussed in the following chapter. The 1 : 7 relationship between the home and the foreigncountry, whereby a small open economy is created in the home country, also applies here.For exactly the same reason, consumers abroad have been given seven times as much labour and
45In recent (empirical) literature the assumption of totally mobile capital is challenged. See, for example,M.S. Feldstein and C. Horioka (1980), Domestic savings and international capital ows, Economic Journal, 90,pp. 314-329, and R.H. Gordon and L. Bovenberg (1996), Why is capital so immobile internationally? Possibleexplanations and implications for capital income taxation, American Economic Review, 86, pp. 1057-1075. Asthe latter article demonstrates, this does not mean that taxing less mobile capital is always favourable.
2.6. PARAMETER CHOICE 21
Table 2.4: Parameter values concerning the consumers
parameters description parameter values
home country (k=h) foreign country (k=f)
C scaling factor (utility function) 25 25
a preference intensity for 1 1consumption good X(utility function)
b preference intensity for 1 1consumption good Y(utility function)
c preference intensity for 0.25 0.25leisure (utility function)
�Lkj number of tradeable units 15 105of labor per consumer
�Kkj number of tradeable units 10 70of capital per consumer
cash distributed pro�ts in the 181 1268theoretical mocel
Note: For the ease of exposition in the experiment the following log-linearized and rescaled version
of the utility function in 2.2 was used: C lnUkj = C(a lnXkj + b lnYkj + c ln(�Lkj � Lkj)):
Furthermore, utility was set to zero if either Xkj , Ykj or (�Lkj � Lkj) was zero.
capital in the experimental economy. The same applies to the cash available to the consumers.The amount was determined using the symmetrical theoretical equilibrium of the economicmodel to be discussed in the following chapter. In the theoretical model, cash is equal to theshare of the consumer in the �rms' pro�ts of the country concerned (�k=nck); see 2.2. Giving theconsumers the theoretical share of the pro�t as cash at the beginning of a trading period makes itpossible to reach the theoretical equilibrium in the experimental economy.46 The chosen valuesfor these amounts and other parameters of relevance for the consumers are given in table 2.4.The values of the preference intensity parameters a and b were chosen such that the two
consumption goods had the same e�ect on the utility of a consumer (and therefore on theirearnings in the experiment). The chosen value for the leisure parameter, c, ensures that, on theone hand, leisure is valued by the consumer - which leads to voluntary unemployment - and,on the other hand, that labour is supplied at a comparatively convenient wage rate (see theprevious methodological comment on the disadvantages of large price di�erences).
Finally, the parameters concerning the social insurance system will be dealt with. Theseconcern the unemployment bene�t, the labour subsidy (in the Van Elswijk system), the wagetax rate (in the wage tax system) and the sales tax rates (in the Van Elswijk system). The chosenvalues of the parameters are given in table 2.5. The chosen value of 70 for the unemploymentbene�t corresponds with a \replacement rate" of approximately 65% in the symmetric theoreticalequilibrium in the wage tax system. A �xed bene�t has been chosen mainly for methodologicalreasons. In this way, the design of each trading period remains identical - except for the changein the tax rate in the periods 9-16. The e�ects of a �xed replacement rate on the theoreticalresults will be discussed in 3.3.
46When determining an equilibrium of the theoretical model, all the equilibrium values for the variables, suchas the pro�t, are determined simultaneously. In a real economy, such as the experimental one, this is not the caseand therefore it cannot be implemented in the experiment.
22 CHAPTER 2. EXPERIMENTAL DESIGN
Table 2.5: Unemployment bene�t, employment subsidy, and tax rates
parameters description home country (k=h) foreign country (k=f)
wage tax Van Elswijk wage taxsystem system system
w0k unemployment bene�t for 70 70 70consumer for each unsold
unit of labor
�wk wage tax rate 0.3777 - 0.3777
�pxk sales tax rate - 0.6521 -good X
�pyk sales tax rate - 0.7518good Y
Note: As mentioned in 2.3 are the sales taxes in the home country under the Van Elswijk system
determined on the basis of the experimental results under the wage tax system. The tax
rates are chosen in a way such that the initial tax burden under the Van Elswijk system
are the same as under the wage tax system. For that the average tax burden in periods 6-8 is
chosen as the basis of calculation.
The wage tax rate of 0.3777 has also been derived from the symmetric theoretical equilibriumof the wage tax system. At this rate all the markets are in equilibrium and the budget isbalanced, as will be discussed in the following chapter. The note under the table has alreadymentioned how the rates are determined in the Van Elswijk system. These are based on theactual experimental outcome in the home country under the wage tax system, in accordancewith the Van Elswijk plan. The initial tax burden for the two production sectors will remain thesame under the chosen values, given the average economic situation in the trading periods 6-8under the wage tax system. In taking this average, the possible learning e�ects are taken intoaccount (by omitting the previous periods) as is the possible noise in the outcomes (by takingthe last three trading periods with a constant tax rate). In both tax systems the tax rates werekept at the same level during the �rst 8 trading periods. After this the rates were altered tosuch an extent that the budget would be balanced if the economic outcomes were the same asin the previous trading period. The exact way in which this occurs is described below.
Dynamic determination of the tax rates
The following shows how the new tax rate for each new trading period is determined.
Wage tax system
The rate � t+1wk in the trading period t+1 is equal to the unemployment expenditure dividedby the total wage sum in period t in country k:
� t+1wk = w0k(�Lk � Ltk)=w
tkL
tk:
Van Elswijk system
In the Van Elswijk system, the initial (�rst trading period) tax rates for the home market aredetermined using the experimental outcomes of the wage tax system (in the foreign country thewage tax system remains operative). If �0wh is the wage tax rate that ensures that the (average)
2.6. PARAMETER CHOICE 23
tax revenue is equal to the (average) unemployment expenditure in the last period(s) - withconstant tax rates - of the experiment with the wage tax system (denoted by the superscript A)then:
�wh0wAhL
Ah = w0h(�Lh � LA
h ):
When adjusting the tax rates an upper limit of 90% was maintained in both tax systems. Thishas been done for methodological reasons since the pilot sessions revealed that participants canbe discouraged to trade if tax rates become higher.
In the Van Elswijk system the initial tax rates, �0pzh and �0pyh should be such that the nettax burden of the production sectors remains the same, that is:
�0pxhpAxX
Ah � w0hL
Axh = �0whw
AhL
Axh and �0pyhp
AyhY
Ah � w0hL
Ayh = �0whw
Ah L
Ayh:
Given the value of �wh0 the values for �pxh0 and �pyh0 can be determined using these equations.
The rates are changed in the subsequent trading periods in accordance with the followingtwo conditions. Firstly, the sum of the unemployment bene�ts and the labour subsidies in periodt must be covered by tax income if behaviour is unchanged:
� t+1pxhptxX
th + � t+1pyh p
tyh = w0h
�Lh:
Secondly, the relationship between the two tax rates must be maintained:
� t+1pxh=�t+1pyh = �0pxh=�
0pyh:
The new tax rates, � t+1pxh and � t+1pyh , can be determined using these two equations.
24 CHAPTER 2. EXPERIMENTAL DESIGN
Part III
Results
25
Chapter 3
Theoretical equilibrium results
3.1 Introduction
The economic model presented in chapter 2, which forms the basis of this experiment, will �rstbe analysed theoretically. For that the following usual assumptions are made. Firstly, it isassumed that producers aim to maximise their pro�ts and that consumers aim for maximumutility. Furthermore, it is assumed that the economic process is in equilibrium, in other words,that supply and demand are equal on all markets and that the contributions received by thegovernment are equal to its expenditures (bene�ts and subsidies). It is important to rememberthat these assumptions do not necessarily hold in practise, and therefore this may also be the casein the experiment. However, they are important to be able to solve the model theoretically. Byusing the chosen values of the parameters (see 2.6) the solution can be determined numerically.The \theoretical equilibrium results" were obtained in this way.
The theoretical equilibrium results were determined for two reasons. Firstly, they are usedto implement the model in the experiment (see 2.5 and 2.6). Secondly, they may provide extrainformation or they may be used as a \benchmark" when evaluating the experimental results.
The results for both tax systems, the wage tax system (system A) and the Van Elswijksystem (system B), are presented in section 3.2 and are compared with one another. A sensitivityanalysis is given in 3.3, in which the e�ects of changes in the values of a number of importantparameters are studied. The chapter closes with a number of conclusions in 3.4.
3.2 Wage tax system vs Van Eslwijk system
Table 3.1 presents the theoretical equilibrium results for the most important economic variablesof both systems, namely: labour employment, employment of capital, production, product prices,market wages and capital price, utility and tax rates. All the prices are relative prices. Todetermine these, the nominal prices were divided by the sum of all nominal prices. In doing so,the experimental results can be compared with the theoretical equilibrium results in a meaningfulway (otherwise the general price level could cause distortions in the interpretation). The (small,open) home economy is dealt with in the top part of the table, while the (large, open) foreigneconomy is dealt with in the lower part of the table. Remember that in the Van Elswijk system,only in the home country the tax system is changed from a tax on wages to a tax on the valueadded (sales revenues) of the production sectors X and Y. In the foreign country always wages
27
28 CHAPTER 3. THEORETICAL EQUILIBRIUM RESULTS
are taxed. Furthermore, in accordance with the Van Elswijk Plan, the ratio between the taxrates for the production sectors is equal to the ratio which results if the tax burden on theproduction sectors is kept equal to the tax burden under the wage tax system, based on theeconomic outcomes of the wage tax system (see 2.5). In the Van Elswijk system, the total taxburden is always equal to the unemployment bene�t (= employment subsidy) times the total(maximum) amount of labour available. Finally, it should be noted that the variables above thesolid line in both parts of the table are exogenously given. These variables are the share of the
Table 3.1: Theoretical equilibrium results
home country wage tax system (A) Van Elswijk system (B)
-/- return on capital -/- 6.45 -/- 6.50 -/- 6.24 -/- 6.61
gross pro�t in X + Y (total) 18.46 18.47 17.60 16.30
-/- tax payment in X + Y (total) -/- 12.63 -/- 12.58 -/- 11.94 -/- 10.30
+/+ subsidy +/+ 0.00 +/+ 0.00 +/+ 0.00 +/+ 0.00
net pro�t in X + Y (total) 5.83 5.89 5.66 6.00
tax rates 0.3777 0.3713 0.3655 0.2769
�xed factor of capital (which determines the share of pro�ts in sales revenues, see 2.2), theunemployment bene�t, the subsidy and the budget surplus. The other variables are determinedendogenously by the model. For example, a condition imposed exogenously is that the budgetshould be balanced (a budget surplus equal to 0), but the market wage is determined by themodel itself.
Since a more detailed discussion of the speci�c results can best be made by comparing themwith the experimental results, we will only discuss the results in general here. Interestingly,there is more than one equilibrium. For each tax system two can be found. Potential reasons forthis will be discussed in the conclusions in 3.4. In the wage tax system (system A), the equilibria
30 CHAPTER 3. THEORETICAL EQUILIBRIUM RESULTS
given in the �rst two columns are very similar. Furthermore, one is symmetrical in the sensethat the foreign market is a 7 times larger copy of the domestic market (see the �rst column).In addition to the �gures in the table, it must be noted that the share of labour income is equalto 0.75 in the relatively capital intensive production sector X and equal to 0.83 in the relativelylabour intensive production sector Y in this equilibrium.
The second striking result is that the equilibria found in the Van Elswijk system (system B)are very di�erent; compare the last two columns. The equilibrium given in the last column showsa large \capital ight" and other unfavourable e�ects for the home economy when comparedwith the wage tax system (symmetric equilibrium). The amount of capital in the domesticmarket is 62% lower and the total level of employment is decreased by 36%. This results in thelevel of welfare (utility) of consumers dropping by 17%. However, the other equilibrium, givenin the last but one column, shows clear positive e�ects on the home economy. Even though thereis a loss of capital, which is now much smaller (7%), the total level of employment increases by17% and the level of consumer utility by 3%. Furthermore, both equilibria show a shift in salesrevenues to the advantage of the relatively labour intensive sector Y , but a shift in productionto the advantage of the relatively capital intensive sector X. Note that under the Van Elswijksystem the tax rate for Y is higher than that for X. This is a consequence of the condition ofthis system that the tax burden for the production sectors should remain the same when thesystem changes (that is, initially). The foreign economy is noticeably a�ected only in the caseof the equilibrium that unfavourably a�ects the home economy. As a result of the capital ightto the foreign country, the amount of capital and the total level of employment increase by 8%while the consumer utility increases by 1% in the foreign country.Apparently, theoretically the Van Elswijk system can have both positive and negative e�ects onthe home economy. This is the case even though the product tax in this system is an implicittax on capital and capital is totally mobile in the theoretical model. This can be explained usingthe theory of taxation.
Before discussing this it should be mentioned that the unemployment bene�t distorts theprice of leisure (as a result it decreases). Taxation to cover unemployment bene�ts leads tofurther distortions which e�ects are dependent on the tax system. In the case of a wage tax andunemployment bene�t, the relative price of labour in relation to capital will also be distorted.Both changes have a negative impact on employment. These distortions - and the consequentialloss in welfare (eÆciency costs) - become larger as the size of the bene�t increases.
In the Van Elswijk system, the employment subsidy ensures that the distortions in therelative price of labour in relation to capital decrease. The subsidy boosts the demand forlabour resulting in a positive e�ect on wages. The distortion in the price of leisure is alsodecreased in this way (indirectly). However, there are also two negative e�ects. Firstly, the taxtake must increase when changing to the Van Elswijk system in order to cover the subsidies.Secondly, the product tax, which is used to �nance bene�ts and subsidies, leads to an (implicit)taxation of capital. If capital were completely mobile, this would on balance (e�ectively) onlylead to a (higher) burden on labour. In that case the net rate of return on capital would bedetermined internationally, and this cannot be a�ected (or hardly) in the case of a small openeconomy. Since labour is assumed to be completely immobile, a tax on capital will only lead tothe exodus of capital until the net income per unit of this production factor in the home countryis again equal to the internationally determined level. However, the model includes a �xed factorwhich can be seen as immobile capital. The pro�ts made by �rms after expenditures for wagesand mobile capital (in the model, the input of capital) relate to this factor. The justi�cation forthe existence of this �xed factor is given in 2.6. The presence of this factor makes it possible to
3.3. SENSITIVITY ANALYSIS 31
shift part of the tax burden from labour to (the immobile part of) capital. Since according tothe theory of taxation, taxing a �xed factor does not lead to distortions (eÆciency costs), thiscan lead to higher welfare under the Van Elswijk system as long as the shift in tax burden islarge enough.
In order to study this further, a sensitivity analysis of the e�ects of changes in the values ofa number of important parameters will be carried out in the following section. Taking the aboveline of reasoning into consideration, the e�ect of a change in the size of the unemplyment bene�t(and the related employment subsidy in the Van Elswijk system) will �rst be discussed. Thiswill be followed by a discussion of the e�ect of the subsidy alone, and then by a discussion ofthe e�ects of a change in the size of the �xed factor. Finally a change in the model is discussedin which the nominally constant unemployment bene�t is replaced by an unemployment bene�trelated to the wage (a constant replacement rate).
3.3 Sensitivity analysis
For the sake of exposition, the discussion of the results of the sensitivity analysis will mainlyfocus on the e�ects on total employment, the total input of capital and the consumer utility(the most suitable measure of welfare in the theoretical model) in the home country.
E�ects of a change of the level of the unemployment bene�t
Tables 3.2 and 3.3 below show the e�ects of changing the level of the bene�t to zero and 35(instead of 70), respectively. When the bene�t is set at zero, and the subsidy in the Van Elswijksystem is thus also zero, only one equilibrium is found for both tax systems. Furthermore, thisequilibrium is the same in both systems. This is obvious since there will be no taxes if the bene�tis zero. In this case, the consumer utility will be highest, since taxes have a distortionary e�ectin the model. There is also just one equilibrium at a bene�t level of 35 (and a subsidy of 35for the Van Elswijk system), but this di�ers per system. Under the wage tax system, both thelevel of employment and the utility of consumers drop as a result of the tax. This also appliesto the Van Elswijk system in which some capital ight also takes place. Even so, this systemdoes better in terms of employment and welfare (utility) then the tax wage system, even at thislow level of unemployment bene�t.
When the bene�t is increased to 70, the levels of employment and welfare drop further underthe wage tax system (see table 3.1) while this is the case to a much lesser extent in the VanElswijk system, where capital ight actually decreases. That is, when the \favourable" equilib-rium (equilibrium 1) of the Van Elswijk system is studied. In the case of the \unfavourable"equilibrium (equilibrium 2) at this level of bene�t, the level of employment, the employment ofcapital and the level of welfare decrease considerably.
In order to gain a better picture of the e�ects of changing the level of the unemploymentbene�t, the level of the utility of consumers (the welfare level) is shown in �gure 3.1 for di�erentvalues of the bene�t. Since the utility level at a bene�t of 35 is very similar to the level at abene�t of 0 only higher levels of unemployment bene�t are studied, up to a level of 100.47 This�gure shows the following. Firstly, both systems have two equilibria only from a bene�t of 45and higher. In contrast to the Van Elswijk system, in the wage tax system a unique equilibriumis again found at a high bene�t level (100). Secondly, the utility level under the wage tax system
47Steps of 5 have been taken. The �gure shows interpolations between these steps.
32 CHAPTER 3. THEORETICAL EQUILIBRIUM RESULTS
60
70
80
90
100
110
120
35 40 45 50 55 60 65 70 75 80 85 90 95 100
unemployment benefit (employment subsidy)
cons
umer
util
ity
wage tax system equilibrium 1 wage tax system equilibrium 2
van Elswijk system equilibrium 1 van Elswijk system equilibrium 2
Figure 3.1: Consumer utility at di�erent levels of unemployment bene�ts (employment subsidies)
continues to decrease with an increase in bene�t in the symmetric equilibrium (equilibrium 1).Moreover, also the level of employment continues to decrease.48 However, in the asymmetricequilibrium (equilibrium 2), we see a continuous increase in utility, accompanied by an increasein labour and capital employment (there is actually some in ow of capital from the foreigncountry at a bene�t level of 95). The utility levels belonging to these equilibria intersect at abene�t of (approximately) 75. Such an intersection does not take place under the Van Elswijksystem. The utility level initially decreases slightly in the \favourable" equilibrium, after whichit gradually rises to a level equal to the level when the bene�t is 35 (but, of course, it remainsunder the highest level at a bene�t of zero). In the \unfavourable" equilibrium the utility levelinitially rises sharply after which it levels o�. The most remarkable result can be considered thefact that the Van Elswijk system always gives the highest level of welfare, whatever the bene�t(in the \favourable equilibrium"). Furthermore, this level hardly changes, even at bene�t levelsabove the benchmark model value of 70; it even increases then (at bene�ts higher than 90,accompanied by some in ow of capital from the foreign country). On the other hand, theperformance of the wage tax system decreases as the bene�t levels increase (in the symmetricequilibrium) and approaches the \unfavourable" equilibrium of the Van Elswijk system.
48The employment of capital stays always constant (30) in the symmetric equilibrium.
3.3. SENSITIVITY ANALYSIS 33
Table 3.2: Theoretical results with no unemployment bene�tand no employment subsidy
home country wage tax system (A)
equilibrium 1
share of �xed factor (%) 10
unemployment bene�t 0
employment subsidy 0
budget surplus 0
budget surplus/GNP 0
employment in X 17.6
employment in Y 20.5
employment in X + Y (total) 38.1
capital in X 18.7
capital in Y 11.3
capital in X + Y (total) 30.0
production in X 26.3
production in Y 24.8
production in X + Y (total) 51.1
price of good X 0.1853
price of good Y 0.1966
net wage 0.1753
capital price 0.0689
utility of a consumer 112.3
revenue in X + Y (total) 9.76
-/- wage sum -/- 6.71
-/- return on capital -/- 2.07
gross pro�t in X + Y (total) 0.98
-/- tax payment in X + Y (total) -/- 0.00
+/+ subsidy +/+ 0.00
net pro�t in X + Y (total) 0.98
tax rates 0.0000
34 CHAPTER 3. THEORETICAL EQUILIBRIUM RESULTS
Table 3.2: (continued)
foreign country wage tax system (A)
equilibrium 1
share of �xed factor (%) 10
unemployment bene�t 0
employment subsidy 0
budget surplus 0
budget surplus/GNP 0
employment in X 123.1
employment in Y 143.5
employment in X + Y (total) 266.6
capital in X 131.0
capital in Y 79.0
capital in X + Y (total) 210.0
production in X 184.2
production in Y 173.7
production in X + Y (total) 357.9
price of good X 0.1853
price of good Y 0.1966
net wage 0.1763
capital price 0.0689
utility of a consumer 221.8
revenue in X + Y (total) 68.29
-/- wage sum -/- 47.00
-/- return on capital -/- 14.46
gross pro�t in X + Y (total) 6.83
-/- tax payment in X + Y (total) -/- 0.00
+/+ subsidy +/+ 0.00
net pro�t in X + Y (total) 6.83
tax rates 0.0000
3.3. SENSITIVITY ANALYSIS 35
Table 3.3: Theoretical results with unemployment bene�tand employment subsidy of 35
home country wage tax system (A) Van Elswijk system (B)
equilibrium 1 equilibrium 1
share of �xed factor (%) 10 10
unemployment bene�t 35 35
employment subsidy 0 35
budget surplus 0 0
budget surplus/GNP 0 0
employment in X 15.4 16.3
employment in Y 17.5 18.1
employment in X+Y (total) 32.9 34.4
capital in X 19.6 17.0
capital in Y 10.4 9.9
capital in X+Y (total) 30.0 26.9
production in X 24.3 24.5
production in Y 21.7 22.2
production in X+Y (total) 46.0 46.7
price of good in X 0.1861 0.1822
price of good in Y 0.2084 0.2074
net wage 0.1751 0.1876
capital price 0.0467 0.0451
utility of a consumer 110.6 111.1
revenue in X+Y (total) 9.06 9.05
-/- wage sum -/- 5.76 -/- 6.46
-/- return on capital -/- 1.40 -/- 1.21
gross pro�t in X+Y (total) 1.90 1.38
-/- tax payment in X+Y (total) -/- 0.99 -/- 3.55
+/+ subsidy +/+ 0.00 +/+ 2.72
net pro�t in X+Y (total) 0.91 0.55
tax rates 0.1724 in X:0.3669in Y:0.4169
36 CHAPTER 3. THEORETICAL EQUILIBRIUM RESULTS
Table 3.3: (continued)
foreign country wage tax system (A) Van Elswijk system (B)
equilibrium 1 equilibrium 1
share of �xed factor (%) 10 10
unemployment bene�t 35 35
employment subsidy 0 0
budget surplus 0 0
budget surplus/GNP 0 0
employment in X 107.7 109.0
employment in Y 122.4 123.2
employment in X+Y (total) 230.1 232.2
capital in X 137.2 139.6
capital in Y 72.8 73.5
capital in X+Y (total) 210.0 213.1
production in X 170.3 172.4
production in Y 152.1 153.1
production in X+Y (total) 322.4 325.5
price of good in X 0.1861 0.1822
price of good in Y 0.2084 0.2043
net wage 0.1751 0.1734
capital price 0.0467 0.0451
utility of a consumer 220.0 220.2
revenue in X+Y (total) 63.40 62.67
-/- wage sum -/- 40.30 -/- 40.26
-/- return on capital -/- 9.82 -/- 9.61
gross pro�t in X+Y (total) 13.28 12.80
-/- tax payment in X+Y (total) -/- 6.95 -/- 6.53
+/+ subsidy +/+ 0.00 +/+ 0.00
net pro�t in X+Y (total) 6.33 6.27
tax rates 0.1724 0.1623
3.3. SENSITIVITY ANALYSIS 37
The e�ect of the subsidy in the Van Elswijk system
In order to gain insight about the e�ect of the employment subsidy in the Van Elswijksystem, its level was set at zero, whereas the unemployment bene�t was held at 70. Theresults are shown in table 3.4. As would be expected, withdrawing the subsidy has a negativee�ect on employment in the \favourable" equilibrium under the Van Elswijk system. In the\unfavourable" equilibrium however, employment is a�ected positively.
Table 3.4: Theoretical results with no subsidy
home country wage tax system (A) Van Elswijk system (B)
-/- return on capital -/- 6.45 -/- 6.50 -/- 6.78 -/- 6.31
gross pro�t in X + Y (total) 18.46 18.47 18.00 18.24
-/- tax payment in X + Y (total) -/- 12.63 -/- 12.58 -/- 11.87 -/- 12.53
+/+ subsidy +/+ 0.00 +/+ 0.00 +/+ 0.00 +/+ 0.00
net pro�t in X + Y (total) 5.83 5.89 6.13 5.71
tax rates 0.3777 0.3713 0.3248 0.3844
The same applies to the level of welfare (utility of consumers). In other words, the twoequilibria approach each other. If these results are compared to those with a subsidy (see table3.1), it appears that 44% of the increase in employment in the \favourable" equilibrium in theVan Elswijk system, as opposed to the wage tax system, is due to the subsidy. Moreover, thispositive e�ect is mainly due to the increase in eÆciency which is achieved by the implicit taxationof capital (the �xed factor) under this system. This can be seen even better when looking atthe employment of capital. It is even higher in the \favourable" equilibrium without subsidythan in the wage tax system, and exhibits some in ow from the foreign country. Despite thepositive e�ects on labour employment, the subsidy leads to a decrease in the employment ofcapital, because of the higher tax rates that accompany a subsidy.
3.3. SENSITIVITY ANALYSIS 39
The following sensitivity analysis concerning the �xed factor is interesting, because of theapparent importance of an eÆciency gain under the Van Elswijk system.
The e�ect of the magnitude of the �xed factor
Tables 3.5 and 3.6 give the results with a larger (20%) and a smaller (5%) �xed factor, incomparison with the results with a �xed factor of 10% in the benchmark model (see table 3.1).
Table 3.5: Theoretical results with a �xed factor of 20%
home country wage tax system (A) Van Elswijk system (B)
-/- return on capital -/- 4.94 -/- 5.08 -/- 4.54 -/- 4.75
gross pro�t in X + Y (total) 20.20 20.23 16.52 19.46
-/- tax payment in X + Y (total) -/- 9.87 -/- 9.59 -/- 6.86 -/- 9.50
+/+ subsidy +/+ 0.00 +/+ 0.00 +/+ 0.00 +/+ 0.00
net pro�t in X + Y (total) 10.33 10.64 9.66 9.96
tax rates 0.3721 0.3440 0.2516 0.3717
When studying the results it has to be mentioned that decreasing the size of the �xed factorhas always a positive e�ect on production at given employment levels of labour and capital,while pro�ts always decrease (this follows from the production function; see 2.2). This also hasa positive e�ect on consumer utility since the consumption of goods can increase. On the otherhand, this leads to less room for shifting the tax burden towards the �xed factor (eÆciency gain).Therefore, we �nd that a decrease in the size of the �xed factor leads to a decrease of capitalemployment in the home country in the \favourable" equilibrium under the Van Elswijk system.There is more capital ight. This also has a negative e�ect on labour employment. When theequilibria under this tax system are compared with those under the wage tax system, we seethat the labour employment levels in the equilibria under the Van Elswijk system come closer
3.3. SENSITIVITY ANALYSIS 41
to those under the wage tax system. However, even when the �xed factor is halved (from 10%in to 5%), the levels of labour employment and welfare in the \favourable" equilibrium underthe Van Elswijk system are still (clearly) higher than the corresponding levels under the wagetax system. Obviously, this factor can be considerably decreased without cancelling the positivee�ects of the Van Elswijk system.
Table 3.6: Theoretical results with a �xed factor of 5%
home country wage tax system (A) Van Elswijk system (B)
-/- return on capital -/- 7.26 -/- 7.21 -/- 7.10 -/- 7.55
gross pro�t in X + Y (total) 17.17 17.19 16.19 15.29
-/- tax payment in X + Y (total) -/- 14.10 -/- 14.12 -/- 13.19 -/- 12.10
+/+ subsidy +/+ 0.00 +/+ 0.00 +/+ 0.00 +/+ 0.00
net pro�t in X + Y (total) 3.07 3.07 3.00 3.19
tax rates 0.3801 0.3840 0.3585 0.2940
3.3. SENSITIVITY ANALYSIS 43
The e�ects of a constant relation of unemployment bene�ts to wages
Finally, the e�ects of a constant \replacement rate" are studied, in which the relation ofbene�ts to wages is kept at 70%. The results are given in table 3.7. The Van Elswijk systemappears to operate better in terms of employment and welfare under the only equilibrium found,although capital ight again occurs. However, the improvement in performance in comparisonto the wage tax system is smaller than when the bene�t is nominally �xed at 70 (see table 3.1).This is because with a �xed nominal bene�t, the ratio of bene�ts to wages decreases under theVan Elswijk system (due to the employment subsidy), in comparison with the wage tax system.
Table 3.7: Theoretical results with an unemployment bene�tand employment subsidy of 70% of the wage
home country wage tax system (A) Van Elswijk system (B)
equilibrium 1 equilibrium 1
share of �xed factor (%) 10 10
unemployment bene�t (% of wage) 70 70
employment subsidy (% of wage) 0 70
budget surplus 0 0
budget surplus/GNP 0 0
employment in X 12.0 13.7
employment in Y 13.1 13.1
employment in X+Y (total) 25.1 26.8
capital in X 21.1 15.1
capital in Y 8.9 7.3
capital in X+Y (total) 30.0 22.4
production in X 20.5 21.2
production in Y 16.9 16.6
production in X+Y (total) 37.4 37.8
price of good in X 0.1905 0.1811
price of good in Y 0.2309 0.2399
net wage 0.1629 0.1815
capital price 0.0220 0.0192
utility of a consumer 103.0 103.8
revenue in X+Y (total) 7.80 7.81
-/- wage sum -/- 4.08 -/- 4.85
-/- return on capital -/- 0.66 -/- 0.43
gross pro�t in X+Y (total) 3.06 2.53
-/- tax payment in X+Y (total) -/- 2.27 -/- 5.72
+/+ subsidy +/+ 0.00 +/+ 3.40
net pro�t in X+Y (total) 0.79 0.21
tax rates 0.5570 in X:0.7006in Y:0.7622
44 CHAPTER 3. THEORETICAL EQUILIBRIUM RESULTS
Table 3.7: (continued)
foreign country wage tax system (A) Van Elswijk system (B)
equilibrium 1 equilibrium 1
share of �xed factor (%) 10 10
unemployment bene�t (% of wage) 70 70
employment subsidy (% of wage) 0 0
budget surplus 0 0
budget surplus/GNP 0 0
employment in X 84.0 85.0
employment in Y 91.4 91.6
employment in X+Y (total) 175.4 176.6
capital in X 147.4 154.3
capital in Y 62.4 63.4
capital in X+Y (total) 209.8 217.7
production in X 143.3 145.3
production in Y 118.2 118.5
production in X+Y (total) 261.5 263.8
price of good in X 0.1905 0.1811
price of good in Y 0.2309 0.2209
net wage 0.1629 0.1574
capital price 0.0220 0.0192
utility of a consumer 221.5 212.7
revenue in X+Y (total) 54.57 52.48
-/- wage sum -/- 28.58 -/- 27.79
-/- return on capital -/- 4.62 -/- 4.17
gross pro�t in X+Y (total) 21.37 20.52
-/- tax payment in X+Y (total) -/- 15.91 -/- 15.26
+/+ subsidy +/+ 0.00 +/+ 0.00
net pro�t in X+Y (total) 5.46 5.26
tax rates 0.5570 0.5492
3.4. CONCLUSIONS AND FINAL REMARKS 45
3.4 Conclusions and �nal remarks
The following conclusions can be drawn on the basis of the theoretical analysis above. The term\basic model" will be used when discussing the results concerning the parameter values usedin the experiment. For the sake of presentation, the conclusions will concentrate on the e�ectsof the tax systems on labour employment, the employment of capital and the level of welfare(consumer utility).
1. In the basic model, two equilibria are found for both tax systems. There is a symmetricequilibrium (where both the home and the foreign economy show the same results, whentaking the scaling factor into account) and an almost identical asymmetric equilibriumunder the wage tax system. Under the Van Elswijk system there is a \favourable" equi-librium in terms of labour employment and welfare (consumer utility) but there is also an\unfavourable" equilibrium.
2. Despite the occurrence of capital ight, the \favourable" equilibrium leads to a clearincrease of labour employment and welfare under the Van Elswijk system in comparisonwith the wage tax system. The level of labour employment increases by 17%. However,labour employment and welfare decrease considerably in the \unfavourable" equilibrium,as the result of a substantial capital ight. In this case, employment decreases by 36% incomparison with the wage tax system.
3. An increase in the level of the unemployment bene�t improves the performance of the VanElswijk system in comparison to the symmetric equilibrium under the wage tax system.The symmetric equilibrium appears to be the only remaining equilibrium at the highestlevel of unemployment bene�t studied (100). The Van Elswijk system always leads to thehighest levels of labour employment and welfare (in the \favourable" equilibrium) at allbene�t levels studied. Labour employment and welfare actually increase and there is anin ow of capital from the foreign country at higher bene�t levels than those in the basicmodel (70).
4. The positive e�ects of the Van Elswijk system seem to result mainly from the eÆciencygain achieved by the implicit taxation of capital (the �xed factor) under this system ratherthan from the subsidy. Even though the subsidy has a positive e�ect on employment, ithas a substantially negative e�ect on capital employment. This is because of the highertax rates the subsidy is accompanied with.
5. When the size of the �xed factor is reduced this leads to a decrease of capital employmentin the \favourable" equilibrium under the Van Elswijk system and has a slightly negativee�ect on labour employment. However, even when this factor is halved (from 10% inthe basic model to 5%), labour employment and welfare are still clearly higher than thecorresponding levels under the wage tax system. Evidently, this factor can be decreasedconsiderably without cancelling the positive e�ects of the Van Elswijk system.
6. A constant relation of bene�ts to wages (of 70%) rather than a �xed nominal bene�t,decreases the positive e�ects of the Van Elswijk system but still gives higher levels oflabour employment and welfare, in comparison to the wage tax system.
All things considered the Van Elswijk system - in the \favourable" equilibrium of this system -appears to perform markedly and systematically (when parameters are changed) better than the
46 CHAPTER 3. THEORETICAL EQUILIBRIUM RESULTS
wage tax system, in particular when the unemployment bene�t increases. However, the resultsin the \unfavourable" equilibrium are clearly worse, even though they improve at higher levelsof the unemployment bene�t.49
The reasons for these favourable results for the Van Elswijk system are discussed in 3.2.Particularly, the presence of a �xed factor (immobile capital) seems to play an important rolebesides the employment subsidy. It is diÆcult to determine what the importance of this factoris in reality. However, it can be noted that recent studies question that there is a large degree ofcapital mobility.50 Furthermore, we have seen that the relative performance of the Van Elswijksystem is not greatly a�ected by changes in the assumed size of the �xed factor.
The chapter will be concluded with some remarks about the existence of multiple equilibriaand the relationship between the theoretical model studied here and the experiment.
From general equilibrium theory we know that, in general, multiple equilibria can be ex-pected.51 Only under strict conditions there will be a unique equilibrium. Little more can besaid on this matter. From the literature, it can be concluded that CES production functions witha small substitution elasticity between labour and capital, as have been used in the model (inline with the CPB's MIMIC model), contribute to the occurrence of multiple equilibria.52 Thisperhaps also applies to the presence of a �xed bene�t. A certain \in exibility" is introduced inthe model in both cases.The role of the theoretical results in this experimental study of the Van Elswijk system wasdiscussed in the introduction to this chapter. On the one hand, they are used for the imple-mentation of the economic model in the experiment. On the other hand, they can be used as asource of information or as a \benchmark" when assessing the experimental results. We wouldlike to add the following comments. Firstly, a choice between the two equilibria found in thewage tax system has to be made in order to implement the model (see 3.1). The symmetricequilibrium was chosen in line with the design of the experiment and because the two equilibriafound barely di�er. Moreover, it is the better of the two equilibria regarding the wage tax sys-tem. The symmetric equilibrium results are shown in the �rst column of table 3.1. Regardingthe theoretical model as a source of information, it must be emphasised that the model assumesthat consumers and producers behave in a paricular way and that the economic process is inequilibrium. However, in the experiment, the participants are free to behave as they choose.Indeed, an important reason for carrying out the experiment is to acquire information on thebehavioural e�ects that can emerge under the Van Elswijk system. Furthermore, time playsan important role in the experiment, as it does in real life. For example, goods can only beproduced after labour and capital have been employed. In other words, one can speak of aninherent sequence of events in the economic process. This sequence and the dynamics relatedto it are expressed in the experiment but not in the theoretical model. Furthermore, it is notclear how these dynamics could be included in the theoretical model on the basis of existingknowledge. The experimental results could contribute to this knowledge and thus aid future
49As an addational information it is noted that the positive e�ects of the Van Elswijk system increase consid-erably if the system is also introduced abroad. In that case, a symmetric equilibrium without capital ight isobtained in the basic model.
50See Feldstein and Horioka (op.cit.), Gordon and Bovenberg (op.cit.).51See for example A. Mas-Colell, M. D. Whinston, and J. R. Green (1995), Micro-economic Theory, Oxford
University Press, New York and Oxford, chapter 17. According to the theory it will be an odd number. That twoare found in the basic model rather than three can be explained using the theoretical �nding that one of thesethree equilibria will be unstable.
52See K. Hildenbrand (1992), Computersimulierte Gleichgewichtsanalyse: Das 2x2x2x2 Modell, DiscussionPaper No. A-363, Universit�at Bonn.
3.4. CONCLUSIONS AND FINAL REMARKS 47
theoretical research. We will return to this subject in the discussion of the experimental resultsin 4.5. The experiment can also be useful with respect to equilibrium selection. In the caseof multiple equilibria, theory does not provide an indication of which to choose. This is ofimportance because under the Van Elswijk system there is not only an equilibrium found thatis better (in terms of labour employment and welfare) than the (symmetric) equilibrium underthe wage tax system; but also one that is worse. The experiment can provide information aboutthe direction in which the economic process will be pushed.
48 CHAPTER 3. THEORETICAL EQUILIBRIUM RESULTS
Chapter 4
Experimental results
4.1 Introduction
In this chapter, the results of the experiments with the wage tax system and the Van Elswijksystem will be presented and compared with each other. Furthermore, the experimental resultswill be compared to the theoretical equilibrium results in the previous chapter. In order to beable to make these comparisons, it is important to remember the set-up of the experimentalsessions (3 per tax system).
As was explained in 2.6, the theoretical equilibrium results of the wage tax system were usedin the implementation of the model in the experiment.53 In the experiment, consumers andproducers (in each of the 16 trading periods) receive a �xed amount of cash to trade with. Sincein the theoretical model the pro�ts are paid to the consumers, the theoretical equilibrium pro�tsof the wage tax system are assigned to the consumers in the experiment as cash. Furthermore,the amount that the producers spend on acquiring production factors (labour and capital) inthis equilibrium is assigned to the producers in the experiment as working capital. Finally, inthe experimental sessions using the wage tax system, the equilibrium wage tax rate is used asthe tax rate for the �rst 8 trading periods, in which it is kept constant. The only di�erencein the Van Elswijk system is that the sales tax rates in the �rst 8 periods are based on theexperimental results of the wage tax system. In the later trading periods (9-16), the tax rateis adjusted to the budget surplus (de�cit) in the previous trading period in both tax systems.This occurs in such a way that, given the economic outcomes of that trading period, the budgetin the previous trading period would have been balanced (except if the maximum rate of 90% isreached). Notice, that the budget surplus is therefore determined by the economic process itself(endogenous) in the experiment, given the tax rates.
Since in the experiment it is only with the wage tax system that the tax rate is at thetheoretical equilibrium level for the �rst 8 periods, the experimental results and the theoreticalequilibrium results can only be compared for these trading periods for this tax system. In theVan Elswijk system such a comparison is only meaningful for trading periods 9-16, when thetax rates can adjust.
Therefore, there will be a separate discussion of the results for trading periods 1-8 and tradingperiods 9-16. The results are presented in three di�erent ways. Firstly, a complete summaryconcerning all trading periods and experimental sessions is given in Appendix C. Secondly,
53As was discussed in 3.4, the symmetrical equilibrium is used for this.
49
50 CHAPTER 4. EXPERIMENTAL RESULTS
tables are presented in the main text showing the results of the three experimental sessionsper tax system for the same variables as in the theoretical analysis. The sessions for a speci�ctax system di�er only in that di�erent subjects participated. An experimental design withmore than one experimental session was chosen in order to be able to study the robustness ofthe results. It cannot be expected that there will be no variation (noise) in such a complexeconomic environment as the one studied here. Since variation will also occur as a result of theparticipants getting used to the experimental environment in the �rst trading periods, the tableswill only give the averages for sessions 5-8 and 13-16. In order to optimise the presentation ofthe evolution of the economic process, �gures can be found in the main text which, in additionto the 'straight scores' presented in C, give the average results per trading period and tax systemfor the di�erent economic variables.
Firstly, the results of the wage tax system will be presented in 4.2. These will be followed by adiscussion of the results of the Van Elswijk system in 4.3. A test of the hypotheses concerning theresults of the experiment, which were formulated by Van Elswijk and the Supervisory Committee,is given in 4.4. A discussion of the results is given in 4.5, thereafter in section 4.6 some conclusionsare drawn. The tables and �gures referred to in this chapter can be found in section 4.7.
4.2 Wage tax system
In tables 4.1 and 4.2 the experimental results per session averaged over trading periods 5-8and 13-16, respectively, are given for the wage tax system, together with a repetition of thetheoretical results. Since the results for the foreign country are similar, we will concentrate onthe results for the home country. The results will be compared with the symmetric theoreticalequilibrium, which is given in the �rst column of tables 4.1 and 4.2. As expected, there isvariation in the results of the di�erent experimental sessions. This is particularly the case at theproduction sector level and for the later trading periods. On the aggregated level the averages(over sessions) of the real variables (total level of labour employment, capital employment andproduction) are more or less close to the theoretical equilibrium, at least in trading periods 5-8.This also applies to the prices, with the exception of the capital price, which is much lowerthan in theory (approximately half the size). Mainly as a result of this, the net pro�t is onaverage higher. The consumer utility is lower, because of the deviations from the theoreticalproduction levels and the level of labour employment. Besides the relatively low capital pricethe (on average) considerable budget de�cit is also striking. This de�cit, which is zero in thetheoretical equilibrium, is on average 8% of the GNP during trading periods 5-8. Even duringperiods 13-16, when the tax rate is adjusted according to the de�cit in the previous period, thede�cit is still considerable (approximately 7%). This explains why (on average) considerablyhigher tax rates are found during these trading periods than in the theoretical equilibrium. Thisappears to have a pronounced negative e�ect on the employment of labour and thus production.Compared to trading periods 5-8, the deviations from the theoretical equilibrium results areconsiderably larger for these variables (approximately 14% instead of less than 5%).
Figures 4.1 to 4.4 give an impression of the evolution of the economic process under thewage tax system. These show the developments of quantities and prices of the inputs (labourand capital) and outputs (X and Y) in the home country and the foreign country for all tradingperiods.54 In addition to this, the change in the budget surplus and the tax rates in bothcountries are shown in �gure 4.5.
54Since, most of the time products were sold out completely only the quatities consumed are shown.
4.3. VAN ELSWIJK SYSTEM 51
The results demonstrate that there is a clear structure in the economic process. The quan-titative variables show a striking trend towards the theoretical equilibrium results on both thehome and foreign markets during the �rst 8 trading periods, when the tax rate is kept constantat the theoretical equilibrium level. This conclusion is supported by the convergence analysisgiven in Appendix D.55 Furthermore, it holds on the average (across periods) that the size of theforeign economy in terms of inputs and outputs does not di�er signi�cantly - with one exception- from the theoretical equilibrium result of being 7 times the size of the home economy.56
However, the development of the relative prices and the budget surplus is di�erent. Themajority of prices, including all the prices of production factors, do not approach the theoreticalequilibrium results in trading periods 1-8 (see �gures 4.2 and 4.4). This is also shown by theconvergence analysis in Appendix D. The general impression is that the prices of the productionfactors are lower - especially the capital price - and those of the products are higher, especiallywhen the later sessions are included. It seems that this is connected with the observation thatthe traded quantities of labour, capital and consumption goods tend to fall below the theoreticalequilibrium level (see �gures 4.1 and 4.3). A low employment level of production factors restrictsproduction. The �rst leads to a downward pressure on the factor prices, the latter to an upwardpush of the product prices. An explanation of these e�ects will be given in 4.5.
When table 4.1 was discussed, it was already mentioned that the wage tax system leads to aconsiderable budget de�cit. Figure 4.5 shows that this is a structural de�cit rather than beinglimited to the trading periods given in the table. The average de�cit however, does decreaseduring periods 9-16, when the tax rate is adjusted. It must be emphasised that changes in thetax rate are based on the de�cit in the previous period and that there is a maximum rate of90%. The tax rates in both countries increase considerably in period 9 as a result of the de�cit.This gives a shock to the economy from which it does not seem to be able to recover properly,with the exception of capital employment, as the already mentioned �gures show.
4.3 Van Elswijk system
Along with a repetition of the theoretical results tables 4.3 and 4.4 show the average resultsper experimental session for trading periods 5-8 and 13-16 for the Van Elswijk system,. Theinitial tax rates for this tax system are determined using the average experimental results fortrading periods 6-8 in the wage tax system. As was also the case in the wage tax system, the taxrates were kept constant during trading periods 1-8. The condition that the tax burden on theproduction sectors should initially remain the same when changing to the Van Elswijk systemhas led to the rate of 65% on the price of product X and that of 75% on the price of productY in the home country. The wage tax system continues to apply in the foreign country, with arate of 38% on wages. Since these rates are not the same as the theoretical equilibria found -see the �rst two columns of the tables - a direct comparison with the theoretical results in thewage tax system cannot be made for periods 1-8. Therefore, we will limit our comparison tothe experimental results of the Van Elswijk system and the wage tax system for these periods.Tables 4.5 and 4.6 enable this comparison to be made.
55The convergence analysis sometimes gives larger di�erences on the non-aggregated level, but this involves avery strict assessment, given the complexity of the economic system. The hypotheses being tested in this study(4.4) do not relate to this.
56A 2-samples t-test with equal variances is used where the home country sample consists of the relevant unitof observation times 7. As the supply of capital is concerned, it was tested whether the di�erence between thedomestic number of units (times 7) and the foreign number of units deviated signi�cantly from 0.
52 CHAPTER 4. EXPERIMENTAL RESULTS
When compared with the wage tax system, table 4.5 shows that in the Van Elswijk system,the economic activity in the foreign country is clearly lower (approximately 25%) on averageover periods 5-8. This is mainly due to the much lower level of employment and input of capitalin session 2 (see Appendix C).57 This mainly a�ects the production of the international goodX, which leads to the high relative price of this product in this session. Also apart from thisexperimental session, the average level of activity is lower in these trading periods (approximately15%). As will be shown, there was a low level of activity from the �rst trading periods onwardswhich also a�ected the home country. Interestingly, the average level of activity in the homeeconomy was already higher in trading periods 5-8 than that in the wage tax system (this alsoholds for experimental session 2). Relatively high product prices and the taxes related to themcaused the net pro�t to be lower at home. The exception is experimental session 2, in whichthe relatively high price of the international good X makes the net pro�t increase as a result ofits limited supply. The higher net pro�t in the foreign country during this session is also partlycaused by this e�ect. During the other experimental sessions, the lower wage and lower taxescompensated the limited revenues. The markedly low level of utility in experimental session 2,both in the home and the foreign country, is also related to the limited production of X in thissession (a considerable amount of the home country production is sold in the foreign country;see Appendix C). The lower level of economic activity, particularly the employment of capital,seems to be an important explanation of the low capital price in relation to the wage tax system.Finally, it is notable that the experimental sessions with the Van Elswijk system demonstrateconsiderable budget surpluses in the home country (more than 10% GNP) during these tradingperiods, while the wage tax system in the foreign country again shows a noticeable de�cit (ofmore than 10% GNP).
In the following discussion of the experimental results during trading periods 13 to 16, itis possible to make a comparison with the theoretical equilibrium results found (see table 4.4).Once again experimental session 2 is di�erent from the others, although this is somewhat less pro-nounced than in the �rst trading periods. Even when this experimental session is included in thecomparison, the results for the home country show remarkable similarities with the \favourable"theoretical equilibrium, given in the �rst column of table 4.4. These similarities include, for ex-ample, the balanced budget, the tax rates (which are adjusted to the level of the budget in theprevious trading period during these periods) the level of employment and production; the lattertwo variables even are approximately 10% higher. Besides the high relative price of product Xin the deviant experimental session 2, the most striking results in this comparison are the muchhigher average employment of capital (approximately 60%) and the much lower capital price(1/40 of the theoretical value). Nevertheless, home country consumer utility drops because asigni�cant part of the production of X is sold in the foreign country, as a result of the smallerlevel of production there (resulting in a relatively high price for this product). As we have al-ready noted in the discussion of the wage tax system, the economy of the foreign country (witha wage tax system in operation) performs worse than in the theoretical equilibrium. Moreover,the high level of activity in the home country means that capital is imported from the foreigncountry as a result of which the level of employment and production being even more negativelyin uenced. It is interesting that the results for the foreign country under the Van Elswijk systembecome more or less the same as the results under the wage tax system (in both countries) ifthis limited employment of capital is taken into account and experimental session 2 is omitted.
57This appears to have occurred because of a foreign country consumer who barely traded labour and capitalduring these periods. Since foreign consumers have 7 times as many units of labour and capital as domestic con-sumers in the experiment, this had observable e�ects. This demonstrates that a suÆcient number of experimentalsessions is necessary, in order to prevent the average results to be strongly a�ected by such variation.
4.3. VAN ELSWIJK SYSTEM 53
If the di�erence in capital export of 21 units is added to the foreign capital use in that case,this leads to an amount of 189, which is almost equal to the average amount in the wage taxsystem (193). If the level of employment and the production level are then increased by a factor21/193, these variables become also more or less the same as in the wage tax system. The taxrates and the budget de�cit abroad also show clear similarities with the wage tax system whenexperimental session 2 is omitted.
Finally, as was the case with the wage tax system, the output prices appear to be higherand the input prices lower than the theoretical equilibrium values, the only exception being thehome country price for product Y (and, when omitting session 2, only the market wage in thehome country).
Figures 4.7 to 4.10 illustrate the economic process under the Van Elswijk system. Theseshow the developments in the quantities and prices of inputs (labour and capital) and outputs(X and Y) during all the trading periods both in the home and the foreign country. Additionally,�gure 4.6 shows the changes in the budget surplus and tax rates. Since the discussion of theresults will be particularly centred on a comparison of the results of the two tax systems,�gures 4.11 to 4.23 also show the results of the wage tax system for the relevant variables.
We will once again start by looking at the developments in trading periods 1-8. Initially, theemployment of both labour and capital in the home and the foreign country appears to lie tothe same degree below the level of the wage tax system, (see �gures 4.11, 4.13 and 4.16 for thehome country, and 4.12 and 4.17 for the foreign country). The market wage in both countries(�gures 4.11 and 4.12) and the capital price (�gure 4.15) are also at a lower level. From tradingperiod 3 onwards, the level of employment in the home country rises above the level in the wagetax system, while it remains lower in the foreign country. As a result of this, the domestic sharein the total level of employment is higher under the Van Elswijk system from trading period3 onwards (�gure 4.13). However, the market wages (excluding tax/subsidy) tend towards thelevel under the wage tax system, in both countries. This results in a considerable di�erencebetween the e�ective wage for producers in the home and foreign country (�gure 4.14). Incontrast to the foreign country, the input of capital at home approaches the level under thewage tax system. This leads to the home country's share of the total employment of capital(from trading session 6 onwards) becoming larger (�gure 4.16). On the capital market, the priceof capital continues to fall (�gure 4.15) with a continual supply surplus.
In both countries, the production of good Y remains at the same level as in the wage taxsystem, helped by the somewhat lower price, especially during later trading periods (�gures 4.21to 4.23). Added to the developments on the factor markets, this means that the production ofthe international product X is at a lower level than under the wage tax system, accompanied bya higher price (�gure 4.18).58 From trading period 7 onwards, production in the home countryincreases above this level, helped by the much lower e�ective wage. The domestic share in thetotal production of X is also higher than under the wage tax system (�gure 4.19). As the �gureindicates, a considerable part of the domestic production is sold in the foreign country.
A further notable di�erence between the economic development in the home and the foreigncountry is that considerable budget surpluses continue to be generated in the home country,while budget de�cits continue to occur in the foreign country (�gure 4.6), as is the case withthe wage tax system (�gure 4.5).
58As has already been discussed, the notably lower level is mainly due to the deviant results in experimentalsession 2.
54 CHAPTER 4. EXPERIMENTAL RESULTS
Because of this, the tax rates in the home country reach a lower and quite stable level duringtrading periods 9 to 16, and the budget is more or less balanced. As we have seen already intable 4.4, the wage tax rate in the foreign country rises to the upper limit of 90% (as is alsothe case under the wage tax system) and de�cits remain. The level of employment rises onthe domestic labour market, accompanied by a rise in the market wage, until it reaches a levelapproximately 45% higher than that of the wage tax system (�gure 4.11). In the foreign countrythe level of employment approaches the level of the wage tax system (but not completely) after aninitial drop, while the market wage reaches the level of the wage tax system again (�gure 4.12).
The total amount of capital traded reaches the full employment level as the price of capitalgradual drops, as is also the case under the wage tax system (�gure 4.15). However, in theforeign country the employment of capital remains below the level reached under the wage taxsystem (�gure 4.17). This is due to the gradually increase in the import of capital by the homecountry, until it �nally (from trading period 14 onwards) leads to an capital employment levelthat is almost 50% higher than under the wage tax system (�gure 4.16). There appears to bea clear relation with the favourable development in employment in the home country comparedto the wage tax system, as just discussed, and the level of employment in the foreign country,which does not completely converge to the level of the wage tax system.
On the market for local product Y, the domestic price initially drops further below the levelof the wage tax system, accompanied by an increase in the amount produced, after which theprice remains more or less stable and the quantity increases to approximately 50% above thelevel of the wage tax system (�gure 4.22). In the foreign country, the level of production initiallydrops, at a somewhat higher price, after which it begins to approach the level under the wagetax system (�gure 4.23)).
On the international market for product X, the price level approaches that observed underthe wage tax system although convergence does not completely take place; the price stabilisesat a markedly higher level than that under the wage tax system (�gure 4.18).59 Comparisonof �gures 4.20 and 4.23 demonstrates that the foreign country production level of this productdi�ers more from the level under the wage tax system, than the production level of the localproduct Y does. This result is clearly related to the fact that the production of X is relativelycapital intensive and the employment of capital in the foreign country is less under the VanElswijk system. The production level in the home country �nally is approximately 20% higherthan that under the wage tax system (�gure 4.19). The persistent budget de�cit in the foreigncountry also indicates that the economic process has not yet (completely) stabilised at the end.
Summarizing conclusion
All things considered, the experimental results of the Van Elswijk system in terms of em-ployment, input of capital and production levels in the home country are more favourable thanwas expected, even taking into account the \favourable" theoretical equilibrium found for thissystem. The negative capital ight (capital import) is particularly surprising. If this is takeninto consideration, the results in the foreign country show clear similarities with the wage taxsystem; which is to be expected, since this system remains operative there under the Van El-swijk system. As is also the case with the results of the wage tax system, the results in the
59The deviant results of experimental session 2 under the Van Elswijk system, previously discussed, mainly a�ectthe production of X and therefore also the price of this product (see tables 4.5 and 4.6). If this experimentalsession were to be omitted, the values of these variables would di�er by approximately 10% from the wage taxsystem.
4.4. TESTING THE HYPOTHESES 55
foreign country di�er from the theoretical equilibrium in an unfavourable way. Employmentand production levels are lower, and the wage tax shows a tendency towards a level higher thanthe enforced maximum of 90% (the remaining budget de�cits suggest this). Furthermore, againthe product prices are higher and the factor prices are lower than the theoretical equilibriumlevels; in particular the lower capital price is noticeable. A possible explanation for this will begiven in section 4.5. In the following section, the hypotheses formulated by Van Elswijk and theSupervisory Committee concerning the results of the experiment will be tested.
4.4 Testing the hypotheses
A total of �ve hypotheses will be tested in which the following variables play a central role:(i) capital ight, (ii) labour employment, (iii) the net wage, (iv) the relative production levelof the production sectors, and (v) welfare. A null hypothesis has been formulated for each ofthese variables, along with one or two alternative hypotheses formulated by the SupervisoryCommittee (SC) and Van Elswijk (VE).
Each of the �ve hypotheses will be tested in two ways. Firstly, by comparing the average ofthe variable in question for consumers and/or producers under the wage tax system (system A)with the average under the Van Elswijk system (system B), per trading period. Secondly, byestimating the e�ect of the tax system on the variable concerned by means of regression analysis.
Three models are subjected to the regression analysis, in which the variable central to thehypothesis is the dependent variable. In models 1 and 2, the e�ect of the transition to theVan Elswijk system is estimated by including a so-called dummy (system B) as an independentvariable in the regression analysis. The value of this dummy is set to 0 in system A and at 1 insystem B. If this dummy is shown to be of signi�cant in uence, this will demonstrate that theVan Elswijk system has a signi�cantly di�erent in uence on the dependent variable. Model 1only considers trading periods 5-8, while model 2 only considers trading periods 8-16. This isbecause the tax rates remain constant in the �rst case, while they are adjusted to the budgetbalance in the previous trading period in the second case. In model 3, the in uence of the taxsystems is not investigated with the help of a dummy but by including the tax rates belonging tothese systems in the regression analysis. This provides information on the in uence of changesin the level of the taxes on the dependent variable, in addition to models 1 and 2. Therefore,this model only considers trading periods 8-16. Since in the Van Elswijk system the tax ratesfor the two production sectors are related to one another (namely determined by the rates whenthe system is introduced), only one of them will be considered; this rate concerns product X.Furthermore, the number of the trading period is included as an explaining variable in theregression analysis in all the models, in order to be able to �gure out possible trends over thetrading periods.
The so-called t-test is used to compare the averages (allowing for unequal variances). The so-called OLS-estimation method is used for the regression analysis (with standard errors correctedfor possible heteroscedasticity).
The following method will be applied to each of the �ve hypotheses. Firstly, the null hy-pothesis and the (alternative) VE-hypothesis and SC-hypothesis will be presented followed, bya precise description of the variable central to the hypothesis. The results of the t-tests andthe regression analyses will then be discussed after which conclusions will be drawn. For thesake of readability, the tables with the detailed test results are presented in an appendix (seeAppendix E).
56 CHAPTER 4. EXPERIMENTAL RESULTS
Hypothesis 1: Capital ight
Null hypothesis: Implementing the Van Elswijk system will not lead to capital ight.VE hypothesis: Same as the null hypothesis.SC hypothesis: Capital ight will take place.
De�nition
Two de�nitions of capital ight are used. Firstly, changes in \capital employment" in the homecountry are studied (Hypothesis 1.1) in which capital employment is de�ned as: the amount ofcapital used by a producer in the home country in a trading period, divided by the total amountof capital used in both the home and foreign country during this trading period.However, this de�nition is problematic in the case of an excess supply of capital. In that case,capital employment in the home country will also change even if only the sale and employmentof capital in the foreign country change. Therefore, the \net capital export" is also studied(Hypothesis 1.2), which is de�ned as: the di�erence between the amount of capital sold bya domestic consumer and the total capital employment in the home country, per domesticconsumer, during a trading period, as a fraction of the total capital employment in the homecountry per domestic consumer during that trading period.
Results for Hypothesis 1.1 Capital input
No signi�cant di�erences between the tax systems are found in the �rst 8 trading periods usingthe t-test. In the later trading periods, signi�cant di�erences are found from period 10 onwards,and these are in the direction of negative capital ight. The capital input then increases underthe Van Elswijk system.The regression analysis supports this result. Only in the later periods, when the tax ratesare adjusted, is a signi�cant (positive) e�ect found for the dummy (Van Elswijk system). Thedomestic tax rates have a negative and the foreign tax rate a positive in uence in these periodsin model 3. However, only the domestic tax rate under the wage tax system has a signi�cant(negative) in uence on the capital employment.
Results for Hypothesis 1.2 Net capital export
The results for this hypothesis are similar. From period 10 onwards, clearly signi�cant di�erencesare found that indicate negative capital ight under the Van Elswijk system. The results of theregression analysis are similar, with the exception that the Van Elswijk system is found to havea signi�cant e�ect in trading periods 5-8, too. Furthermore, both the domestic tax rates underthe Van Elswijk system, and the foreign wage tax, appear to have a signi�cant e�ect (in model3). These e�ects are similar to those concerning Hypothesis 1.1.
Conclusion
It can be concluded that the experimental results appear to give more support to Van Elswijk'shypothesis.
4.4. TESTING THE HYPOTHESES 57
Hypothesis 2: Labour employment
Null hypothesis: Implementing the Van Elswijk system will not lead to a change in total em-ployment.VE hypothesis: Employment will increase.SC hypothesis: Employment will decrease.
De�nition
When testing this hypothesis the variable \employment" is de�ned as: the amount of labourused by a home country producer in a trading period.
Results
The t-test shows that clear di�erences between the tax systems occur from trading period 9onwards. They indicate higher employment under the Van Elswijk system.The regression analysis supports this result. A signi�cant (positive) e�ect is only found fortrading periods 8-16 for the dummy (the Van Elswijk system). In model 3, negative e�ects areonce again found for the domestic tax rates in both tax systems and a positive e�ect is foundfor the foreign wage tax rate. However, only the e�ect of the domestic tax rate under the wagetax system is signi�cant.
Conclusion
It can be concluded that the experimental results give some support to the hypothesis of VanElswijk.
Hypothesis 3: Net wage
Null hypothesis: Implementing the Van Elswijk system will not lead to a change in the netwages of employees.VE hypothesis: Net wages will not decrease.SC hypothesis: Net wages will decrease.
De�nition
When testing the hypothesis, \net wage" is de�ned as: the nominal market price (per transac-tion) of labour during a trading period in the home country, divided by the sum of all averagemarket prices in this period. This normalization corrects for changes in the general price level.
Results
The t-test shows that for the Van Elswijk system the net wage is signi�cantly lower in the �rst 4trading periods as well as in period 8. However, from period 10 onwards the net wage becomessigni�cantly higher under this tax system.The regression analysis supports these results. While the dummy (the Van Elswijk system) issigni�cantly negative for trading periods 5-8, a signi�cantly positive e�ect is found for periods8-16. Model 3 demonstrates that the domestic tax rate has a signi�cantly negative impact
58 CHAPTER 4. EXPERIMENTAL RESULTS
under the wage tax system, which is also the case for the wage tax rate in the foreign country.Furthermore, the product tax rates have a (weakly) signi�cantly positive e�ect under the VanElswijk system. The undi�erentiated e�ect of the labour subsidy under this system could playa role here.
Conclusion
The experimental results support the SC hypothesis to a certain extent for those trading periodsin which the tax rates are not adjusted to the budget balance.60 The results seem to supportthe VE hypothesis in those trading periods in which that is the case.
Hypothesis 4: Shifts in production
Null hypothesis: Implementing the Van Elswijk system will not lead to a shift in relative pro-duction levels of the production sectors.VE hypothesis: Conditional.SC hypothesis: The capital intensive sector will become smaller in relation to the labour inten-sive sector.
De�nition
When testing this hypothesis, changes in the relative size of the \labour intensive productionsector" are studied. This is de�ned as: the number of units produced by a home countryproducer in the labour intensive sector Y in a trading period, divided by the total productionof the labour intensive sector Y and the capital intensive sector X in the home country in thisperiod.
Results
Neither with the t-test nor with the regression analysis signi�cant di�erences are found.
Conclusion
It can be concluded that the experimental results support the null hypothesis.
Hypothesis 5: Welfare
Null hypothesis: Implementing the Van Elswijk system will not lead to a change in welfare.VE hypothesis: Implementing will lead to an increase in welfare.SC hypothesis: Implementing will lead to a decrease in welfare.
60N.B. the negative e�ect of the Van Elswijk system in the regression analysis for the trading periods 5-8changes to a (signi�cantly) positive e�ect if the deviant experimental session 2 is omitted.
4.4. TESTING THE HYPOTHESES 59
De�nition
Four di�erent measures of welfare are studied when testing the hypothesis. Firstly, changes inthe \utility" are studied (Hypothesis 5.1), de�ned as: the earnings of a home country consumerin a trading period.From economic theory it is known that this variable can be used as a measure of the eÆciencyof a tax system in case of the theoretical equilibrium results. However, we have already seenthat the experimental results di�er from the theoretical results. Therefore, the measure \systemeÆciency", a measure used in experimental economics, is also used. For this we look at the\earnings" (Hypothesis 5.2), de�ned as: the earnings of a home country consumer or producerin a trading period.A problem of this measure is that the budget balance is not taken into account. Therefore,also \earnings including the government surplus" are studied (Hypothesis 5.3), de�ned as: theearnings of a home country consumer or producer plus (minus) the per capita government surplus(de�cit) in the home country in a trading period.Since the measure \gross national product" is seen as important in policy discussions, it will alsobe studied (Hypothesis 5.4). It is de�ned as: the average market price of X times the domesticproduction of X plus the average domestic market price of Y times the domestic production ofY, divided by the sum of these average market prices, in a trading period.61
Results for hypothesis 5.1: Utility
The t-test shows that the utility decreases signi�cantly under the Van Elswijk system in 5 of the�rst 8 trading periods, while no clear di�erences are found in the later trading periods (with theexception of periods 14 and 16, where the utility is signi�cantly larger under the Van Elswijksystem).The results of the regression analysis support this. The dummy (the Van Elswijk system) isfound to have a signi�cantly negative e�ect in trading periods 5-8 and no e�ect in the laterperiods. In model 3, a signi�cant (negative) e�ect is only found for the foreign wage tax rate,which could be due to its negative e�ect on the production of the internationally traded productX (through which there is less consumption possible). The wage tax rate in the home countryonce again has a negative e�ect, while the product tax rates have a positive e�ect; however,neither e�ects are signi�cant.
Results for hypothesis 5.2: Earnings
The results of the t-test show that there are no signi�cant di�erences between the tax systems intrading periods 1-8. The results of the later trading periods clearly show (signi�cantly) higherearnings under the Van Elswijk system.Contrary to the t-test, the regression analysis shows that the Van Elswijk system has a signi�-cantly positive e�ect, not only in trading periods 8-16 but also in periods 5-8. Model 3 once again
61Besides these measures, the relation between the actual production in a production sector and the maximumproduction that would have been achievable given the total amount of labour and capital employed for theproduction in the production sector has also been studied (a measure for production eÆciency). The average ofthis relation during the trading periods 5-8 for the production of Y in the home country is: 0.92 for system Aand 0.97 for system B (for the foreign country: 0.97 and 0.99, respectively). For the production of X the result is0.99 for both systems (also in the foreign country). The average of this relation during the trading periods 13-16for the production of Y in the home country is: 0.98 for system A and 0.95 for system B (for the foreign country:0.98 and 0.99, respectively). For the production of X the result is 0.97 for both systems (for the foreign country:0.98). See also Appendix F.
60 CHAPTER 4. EXPERIMENTAL RESULTS
shows the wage tax rates and the product tax rates in the home country to have (signi�cantly)negative e�ects, and the foreign wage tax rate to have a (signi�cantly) positive e�ect.
Results for hypothesis 5.3: Earnings including the government surplus
According to this criterion, the t-test shows earnings to be signi�cantly higher under the VanElswijk system both during trading periods 1-8 and in later periods.The results of the regression analysis support this. The e�ect of the dummy (the Van Elswijksystem) appears to be signi�cantly positive during trading periods 5-8 and also during the laterperiods. Model 3 once more shows the tax rates in the home country to have a negative e�ectwhile the tax rate in the foreign country has a positive e�ect, although only the e�ect of thedomestic wage tax rate is signi�cant.
Results for hypothesis 5.4: Gross national product
The results only concern the t-tests for trading periods 5-8 and 13-16 because of the smallnumber of observations (one per trading period and experimental session). The test shows thatthe GNP is signi�cantly larger under the Van Elswijk system in both cases.
Conclusion
It can be concluded that the experimental results seem to support the VE hypothesis more thanthe SC hypothesis. The only indication of a decrease in welfare under the Van Elswijk system isfound for the criterion of earnings (utility) of the consumers for those trading periods in whichthe tax rates do not adjust.62
Summarizing conclusion
As was to be expected on the basis of the discussion of the experimental results in the pre-ceding section, the results of the tests of the �ve hypotheses favour the Van Elswijk system. Theonly results that support the \pessimistic" hypotheses of the Supervisory Committee concernthe (negative) e�ects of this system on the net wage and the consumer utility in trading periods1-8, when the tax rates are not adjusted to the budget de�cit or surplus. Furthermore, thesenegative e�ects disappear if the deviant experimental session 2 under the Van Elswijk system isomitted.Interestingly, the regression analysis (model 3) always shows the wage tax rate in the homecountry to have a negative e�ect on the variables central to the hypotheses: capital and labouremployment in the home country, the net wage, the relative size of the labour intensive sectorY and the welfare measures. Furthermore, this negative e�ect is always signi�cant, with theexception of consumer utility and the relative size of the labour intensive sector. The e�ect ofthe foreign wage tax rate on these variables is generally positive but not signi�cant, while thee�ect of the home country product tax rates (for the Van Elswijk system) is generally negativebut not signi�cant. Moreover, the e�ect of the tax rates under the Van Elswijk system is alwaysless negative than the e�ect of the wage tax rate in the home country under the wage tax sys-tem. In the discussion of the experimental results in the following subsection these �ndings arediscussed more thoroughly.
62N.B. The regression analysis shows that the negative e�ect of the Van Elswijk system on the utility is nolonger signi�cant for trading periods 5-8 if the deviant session 2 is omitted.
4.5. THE \RISK-COMPENSATED PRICE MECHANISM" 61
4.5 The \risk-compensated price mechanism"
The favourable experimental results for the Van Elswijk system, expressed in the summarizingconclusions in 4.3 and 4.4 will be explained further in this section. Furthermore, attentionwill also be paid to the performance of the capital market and the relatively low capital priceobserved in the experiment.
The theoretical equilibrium analysis in chapter 3 has already demonstrated that the VanElswijk system can lead to more favourable results than the wage tax system, but these resultscan also be less favourable. Theoretically, the following two factors are of importance: on theone hand, the implicit taxation of (mobile) capital induces capital ight; on the other hand, theprice distortion with respect to the labour input (caused by the unemployment bene�t and therelated tax) is decreased because some of the tax burden can be shifted to the �xed factor (theimmobile capital; see 3.2).
The experimental results suggest that the \favourable" theoretical equilibrium is the moreattractive of the two. However, it should be added that the experimental results are clearlydi�erent from the theoretical ones. For example, the economic performance of the home countryeconomy - in terms of labour and capital employment, and production - is even better than thatpredicted by the \favourable" theoretical equilibrium.63 However, striking di�erences can alsobe seen with respect to other points. Firstly, it must be noted that the factor prices (the capitalprice in particular) fall below the theoretical equilibrium level and that the product prices riseabove the theoretical equilibrium levels. Relatedly, it appears that the input quantities usedand the production levels tend to lie at a lower level. This is particularly the case for the wagetax system and for the foreign country under the Van Elswijk system (where the wage taxsystem remains in place). Furthermore, it is striking that the wage tax leads to clearly negativeeconomic e�ects according to the regression analysis, whereas these negative e�ects are muchless prominent for the product tax under the Van Elswijk system. In our view these di�erentresults are related to an important and realistic di�erence between the experimental economyand the theoretical economic model, to which some attention was paid in 3.4.
We do not think that the aims of the consumers and/or producers in the experiment di�eredvery much from those assumed in the theoretical model (that is, utility and pro�t maximisation,respectively), since there are no indications that this is the case. This also holds for the possibilitythat they were confused about the way in which they could increase their earnings (utility, pro�t)after the �rst four trading periods - which were never in favour of the Van Elswijk system inthe analysis. The di�erence which we would like to focus on is the role that time plays. Inthe experiment, products can only be produced and traded once production factors (labourand capital) have been acquired. Because of this, both producers and consumers face risks.The producers are uncertain of the extent to which the production costs incurred will lead toa pro�t. Consumers experience similar time-related uncertainty with respect to the supply oflabour. After all, they are certain about the value of leisure, but not about the number of goodsthat they will be able to buy with the extra income generated from the labour time they trade.The supply of capital cannot lead to a similar alternative bene�t and so it is always better
63The following has to be noted. Firstly, the variation in the results of the experimental sessions is accountedfor in the statistical tests. Moreover, the results become even more favourable for the Van Elswijk system ifexperimental session 2, which is clearly deviant, is not taken into consideration. Secondly, consumer utility islower as a result of the deviations from the (eÆcient) theoretical equilibrium. It must be emphasised, though,that the relevant comparison is with the experimental results of the wage tax system and not the theoreticalequilibrium.
62 CHAPTER 4. EXPERIMENTAL RESULTS
to o�er it, irrespective of the price.64 Producers can be expected to be careful when buyingproduction factors because of the uncertainties mentioned, and this will have a restrictive e�ecton production.65 This results in a downward pressure on the factor prices and an upwardpressure on the product prices. The e�ect of this downward pressure on the market wage isopposed by the aforementioned uncertainty of consumers, and so the �nal e�ect on wages isunclear. On the other hand, the tendency for less labour to be traded is strengthened by this.The e�ects of these uncertainties may decrease as the economic process stabilises.
The experimental results are in accordance with this assumed process of price-formationwhich has also been found observed in other experimental markets and was elsewhere denotedas the \risk-compensated input/output price adjustment process".66 For the sake of conveniencewe will call it a \risk-compensated price mechanism". This mechanism - in combination withthe possibility of shifting the tax burden to the �xed factor - can explain the better economicperformance of the Van Elswijk system in comparison with the wage tax system and also, inmany respects, in comparison with the \favourable" theoretical equilibrium. With this mecha-nism namely, it really makes a di�erence for producers whether they have to pay taxes for theacquisition of labour up front (as in the wage tax system), or whether they are initially givena labour subsidy and are taxed later, in relation to price developments (as in the Van Elswijksystem). In the latter case, under the Van Elswijk system, a form of risk-sharing takes placebetween the producer and the government.67 In this perspective the results of the regressionanalysis that showed that the product tax rates have a clearly less negative e�ect than thedomestic wage tax rates can also be explained.
The sequential nature of the production process and the related uncertainties are not takeninto account in the theoretical equilibrium model. As we have already indicated, this is notthe case because of the wrong model choice, but re ects the present state of our theoreticalknowledge. In our view this is an important lacuna. In practise, there are possibilities fordecreasing this uncertainty (for example, via future contracts) but these are limited and costly.68
The risk-compensated price mechanism can be used to explain why the capital price is lowerthan in the theoretical equilibrium, while this is less so for the wage rate. We have alreadyindicated that, in contrast to the labour market (where, moreover, the unemployment bene�tforms a lower bound for the price-formation process), there is only a downward pressure on the
64Dynamic reputation e�ects, whereby it could be rational to constrain the supply of capital at low prices, arenot to be expected because of competition on the capital market.
65This will be the case if risk aversion is assumed. Contrary to what is generally assumed, empirical evidenceindicates that there are no di�erences between producers (entrepreneurs and managers) and the population atlarge as far as risk preferences are concerned; see R.H. Brockhaus, Sr., The psychology of the entrepreneur, in:C.A. Kent, D.L. Sexton and K.H. Vesper (eds.), Encyclopaedia of Entrepreneurship, Prentice-Hall, EnglewoodCli�s, 1982. At any rate, there is no clear empirical evidence that producers are not risk averse. In relation tothis also see G. Zhang (1998), Ownership concentration, risk aversion, and the e�ect of �nancial structure oninvestment decision, European Economic Review, 42, pp. 1751-1778.
66Noussair, Plott and Riezman (op.cit.). In this study - in which simultaneous markets are implemented (see2.5) - unsold labour is of no value for consumers. The opposite is the case in the following studies, in which an\undersupply" of labour is observed.: J. Hey and D. Di Cagno (1998), Sequential markets: An ExperimentalInvestigation of Clower's Dual Decision Hypothesis, Experimental Economics, 1, pp. 63-85; and P. Liang en C.R. Plott, General equilibrium, macroeconomics and money in a laboratory experimental environment, EconomicTheory, to be published.
67This risk-sharing only concerns the price risk and not the sales risk, in as far as the tax on value added underthe Van Elswijk system is concerned with the production value, irrespective of whether it is actually sold. In thiscontext it is noted, however, that production almost never di�ered from sales in the experiment (see Appendix C).
68The uncertainty in the experiment is limited for producers because the employment of capital (as of labour)only leads to �xed costs within a trading period.
4.6. CONCLUSIONS 63
capital price. Due to the resulting excess supply, more pressure is put on the capital price, sincecapital that is not traded does not yield any earnings to consumers. This pressure is furtherincreased because the employment of labour is also negatively a�ected by the risk-compensatedprice mechanism, which has a negative e�ect on the demand for capital via the small substitutionelasticity between labour and capital (in line with the MIMIC model). The reason why thecapital price is lower under the Van Elswijk system than under the wage tax system seems to bethe lower level of employment, initially in both countries and later only in the foreign country.It is not likely that the lower capital price under the Van Elswijk system is caused by a badperformance of the capital market. Firstly, �gure 4.15 shows that a decrease in the capital priceis accompanied by a higher input of capital, which eventually converges to full employment.This consideration is further supported by a regression analysis in which - on the basis of thetheoretical model - the relationship between the labour-capital ratio for producers on the onehand, and the capital price and the wage per unit of labour on the other hand, is estimated.69 All coeÆcients have the correct positive or negative sign and all but two of the eight aresigni�cant. Both non-signi�cant coeÆcients relate to the e�ect of the capital price on the labour-capital ratio in the labour intensive sector both in the home and the foreign country. A possibleexplanation for this is that capital expenditures play a less important role in this sector.
4.6 Conclusions
The experimental results presented above lead to the following conclusions.
1. Under the wage tax system, the quantities (particularly on an aggregated level) tendtowards the theoretical equilibrium when the wage tax rate is kept constant at the equi-librium level (trading periods 1-8). However, this development is accompanied by con-siderable government de�cits both in the home and the foreign country. Once the wagetax rates are adjusted to the government de�cits (trading periods 9-16) the de�cits de-crease in size. However, employment and production reach a level that is clearly belowthe theoretical level and this is also the case for welfare (consumer utility).
2. Regarding the (relative) prices, there is no clear convergence towards the theoretical equi-librium results. Generally, the prices of production factors - the capital price in particular- tend to be lower and the product prices tend to be higher than the theoretical equilib-rium results. This is particularly the case once the wage tax is adjusted to the governmentde�cit in the previous trading period (trading periods 9-16).
3. The previous conclusion can be related to the observation that the quantities traded on themarkets for labour, capital and consumption goods generally fall short of the theoreticalequilibrium level. A lower level of employment of production factors has a restrictive e�ecton production. The former has a negative e�ect on the factor prices while the latter hasa positive e�ect on the product prices.
4. The results for the Van Elswijk system, in terms of labour and capital employment andproduction in the home country, are more favourable than the \favourable" theoreticalequilibrium for this system, and are therefore also much more favourable than that of thewage tax system. The negative capital ight (capital import) is particularly striking. Ifthis is taken into account the results for the foreign country show clear similarities with
69See Appendix F.
64 CHAPTER 4. EXPERIMENTAL RESULTS
the wage tax system. This is not surprising since this system remains operative in theforeign country under the Van Elswijk system.
5. Under the Van Elswijk system, the product prices also remain higher and the factor priceslower than the theoretical equilibrium levels.
6. The results of the tests of the �ve hypotheses formulated by Van Elswijk and the Supervi-sory Committee in order to compare the performance of the Van Elswijk system with thatof the wage tax system favour the Van Elswijk system. The hypotheses concern capital ight, labour employment, the net wage, the relative size of the labour intensive (shel-tered) sector and welfare. Only the negative e�ects of this system on the net wage andconsumer utility (one of the welfare measures) in the �rst trading periods, in which the taxrates are not adjusted to the government balance, support the Supervisory Committee'shypotheses. However, these negative e�ects disappear once a clearly deviant session underthe Van Elswijk system is omitted.
7. Under the wage tax system, the wage tax rate in the home country always has a negativee�ect on the variables playing a central role in these hypotheses, and this e�ect tends tobe signi�cant. Under the Van Elswijk system, the e�ect of the product tax rates tends tobe similar but it is always less negative and mostly not signi�cant.
8. A \risk-compensated price mechanism" provides an explanation for the evolution of pricesand quantities (see conclusions 2, 3 and 5) observed under both tax systems. Consumersand producers are confronted with uncertainty because production factors must be tradedbefore consumption goods become available; a fact ignored in the theoretical model. Thus,risk aversion leads to compensatory developments in the process of price-formation whichcan explain the results, as is discussed in 4.5.
9. The fact that it is possible to shift the tax burden to the �xed factor (immobile capital)can by itself already explain why the Van Elswijk system can lead to a better economicperformance than the wage tax system; the theoretical analysis in chapter 3 demonstratedthis. The result that the experimental economy was more attracted to the \favourable"than the \unfavourable" theoretical equilibrium, as well as the observation that the exper-imental results for the home economy are generally even better than in the \favourable"theoretical equilibrium can be explained by the aforementioned \risk-compensated pricemechanism". Under the Van Elswijk system, on the one hand, labour is no longer taxedbut actually subsidised, and on the other hand, taxes are levied later and in proportionto the development in prices. In contrast to the wage tax system, the sales risks of theproducers are shared by the government. This can also explain why the product tax ratesunder the Van Elswijk system have a more limited negative e�ect than the wage tax rateunder the wage tax system (see conclusion 7).
4.7. TABLES AND FIGURES WITH RESULTS 65
4.7 Tables and �gures with results
Table 4.1: Experimental results wage tax system (A), averageover periods 5 to 8
home country wage tax system (A) wage tax system (A)
labour foreign country labour foreign country (theoretical)
capital total capital total (theoretical)
capital foreign country capital foreign country (theoretical)
labour home country labour home country (theoretical)
capital home country capital home country (theoretical)
Figure 4.1: Evolution of input quantities under the wage tax system
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
wag
e (r
elat
ive)
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
capi
tal p
rice
(rel
ativ
e)
wage home country wage (theoretical) wage foreign country
capital price capital price (theoretical)
Figure 4.2: Evolution of input prices under the wage tax system
78 CHAPTER 4. EXPERIMENTAL RESULTS
0
20
40
60
80
100
120
140
160
180
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
fore
ign
coun
try
and
tota
l
0
10
20
30
40
50
60
hom
e co
untr
y
X consumption total X total (theoretical)
X sold by foreign producers X foreign country (theoretical)
Y consumption foreign country Y foreign country (theoretical)
X sold by home producers X home country (theoretical)
Y consumption home country Y home country (theoretical)
Figure 4.3: Evolution of output quantities under the wage tax system
0
0.05
0.1
0.15
0.2
0.25
0.3
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
pric
es (
rela
tive)
X price X price (theoretical) Y price home country
Y price (theoretical) Y price foreign country
Figure 4.4: Evolution of output prices under the wage tax system
4.7. TABLES AND FIGURES WITH RESULTS 79
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
surp
lus/
GN
P
-90%
-60%
-30%
0%
30%
60%
90%
tax
rate
s
surplus home country surplus foreign country
wage tax rate home country wage tax rate foreign country
Figure 4.5: Budget surplus and tax rates under the wage tax system
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
surp
lus/
GN
P
-90
-60
-30
0
30
60
90
tax
rate
s (%
)
surplus home country surplus foreign country
wage tax rate foreign country tax rate on X home country
tax rate on Y home country
Figure 4.6: Budget surplus and tax rates under the Van Elswijk system
80 CHAPTER 4. EXPERIMENTAL RESULTS
0
20
40
60
80
100
120
140
160
180
200
220
240
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
fore
ign
coun
try
and
tota
l
0
10
20
30
40
50
60
70
80
90
100
hom
e co
untr
y
labour foreign country capital total capital foreign country
labour home country capital home country
Figure 4.7: Evolution of input quantities under the Van Elswijk system
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
wag
e (r
elat
ive)
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
capi
tal p
rice
(rel
ativ
e)
wage home country wage foreign country capital price
Figure 4.8: Evolution of input prices under the Van Elswijk system
4.7. TABLES AND FIGURES WITH RESULTS 81
0
20
40
60
80
100
120
140
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
fore
ign
coun
try
and
tota
l
0
10
20
30
40
50
60
70
80
hom
e co
untr
y
X consumption total Y consumption foreign country
X sold by foreign producers Y consumption home country
X sold by home producers
Figure 4.9: Evolution of output quantities under the Van Elswijk system
0
0.05
0.1
0.15
0.2
0.25
0.3
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
pric
es (
rela
tive)
X price Y price home country Y price foreign country
Figure 4.10: Evolution of output prices under the Van Elswijk system
82 CHAPTER 4. EXPERIMENTAL RESULTS
0
5
10
15
20
25
30
35
40
45
50
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
quan
tity
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
wa
ge
(rel
ativ
e)
labour home country (wage tax system) labour home country (Van Elswijk system)
wage home country (wage tax system) wage home country (Van Elswijk system)
Figure 4.11: Labour employment and wage in the home country
0
50
100
150
200
250
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
quan
tity
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
wag
e (r
elat
ive)
labour foreign country (wage tax system) labour foreign country (Van Elswijk system)
wage foreign country (wage tax system) wage foreign country (Van Elswijk system)
Figure 4.12: Labour employment and wage in the foreign country
4.7. TABLES AND FIGURES WITH RESULTS 83
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
shar
e
home country's share of total labour employment (wage tax system)
home country's share of total labour employment (Van Elswijk system)
foreign country's share of total labour employment (wage tax system)
foreign country's share of total labour employment (Van Elswijk system)
Figure 4.13: Shares of total labour employment
0
0.05
0.1
0.15
0.2
0.25
0.3
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
wag
e (r
elat
ive)
home country's wage (wage tax system)
home country's effective (after tax) wage (wage tax system)
home country's wage (Van Elswijk system)
home country's effective (after subsidy) wage (Van Elswijk system)
Figure 4.14: Wage and e�ective (after tax/subsidy) wage
84 CHAPTER 4. EXPERIMENTAL RESULTS
0
40
80
120
160
200
240
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
quan
tity
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
cap
ital
pric
e (r
elat
ive
)
total capital employment (wage tax system)
total capital employment (Van Elswijk system)
capital price (wage tax system)
capital price (Van Elswijk system)
Figure 4.15: Total capital employment and capital price
-5
5
15
25
35
45
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
quan
tity
-0.05
0.05
0.15
0.25
0.35
0.45
shar
e
capital in home country (wage tax system)
capital in home country (Van Elswijk system)
home country's net capital import (wage tax system)
home country's net capital import (Van Elswijk system)
home country's share of total capital employment (wage tax system)
home country's share of total capital employment (Van Eswijk system)
Figure 4.16: Capital employment in the home country
4.7. TABLES AND FIGURES WITH RESULTS 85
0
50
100
150
200
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
quan
tity
0.70
0.75
0.80
0.85
0.90
0.95
1.00
1.05
1.10
1.15
1.20
shar
e
capital in foreign country (wage tax system)
capital in foreign country (Van Elswijk system)
foreign country's share of total capital employment (wage tax system)
foreign country's share of total capital employment (Van Eswijk system)
Figure 4.17: Capital employment in the foreign country
0
20
40
60
80
100
120
140
160
180
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
quan
tity
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
pric
e (r
elat
ive)
sold units of good X (wage tax system) sold units of good X (Van Elswijk system)
price of good X (wage tax system) price of good X (Van Elswjk system)
Figure 4.18: Market for the international good X
86 CHAPTER 4. EXPERIMENTAL RESULTS
-15
-10
-5
0
5
10
15
20
25
30
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
quan
tity
-0.5
-0.25
0
0.25
0.5
0.75
1
shar
e
units of good X sold in the home country (wage tax system)
units of good X sold in the home country (Van Elswijk system)
home country's net import of good X (wage tax system)
home country's net import of good X (wage tax system)
home country's share of total production of good X (wage tax system)
home country's share of total production of good X (Van Elswijk system)
Figure 4.19: Home country's market for the international good X
0
20
40
60
80
100
120
140
160
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
quan
tity
0
0.5
1
1.5
2
2.5
3
shar
e
units of good X sold in the foreign country (wage tax system)
units of good X sold in the foreign country (Van Elswijk system)
foreign country's share of total production of good X (wage tax system)
foreign country's share of total production of good X (Van Elswijk system)
Figure 4.20: Foreign country's market for the international good X
4.7. TABLES AND FIGURES WITH RESULTS 87
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
shar
e
home country's share of consumption of good Y (wage tax system)
foreign country's share of consumption of good Y (wage tax system)
home country's share of consumption of good Y (Van Elswijk system)
foreign country's share of consumption of good Y (Van Elswijk system)
Figure 4.21: Shares of total consumption of the local good Y
0
5
10
15
20
25
30
35
40
45
50
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
quan
tity
0
0.05
0.1
0.15
0.2
0.25
pric
e (r
elat
ive)
home country's consumption of good Y (wage tax system)
home country's consumption of good Y (Van Elswijk system)
home country's price of good Y (wage tax system)
home country's price of good Y (Van Elswijk system)
Figure 4.22: Home country's market of the local good Y
88 CHAPTER 4. EXPERIMENTAL RESULTS
0
20
40
60
80
100
120
140
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
period
quan
tity
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
pric
e (r
elat
ive
)
foreign country's consumption of good Y (wage tax system)
foreign country's consumption of good Y (Van Elswijk system)
foreign country's price of good Y (wage tax system)
foreign country's price of good Y (Van Elswijk system)
Figure 4.23: Foreign country's market of the local good Y
Part IV
Appendices
89
Appendix A
Training, closed economy, payment
procedure
As mentioned in 2.5, all subjects took part in a training session and an experiment with a closedeconomy before participating in the (open economy) experiment. This appendix explains thetraining procedure, the design of the closed economy experiment and the procedures for partic-ipation and payment. As the design of an experiment such as this is not simple, it is a commonprinciple in experimental economics to give subjects the opportunity to acquire suÆcient experi-ence with the trading rules and trading itself before participating in the experiment. Therefore,each subject participated in 3 sessions: a training session, a session with a closed economy anda session with an open economy (the experiment itself). Each subject participating in a sessionwith the wage tax system in the experiment was also faced with this tax system in the trainingsession and the closed economy session. By the same token, each subject participating in asession with the Van Elswijk system was also faced with this system in the training session andthe closed economy session. Furthermore, for the training, separate sessions were organised forthose playing the role of consumer or the role of producer. Subjects kept the same role in allthree sessions. Tables A.1 and A.2 give an overview of all the sessions that were organised forthe experiment with the wage tax system and the experiment with the Van Elswijk system,respectively.
Participation and procedure of payment
Subjects were recruited through announcements on notice boards in the building whichhouses the CREED laboratory of the University of Amsterdam. Each participant had to signup for three sessions in this laboratory (a training session, a closed economy session and anopen economy session). As can be seen from the tables, more subjects took part in the trainingsessions than were necessary for the closed economy sessions, and more participated in the lattersessions than were necessary for the experiment itself (the open economy sessions). The reasonfor this is twofold. Firstly, this was done so that there would be some subjects in reserve inthe case of sickness or other eventualities. Secondly, this provided the possibility for subjectshaving diÆculties with understanding the instructions or with trading to drop out. Participationin the closed economy sessions was on the basis of \�rst come, �rst served", or if this was notpossible, on the basis of drawing lots. For the experiment itself (the open economy sessions), theparticipants were selected on the basis of their performance (earnings) in the closed economy.
91
92 APPENDIX A. TRAINING, CLOSED ECONOMY, PAYMENT PROCEDURE
Table A.1: Summary of all sessions concerning the wage tax system
date session # of subjects # of trading periods(practice rounds)
280998/M training consumers 1 13 -
280998/A training consumers 2 11 -
290998/M training consumers 3 14 -
290998/A training producers 1 15 -
011098/M training producers 2 10 -
011098/A training producers 3 10 -
021098/M training producers 4 14 -
051098/M closed economy 1 16 14 (3)
051098/A closed economy 2 16 13 (3)
061098/M closed economy 3 16 12 (3)
061098/A closed economy 4 16 14 (3)
081098/M open economy 1 16 16 (2)
081098/A open economy 2 16 16 (2)
091098/M open economy 3 16 16 (2)
Note: `M' indicates a morning session and `A' an afternoon session.
Table A.2: Summary of all sessions concerning the Van Elswijk system
date session # of subjects # of trading periods(practice rounds)
121098/M training consumers 1 12 -
121098/A training consumers 2 13 -
131098/M training consumers 3 10 -
131098/A training producers 1 11 -
151098/M training producers 2 11 -
151098/A training producers 3 11 -
161098/M training producers 4 13 -
191098/M closed economy 1 16 14 (3)
191098/A closed economy 2 16 13 (3)
201098/M closed economy 3 16 12 (3)
201098/A closed economy 4 16 14 (3)
221098/M open economy 1 16 16 (2)
221098/A open economy 2 16 16 (2)
231098/M open economy 3 16 16 (2)
Note: `M' indicates a morning session and `A' an afternoon session.
93
Participants were only paid if they showed up at all three sessions. Those who ultimatelydid not take part in the open economy session were paid at the beginning of this session, whilethose who did participate were paid at the end. Each session lasted approximately three anda half hours. Each participant received a show-up fee of 70 guilders for the training session.A show-up fee of 40 guilders was paid for the closed economy session and a show-up fee of 10guilders for the open economy session. The earnings of the participants in the closed and openeconomy sessions were also included in the payments. Possible losses were also included with theone condition that the minimum participants could end up with was 0 guilders. All participantswere informed of the payment procedure at the beginning of the training session. On average,participants earned 26 guilders in the closed economy and 120 guilders in the open economy(besides the show-up fee). The total number of subjects was 168, the majority of which (60%)were students at the Faculty of Economics and Econometrics. In the actual experiment, the 3open economy sessions with the wage tax system and the 3 open economy sessions with the VanElswijk system, these numbers were 48 (75% economics students) for the wage tax system and48 (65% economics students) for the Van Elswijk system.
Training session
The training session was organised as follows. After arriving at the reception room, thesubjects were informed about the procedures of participation and payment described above.Then they each picked an envelope containing a seat number in the laboratory. Each trainingsession consisted of three phases. In phase 1, the participants familiarised themselves with thecomputer screens, the use of the keyboard, �lling in the personal registration form, calculatingthe earnings, and the trading rules (for the latter, see Appendix B). In phase 2 they had toanswer a number of questions, focused on their role as consumer or producer, to test theirunderstanding of the trading rules, the way the markets worked, and so on. This phase wasonly �nished when everybody had answered the questions correctly. After a short break, phase3 began in which they had to trade on the factor markets and the product markets a numberof periods. This phase was organised in the same way as the later closed and open economysessions, except that in this session the experimenters played the role of the other side of themarket. That is, during the consumer training session the experimenters played the role ofproducers, and during the producer training session they played the role of consumers. To avoidthis having an e�ect on trading in the experiment, the experimenters randomly and greatlyvaried the prices and quantities o�ered and demanded in all training sessions.
Finally, it is noted that in the training sessions as well as in the closed and open economysessions instead of using the words `consumer', `producer', `labour' and `capital', the terms `type-I trader', `type-II trader', `product V' and `product W' were used. As it is common practice inexperimental economics this was done to avoid uncontrollable behavioural rules that subjectsmight attach to concepts such as labour or capital.
Closed economy experiment
The purpose of the closed economy sessions was to provide participants with the opportunityto experience the experimental economy. This was considered particularly important becausein the actual (open economy) experiment, some of the subjects (through picking an envelopein the reception room) would play the role of a \large" consumer or producer in the foreigncountry. Therefore, subjects had to be able to trade many units of a product within a limitedtime period. The design of the closed economy sessions was the same as the open economy
94 APPENDIX A. TRAINING, CLOSED ECONOMY, PAYMENT PROCEDURE
sessions. Therefore, only the most important di�erences will be mentioned here. For the rest,we refer to the description of the experimental sessions in 2.5 and the instructions for the openeconomy in Appendix B. Firstly, there was obviously no foreign country in the closed economyexperiment. Nevertheless, the total number of producers for the products X (4) and Y (6)remained the same as did the total number of consumers (6). Secondly, the parameter values,that is the scaling factors of the production functions and the amounts of labour and capital,were chosen such that in a theoretical equilibrium situation the closed economy would be a copyfour times the size (in terms of production, consumption etc.) of the home economy in theactual (open economy) experiment. For example, consumers in these sessions had four times asmany units of labour and capital at their disposal than the home country consumers in the openeconomy. Apart from these di�erences, the design of the closed economy sessions was the sameas the open economy sessions.
Appendix B
Instructions, trading rules, computer
screens
This appendix contains a translation from Dutch of the instructions (incl. redemption valuesand production schedules), and the trading rules. Print-outs of the computer screens are addedfor further information.
B.1 Instructions
General Instructions (part 1)
General Instructions
These instructions are part of an experimental study into decision-making on markets. If you follow theinstructions carefully you can earn a considerable amount of money. Your earnings will be paid out toyou in cash, personally and con�dentially, at the end of the experiment.
During todays session, you will participate on a number of markets during a number of time periods inwhich trading can take place. During these trading periods, you will act as either a Type-I trader ora Type-II trader. During all the trading periods, you will be the same type of trader. You will �nd aREGISTRATION TABLE in the envelope on your table. Look up the type you represent at the beginningof the form. If you are a Type-I trader you will also �nd a REDEMPTION VALUES TABLE. If youare a Type-II trader you will �nd a PRODUCTION TABLE. These tables will help you to weigh upthe decisions you can make. It is not allowed to exchange the information on the tables withothers. It concerns your own personal information.In the training session, you have already familiarized yourself with the trading rules. You will �nd asummary of these rules in your envelope.
On the markets, the \franc" will be used as the unit of currency. All trading will take place in francs. Atthe end of the experiment, you will be paid in guilders. The exchange rate - which indicates how much afranc is in guilders (or cents) - is shown on your Registration table. The more francs you earn, the moreguilders you will earn.
Four di�erent goods can be traded: V, W, X and Y. Goods W and X will be traded on two separatemarkets, one for W (denoted W1) and one for X (denoted X1). Goods V and Y will be traded on fourseparate markets, two for V (denoted V1 and V2) and two for Y (denoted Y1 and Y2). Your tradingwill be restricted to some of these markets. The markets on which you can trade are shown on yourRegistration table.
95
96 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
Each trading period consists of two phases. In phase 1, only the markets for V and W are open. Onthese markets, the Type-I traders are sellers and the Type-II traders are buyers. Once phase 1 is �nishedphase 2 begins. In phase 2, only the markets for X and Y are open. On these markets, Type II-tradersare now sellers and Type-I traders are buyers. Type-I traders can earn money by consuming goods, whileType-II traders can earn money by making a pro�t.
The Speci�c Instructions for Type-I and Type-II traders that you will �nd below, show you preciselyhow your earnings will be determined. We ask you now to read these carefully. If you have a questionregarding these instructions, which concern personal information, put up your hand. We will then cometo your table to answer your question. If everybody has �nished with the Speci�c Instructions we willcarry on with some �nal instructions for both types of traders.
In the experiment the Speci�c Instructions were given here. (For their content, see below.)
General Instructions (part 2)
Instructions for both types of traders
HOW THE SYSTEM WORKS
Type-I traders receive an initial endowment of cash (an initial cash) and an initial endowment of V andW. They want to consume V but also X and Y. In phase 1 they can sell V and W to Type-II traders inorder to increase their amount of cash to be able to buy X and Y in phase 2.
Type-II traders receive an initial endowment of cash (an initial cash). In phase 1 they can use this tobuy units of V and W. They need these units to produce units of X and Y, which they can sell to TypeI-traders in phase 2 to make a pro�t.
EARNINGS
In the Speci�c Instructions it was explained how your earnings in a period are determined. Your earningswill be expressed in francs. At the end of todays meeting, your total earnings during all the tradingperiods will be exchanged into guilders using the exchange rate stated on your Registration table.
INITIAL ENDOWMENTS
At the beginning of each trading period, you will receive an initial cash with which to buy goods. Inaddition, Type-I traders also receive some units of V andW which they can sell. These initial endowments,which will be the same in each period, are mentioned on your Registration table.
MARKET RESTRICTIONS (in wage tax system)
As indicated in the General Instructions, there are di�erent markets for the goods V, W, X and Y. Themarkets for V and W (V1, V2 and W1) will be open for 4 minutes and 30 seconds. After a short pause of20 seconds, the markets for X and Y (X1, Y1 and Y2) will be open for 3 minutes and 30 seconds. Thereis only one exception to this rule which occurs if a good is sold out. In this case, the market for thisgood will close. The cash which is involved with an outstanding bid on this market will then be availableagain for other purchases.
Your trading will be restricted to some of the markets. The markets on which you can trade are mentionedon your Registration table.
Some of you will have to pay taxes and/or receive subsidies. All the tax rates and subsidies relevant tothe markets on which you can trade are mentioned on your Registration table.N.B. On each separate market, all traders of a certain type (Type-I or Type-II) will be faced with thesame tax rates and/or subsidies concerning the good that is being traded on this market.
You will encounter market prices as well as inclusive prices. Market prices are of direct importance forType-I traders. Inclusive prices are of direct importance for Type-II traders because these prices includethe taxes or subsidies which apply to them.
Apart from two training periods (in which you cannot earn any money), there will be 16 trading periodsin total. After each period there will be a 2 minutes pause before the next period begins. For those of
B.1. INSTRUCTIONS 97
you who have to pay taxes it is important to note that the tax rates during the �rst 8 periods (periods 1- 8) will remain the same. Thereafter, beginning with period 9, the tax rates can be di�erent in di�erentperiods. We will warn you when this occurs.
If you have a question put up your hand. We will then come to you to answer your question.
MARKET RESTRICTIONS (in Van Elswijk system)
As indicated in the General Instructions, there are di�erent markets for the goods V, W, X and Y. Themarkets for V and W (V1, V2 and W1) will be open for 4 minutes and 30 seconds. After a short pause of20 seconds, the markets for X and Y (X1, Y1 and Y2) will be open for 3 minutes and 30 seconds. Thereis only one exception to this rule which occurs if a good is sold out. In this case, the market for this goodwill close. The cash which is involved in an outstanding bid on this market will then be available againfor other purchases. Your trading will be limited to some of the markets. The markets on which you cantrade are mentioned on your Registration table.
Some of you will have to pay taxes and/or receive subsidies. All the tax rates and subsidies relevant tothe markets on which you can trade are mentioned on your Registration table.N.B. On each separate market, all traders of a certain type (Type-I or Type-II) will be faced with thesame tax rates and/or subsidies concerning the good that is being traded on this market, except onmarket X1. On market X1, some Type-II traders will have to pay taxes on top of the market price of X,while other Type-II traders will not have to do this.
You will encounter market prices as well as inclusive prices. Market prices are of direct importance forType-I traders. Inclusive prices are of direct importance for Type-II traders because these prices includethe taxes or subsidies which apply to them. Note that the inclusive prices for Type-II traders will di�erbecause these traders do not all have to pay taxes on top of the market price for X. This will be indicatedwith an asterisk (*).
Apart from two training periods (in which you cannot earn any money), there will be 16 trading periodsin total. After each period there will be a 2 minutes pause before the next period begins. For those ofyou who have to pay taxes it is important to note that the tax rates during the �rst 8 periods (periods 1- 8) will remain the same. Thereafter, beginning with period 9, the tax rates can be di�erent in di�erentperiods. We will warn you when this occurs.
If you have a question put your hand up. We will then come to you and answer it.
Speci�c instructions for consumers (Type-I) and producers (Type-II)
Domestic consumers in the wage tax system and the Van Elswijk system
Speci�c instructions for Type-I traders
Each trading period consists of 2 phases: phase 1 and phase 2.
Phase 1:
In each period, you will receive an initial endowment of cash (an initial cash) and an initial endowmentof V and W at the beginning of phase 1. Your initial endowment of V and W and your initial cash, whichwill be the same in each period, are mentioned on your Registration table.
In phase 1, the markets for V and W are open and you are free to sell however much of your endowmentof goods V and W to whoever wishes to buy these goods. The amount of francs that you receive pertransaction depends on the market price at which you are selling and the amount of units that you areselling at this price.
However, note that you will receive francs for every unit of V that you do not sell in phase 1 on top ofyour sales proceeds and initial cash. For every unit of V that you do not sell you will receive a subsidyof 70 francs.
Your total cash inventory in francs at the end of phase 1 is therefore equal to:
98 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
Cash inventory end of phase 1 = initial cash + sales proceeds at market prices + subsidies.
In phase 2, you will need this cash inventory to buy units of X and Y. These units of X and Y, togetherwith the units of V that you did not sell in phase 1 will determine your earnings for the trading period,as will be explained below.
Phase 2:
In phase 2, the markets for X and Y are open. In this phase, you are free to buy units of X and Y fromwhoever wishes to sell these units. However, to be able to buy these units you need to have enough cashto be able to pay the market price.
Each unit that you buy will be added to your endowment of the good in question. All the units youhave in inventory at the end of the period, including the units of V which you did not sell in phase 1 areconsidered as having been consumed by you. Your earnings in francs in a trading period are determinedby the redemption value of the goods consumed by you. The Redemption value table shows the amountthat you will receive. Take the table out of the envelope in front of you.
Note �rst that the redemption value is always 0 as long as you do not have at least one unit of all thegoods (V, X and Y). There is a warning at the top of the table about this. For the �rst unit of V thatyou consume in a trading period you will receive the amount mentioned in the �rst row of the column\V unit value". If you consume a second unit of V, you will receive the amount mentioned in the secondrow of the column \V unit value". The total amount that you will receive through your consumption ofboth units of V is in the second row of the column \V total value". You can calculate the amount thatyou will receive through consuming X and Y in exactly the same way. Your earnings in a period are thendetermined by the sum of your earnings through consuming X, Y and V. Thus:
Earnings in period = V total value + X total value + Y total value
You are already familiar with reading the Redemption value table. We now ask you to answer thefollowing questions using the table that you have received.
Question 1
Suppose that in a certain period you have not sold 14 units of V during phase 1. Furthermore, supposethat you have bought 1 unit of Y and 10 units of X during phase 2. Therefore, at the end of this periodyou have 14 units of V, 1 unit of Y and 10 units of X. What are your total earnings for this period?
Your answer: .........................................................................................
Question 2
What are your total earnings in this period if you have 0 units of Y instead of 1 unit at the end of theperiod?
Your answer: .........................................................................................
Question 3
Suppose that in a certain period you have not sold 2 units of V during phase 1. Furthermore, supposethat you have bought 7 units of Y and 3 units of X during phase 2. Therefore, at the end of the periodyou have 2 units of V, 7 units of Y and 3 units of X. What are your total earnings for this period?
Your answer: .........................................................................................
Warnings: In the �rst place, note that you will earn nothing if you do not consume any units of acertain good (V, X or Y), regardless of the number of other goods you consume. Also note from the �rstrow of the Redemption value table that you will only start earning money once you have more than 1unit of a product.Further note that unsold units of W and leftover cash at the end of phase 2 will not provide you withany earnings.Finally note that you will come across market prices as well as inclusive prices. Market prices are of direct
B.1. INSTRUCTIONS 99
importance for Type-I traders. Inclusive prices are of direct importance for Type-II traders because theseprices include the taxes or subsidies that apply to them.
REGISTRATION TABLE
Take the Registration table out of the envelope in front of you. The top of this table gives you personalinformation concerning: your type, the markets on which you can trade, your exchange rate, your initialendowments, the subsidies you will receive and/or the taxes you have to pay, as well as other relevantsubsidies and/or taxes. In the table you should register your sales, your purchases and your earnings foreach period. You are already familiar with �lling in this table. We would just like to remind you thatyou should �ll in the number of unsold units of V - in the �rst row under \your sales" - in the pausebetween phases 1 and 2. You should �ll in the rest in the pause at the end of the period. You can usethe example in the grey column to help you with this.
Foreign consumers in the wage tax and Van Elswijk system
Speci�c Instructions for Type-I traders
Each trading period consists of 2 phases: phase 1 and phase 2.
Phase 1:
In each period, you will receive an initial endowment of cash (an initial cash) and an initial endowmentof V and W at the beginning of phase 1. Your initial endowment of V and W and your initial cash, whichwill be the same in each period, are mentioned on your Registration table.
In phase 1, the markets for V and W are open and you are free to sell however much of your endowmentof goods V and W to whoever wishes to buy these goods. The amount of francs that you receive pertransaction depends on the market price at which you are selling and the amount of units that you areselling at this price.
However, note that you will receive francs for every unit of V that you do not sell in phase 1 on top ofyour sales proceedings and initial cash. For every unit of V that you do not sell you will receive a subsidyof 70 francs.
Your total cash endowment in francs at the end of phase 1 is therefore equal to:
Cash inventory end of phase 1 = initial cash + sales proceeds at market prices + subsidies
In phase 2, you will need this cash inventory to buy units of X and Y. These units of X and Y, togetherwith the units of V that you did not sell in phase 1 will determine your earnings during the tradingperiod, as will be explained below.
Phase 2:
In phase 2, the markets for X and Y are open. In this phase, you are free to buy units of X and Y fromwhoever wishes to sell these units. However, to be able to buy these units you need to have enough cashto pay the market price.
Each unit that you buy will be added to your endowment of the good in question. All the units youhave in inventory at the end of the period, including the units of V which you did not sell in phase 1are considered as having been consumed by you. Your earnings in francs during a trading period aredetermined by the redemption value of the goods consumed by you. The Redemption value table indicatesthe amount that you will receive. Take the table out of the envelope in front of you.
Note �rst that the redemption value is always 0 as long as you do not have at least one unit of all thegoods (V, X and Y). There is a warning at the top of the table about this. For the �rst unit of V that youconsume during a trading period you will receive the amount mentioned in the �rst row of the column\V unit value". If you consume a second unit of V, you will receive the amount mentioned in the secondrow of the column \V unit value". The total amount that you will receive through your consumption of
100 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
both units of V is in the second row of the column \V total value". You can calculate the amount thatyou will receive through consuming X and Y in exactly the same way. Your earnings in a period are thendetermined by the sum of your earnings through consuming X, Y and V. Thus:
Earnings in period = V total value + X total value + Y total value
You are already familiar with reading the Redemption value table. We now ask you to answer thefollowing questions using the table that you have received.
Question 1
Suppose that in a certain period you have not sold 98 units of V during phase 1. Further suppose thatyou have bought 7 units of Y and 70 units of X during phase 2. Therefore, at the end of this period youhave 98 units of V, 7 units of Y and 70 units of X. What are your total earnings for this period?
Your answer: .........................................................................................
Question 2
What are your total earnings in this period if you have 0 units of Y instead of 7 units at the end of theperiod?
Your answer: .........................................................................................
Question 3
Suppose that in a certain period you have not sold 14 units of V during phase 1. Further suppose thatyou have bought 49 units of Y and 21 units of X during phase 2. Therefore, at the end of the period youhave 14 units of V, 49 units of Y and 21 units of X. What are your total earnings for this period?
Your answer: .........................................................................................
Warnings: In the �rst place, note that you will earn nothing if you do not consume any units of acertain good (V, X or Y), regardless of the number of other goods you consume. Also note from the �rstrow of the Redemption value table that you will only start earning money once you have more than 1unit of a good.Furthermore, note that unsold units of W and leftover cash at the end of phase 2 will not provide youwith any earnings.Finally note that you will come across market prices as well as inclusive prices. Market prices are of directimportance for Type-I traders. Inclusive prices are of direct importance for Type-II traders because theseprices include the taxes or subsidies that apply to them.
REGISTRATION TABLE
Take the Registration table out of the envelope in front of you. The top of this table gives you personalinformation concerning: your type, the markets on which you can trade, your exchange rate, your initialendowments, the subsidies you will receive and/or the taxes you have to pay, as well as other relevantsubsidies and/or taxes. In the table you should register your sales, your purchases and your earnings foreach period. You are already familiar with �lling in this table. We would just like to remind you thatyou should �ll in the number of unsold units of V - in the �rst row under "your sales" - in the pausebetween phases 1 and 2. You should �ll in the rest in the pause at the end of the period. You can usethe example in the grey column to help you with this.
Foreign producers in the wage tax and Van Elswijk system
Speci�c Instructions for Type-II traders
Each trading period consists of two phases: phase 1 and phase 2. In all trading periods you will eitherbe a producer of X or a producer of Y. Your Registration table and Production table indicates whetheryou are an X producer or a Y producer.
B.1. INSTRUCTIONS 101
Phase 1:
In each period, you will receive an initial endowment of cash (an initial cash) at the beginning of phase1. Your initial cash, which is mentioned on the Registration table, is the same in each period.
In phase 1, the markets are open for V and W and you are free to buy units of these goods from whoeverwishes to sell them. If you buy products your endowment of cash decreases. In the �rst place, youwill lose francs because of the costs, which are determined by the market price at which you buy andthe number of units that you buy; these costs determine the purchasing costs at market prices. In thesecond place, you have to pay taxes on top of the market price of good V. The tax rate, expressed as apercentage of the market price of V, is mentioned on your Registration table. There are no taxes on topof the market price of good W. Your purchasing costs at market prices plus the taxes you have to paydetermine your total purchasing costs. These are equal to the purchasing costs at inclusive prices. If youwish to buy a units of V or W you need to have a suÆcient endowment of cash to pay the purchasingcosts at inclusive prices.
Your total cash inventory in francs at the end of phase 1 is equal to:
Cash inventory at end of phase 1 = initial cash - purchasing costs at market prices - taxes = initial cash- purchasing costs at inclusive prices
Warnings: You need both V and W to be able to produce units of X or Y, which you can sell in phase2 of the trading period in order to make a pro�t. Your earnings in a period are only determined by thepro�t you make in phase 2.
The total number of units of V and W that you have in inventory at the end of phase 1 determines howmany units of X or Y you will produce. Your Production table mentions how many units of X or Y youwill produce with a given quantity of V and W.
You are already familiar with reading the Production table. Please answer the following questions usingthe table you have received.
Question 1
Suppose that in a certain period you have bought 57 units of V and 51 units of W during phase 1.Therefore, at the end of this phase you have 57 units of V and 51 units of W in inventory. What is yourlevel of production?
Your answer: .........................................................................................
Question 2
What is your production level if you have 0 instead of 57 units of V in inventory at the end of phase 1?
Your answer: .........................................................................................
Question 3
Suppose that in a certain period you have bought 25 units of V and 29 units of W during phase 1.Therefore, at the end of this phase you have 25 units of V and 29 units of W in inventory. What is yourlevel of production?
Your answer: .........................................................................................
Your level of production is determined automatically by the computer at the end of phase 1.
Warnings: Note that if you do not buy any units of V and W (that is, you buy nothing) in phase 1,you do not produce anything and you do not make any pro�t or loss. If you do not buy any unit of Vor W (that is, you buy units of V only or units of W only) in phase 1, then you still do not produceanything but you do make a loss because of the total purchasing costs for the units bought.
Phase 2:
In phase 2 the markets for X and Y are open. As a seller of X or Y, your trading is restricted to one ofthese markets. In this phase, you are free to sell units of your production inventory of X or Y to whoever
102 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
wishes to buy them. If you sell goods you receive francs. Your sales revenues depend on the market priceat which you sell and the number of units you sell. This determines your sales revenues at market priceswhich are the same as your sales revenues at inclusive prices.
Your pro�ts, which determine your earnings in a period, consist of your sales revenues at market pricesin phase 2 minus your purchasing costs of V and W at market prices in phase 1 minus the taxes youpay for V in phase 1. Your earnings in a period are therefore equal to:
Earnings in a period = sales revenues at market prices - purchasing costs at market prices - taxes= sales revenues at inclusive prices - purchasing costs at inclusive prices
Warnings: Note that the units of X and Y that you do not sell in phase 2 will not provide you withany earnings. Furthermore, note that if you do not wish to make a loss in a period, you must try not tosell your inventory for less than the purchasing costs (including taxes). In other words, you should notspend more on the purchase of V and W in phase 1 than you expect to earn back via the sales revenuesin phase 2.
The information between brackets of the following paragraph was only given in the Van Elswijk system
Finally, note that on each separate market, all traders of a certain type (Type-I and Type-II) are facedwith the same tax rates and/or subsidies regarding the good that is being traded on that market[, excepton market X1. On market X1, some Type II traders have to pay a tax on top of the market price of X,but others do not. The result is that their inclusive prices will di�er].
REGISTRATION TABLE
Take the Registration table out of the envelope in front of you. The top of this table gives you personalinformation concerning: your type, the markets on which you can trade, your exchange rate, your initialendowment, the subsidies you will receive and/or the taxes you have to pay, as well as other relevantsubsidies and/or taxes. In the table you should register your purchases, sales and earnings for each period.You are already familiar with �lling this table. We would just like to point out again that you should �llin the required information in the break at the end of the period. You can use the example in the greycolumn to help you with this.
Domestic producers in the wage tax system
Speci�c instructions for Type II traders
Each trading period consists of two phases: phase 1 and phase 2. In all trading periods you will eitherbe a producer of X or a producer of Y. Your Registration table and Production table indicate whetheryou are an X producer or a Y producer.
Phase 1:
In each period, you will receive an initial endowment of cash (an initial cash) at the beginning of phase1. Your initial cash, which is indicated on the Registration table, is the same in each period. In phase1, the markets are open for V and W and you are free to buy units of these goods from whoever wishesto sell them. If you buy goods your inventory of cash decreases. In the �rst place, you will lose francsbecause of the costs, which is determined by the market price, at which you buy and the number of unitsthat you buy; these costs determine your purchasing costs at market prices. In the second place, youhave to pay taxes on top of the market price of good V. The tax rate, expressed as a percentage of themarket price of V, is mentioned on your Registration table. There are no taxes on top of the market priceof good W. Your purchasing costs at market prices plus the taxes you have to pay determine your totalpurchasing costs. These are equal to the purchasing costs at inclusive prices. If you wish to buy units ofV or W you need to have suÆcient cash in your inventory to pay the purchase costs at inclusive prices.
Your total cash inventory in francs at the end of phase 1 is equal to:
B.1. INSTRUCTIONS 103
Cash inventory at end of phase 1 = initial cash - purchasing costs at market prices - taxes= initial cash - purchasing costs at inclusive prices
Warning: You need both V and W to be able to produce units of X or Y, which you can sell in phase2 of the trading period in order to make a pro�t. Your earnings in a period are only determined by thepro�t you make in phase 2.
The total number of units of V and W that you have in your inventory at the end of phase 1 determineshow many units of X or Y you will produce. Your Production table shows how many units of X or Y youwill produce with a given quantity of V and W.
You are already familiar with reading the Production table. We ask to answer the following questionsusing the table you have received.
Question 1
Suppose that in a certain period you have bought 19 units of V and 12 units of W during phase 1.Therefore, at the end of this phase you have 19 units of V and 12 units of W in your inventory. What isyour level of production?
Your answer: .........................................................................................
Question 2
What is your production level if you have 0 instead of 19 units of V in your inventory at the end of phase1?
Your answer: .........................................................................................
Question 3
Suppose that in a certain period you have bought 5 units of V and 3 units of W during phase 1. Therefore,at the end of this phase you have 5 units of V and 3 units of W in your inventory. What is your level ofproduction?
Your answer: .........................................................................................
Your level of production is determined automatically by the computer at the end of phase 1.
Warnings: Note that if you do not buy any units of V and W (that is, you buy nothing) in phase 1,you do not produce anything and you do not make any pro�t or loss. If you do not buy any unit of Vor W (that is, you only buy units of V or units of W) in phase 1, then you still do not produce anythingbut you do make a loss because of the total purchasing costs for the units bought.
Phase 2:
In phase 2 the markets for X and Y are open. As a seller of X or Y, your trading is restricted to one ofthese markets. In this phase, you are free to sell units of your production inventory of X or Y to whoeverwishes to buy them. If you sell goods you receive francs. Your sales revenues depend on the market priceat which you sell and the number of units you sell. This determines your sales revenues at market priceswhich are the same as your sales revenues at inclusive prices.
Your pro�ts, which determine your earnings in a period, consist of your sales revenues at market pricesin phase 2 minus your purchasing costs of V and W at market prices in phase 1 minus the taxes youpay for V in phase 1. Your earnings in a period are therefore equal to:
Earnings in period = sales revenues at market prices - purchasing costs at market prices - taxes= sales revenues at inclusive prices - purchasing costs at inclusive prices
Warnings: Note that the units of X and Y that you do not sell in phase 2 will not provide you withany earnings. Furthermore, note that if you do not wish to make a loss in a period, you must try not tosell your inventory for less than the purchasing costs (including taxes). In other words, you should notspend more on the purchase of V and W in phase 1 than you expect to earn back via the sales revenuesin phase 2.
104 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
REGISTRATION TABLE
Take the Registration table out of the envelope in front of you. The top of this table gives you personalinformation concerning: your type, the markets on which you can trade, your exchange rate, your initialendowments, the subsidies you will receive and/or the taxes you will pay, as well as other relevant subsidiesand/or taxes. In the table you should register your purchases, sales and earnings for each period. Youare already familiar with �lling this table. We would just like to point out again that you should �ll inthe required information in the break at the end of the session. You can use the example in the greycolumn to help you with this.
Domestic producers in the Van Elswijk system
Speci�c instructions for Type II traders
Each trading period consists of two phases: phase 1 and phase 2. In all trading periods you will eitherbe a producer of X or a producer of Y. Your Registration table and Production table indicates whetheryou are an X producer or a Y producer.
Phase 1:
In each period, you will receive an initial endowment of cash (an initial cash) at the beginning of phase1. Your initial cash, which is mentioned on the Registration table, is the same in each period.
In phase 1, the markets are open for V and W and you are free to buy units of these goods from whoeverwishes to sell them. If you buy goods your cash inventory changes. In the �rst place, you will losefrancs because of the costs, which is determined by the market price, at which you buy the goods andthe number of units that you buy; these costs determine your purchasing costs at market prices. On theother hand, your cash inventory increases if you buy units of V, as a result of a subsidy on V. You receivea subsidy for each unit of V that you buy. The amount of the subsidy is mentioned on your Registrationtable. There is no subsidy for units of W. Your purchasing costs at market prices minus the subsidiesyou receive together determine your total purchasing costs, which are equal to the purchasing costs atinclusive prices. If you wish to buy units of V or W you need to have suÆcient cash in your inventory topay the purchasing costs at inclusive prices.
Your total cash inventory in francs at the end of phase 1 is equal to:
Cash inventory at end of phase 1 = initial cash - purchasing costs at market prices + subsidies= initial cash - purchasing costs at inclusive prices
Warning: You need both V and W to be able to produce units of X or Y, which you can sell in phase2 of the trading period in order to make a pro�t. Your earnings in a period are only determined by thepro�t you make in phase 2.
The total number of units of V and W that you have in inventory at the end of phase 1 determine howmany units of X or Y you will produce. Your Production table shows how many units of X or Y you willproduce with a given quantity of V and W.
You are already familiar with reading the Production table. We ask you to answer the following questionsusing the table you have received.
Question 1
Suppose that in a certain period you have bought 19 units of V and 12 units of W during phase 1.Therefore, at the end of this phase you have 19 units of V and 12 units of W in your inventory. What isyour level of production?
Your answer: .........................................................................................
Question 2
B.1. INSTRUCTIONS 105
What is your production level if you have 0 instead of 19 units of V in your inventory at the end of phase1?
Your answer: .........................................................................................
Question 3
Suppose that in a certain period you have bought 5 units of V and 3 units of W during phase 1. Therefore,at the end of this phase you have 5 units of V and 3 units of W in your inventory. What is your level ofproduction?
Your answer: .........................................................................................
Your level of production is determined automatically by the computer at the end of phase 1.
Warning: Note that if you do not buy any units of V and W (that is, you buy nothing) in phase 1, youdo not produce anything and you do not make any pro�t or loss. If you do not buy any unit of V or W(that is, you only buy units of V or units of W) in phase 1, then you still do not produce anything butyou do make a loss because of the total purchasing costs for the units bought.
Phase 2:
In phase 2 the markets for X and Y are open. As a seller of X or Y, your trading is restricted to oneof these markets. In this phase, you are free to sell units of your production inventory of X or Y towhoever wishes to buy them. If you sell goods you receive francs. Your sales revenues depend on themarket price at which you sell and the number of units you sell. This determines your sales revenues atmarket prices. However, note that you will have to pay a tax on top of the market price of the good thatyou sell. The tax rate, expressed as a percentage of the market price of the good, is mentioned on yourRegistration table. The di�erence between your sales revenues at market prices and the taxes you haveto pay determine your sales revenues at inclusive prices.
Your pro�ts, which determine your earnings in a period, consist of your sales revenues at market pricesin phase 2 minus the taxes you have to pay in phase 2 minus your purchasing costs at market prices ofV and W in phase 1 plus the subsidies that you receive for V in phase 1. Your earnings in a period aretherefore equal to:
Earnings in a period = sales revenues at market prices - purchasing costs at market prices - taxes +subsidies = sales revenues at inclusive prices - purchasing costs at inclusive prices
Warnings: Note that the units of X and Y that you do not sell in phase 2 will not provide you withany earnings. Furthermore, note that if you do not wish to make a loss in a period, you must try not tosell your inventory for less than the purchasing costs (including taxes). In other words, you should notspend more on the purchase of V and W in phase 1 than you expect to earn back via the sales revenuesin phase 2.Finally, note that on each separate market, all traders of a certain type (Type-I or Type-II) are facedwith the same tax rates and/or subsidies regarding the good that is being traded on that market, excepton market X1. On market X1, some Type-II traders have to pay a tax on top of the market price of X,but others do not. The result is that their inclusive prices will di�er.
REGISTRATION TABLE
Take the Registration table out of the envelope in front of you. The top of this table gives you personalinformation concerning: your type, the markets on which you can trade, your exchange rate, your initialendowments, the subsidies you will receive and/or the taxes you have to pay, as well as other relevantsubsidies and/or taxes. In the table you should register your purchases, sales and earnings for each period.You are already familiar with �lling this table. We would just like to point out again that you should �llin the required information in the break at the end of the session. You can use the example in the greycolumn to help you with this.
106 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
B.2 Registration tables
Domestic consumers in wage tax system (selection)
Your type: Type-IMarkets on which you can trade: V1, W1, X1, Y1Your exchange rate: 10 francs = 36 cents
Your initial endowment:
Inititial cash 181Initial endowment V 15Initial endowment W 10
Subsidies you receive: Other relevant subsidies: Type-II traders receive
for every not sold unit V: 70 noneTaxes you have to pay: Other relevant taxes: Type-II traders pay
tax on market price V: 37.8% N.B. Beginning with period 9none these taxes can change
practice practice
example period 1 period 2 period 1
Your sales good good good goodV W V W V W V W
. . . .
not sold units V A . . . .
. . . .
total V subsidy B . . . .
sold units C Eaverage sales price(in market prices) D F
total sales proceeds G=B+C*D+E*F
end cash phase 1 H=initial cash+G
Your purchases good good good goodX Y X Y X Y X Y
bought units I Javerage purchase price(in market prices) K L
total purchase costs M=I*K+J*L
end cash N=H-M
Your earnings good good good goodV X Y V X Y V X Y V X Y
number of units =A =I =Jtotal valueV, X, Y separately O P Q
earnings R=O+P+Q none none
B.2. REGISTRATION TABLES 107
Foreign consumers in wage tax system (selection)
Your type: Type-IMarkets on which you can trade: V2, W1, X1, Y2Your exchange rate: 10 francs = 18 cents
Your initial endowment:
Inititial cash 1268Initial endowment V 105Initial endowment W 70
Subsidies you receive: Other relevant subsidies: Type-II traders receive
for every not sold unit V: 70 noneTaxes you have to pay: Other relevant taxes: Type-II traders pay
tax on market price V: 37.8% N.B. Beginning with period 9none these taxes can change
practice practice
example period 1 period 2 period 1
Your sales good good good goodV W V W V W V W
. . . .
not sold units V A . . . .
. . . .
total V subsidy B . . . .
sold units C Eaverage sales price(in market prices) D F
total sales proceeds G=B+C*D+E*F
end cash phase 1 H=initial cash+G
Your purchases good good good goodX Y X Y X Y X Y
bought units I Javerage purchase price(in market prices) K L
total purchase costs M=I*K+J*L
end cash N=H-M
Your earnings good good good goodV X Y V X Y V X Y V X Y
number of units =A =I =Jtotal valueV, X, Y separately O P Q
earnings R=O+P+Q none none
108 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
Domestic consumers in Van Elswijk system (selection)
Your type: Type-IMarkets on which you can trade: V1, W1, X1, Y1Your exchange rate: 10 francs = 37 cents
Your initial endowment:
Inititial cash 181Initial endowment V 15Initial endowment W 10
Subsidies you receive: Other relevant subsidies: Type-II traders receive
for every not sold unit V: 70 for every bought unit V: 70Taxes you have to pay: Other relevant taxes: Type-II traders pay
tax on market price X: 65.2%� � Not all Type-II traders pay taxnone on market price X
tax on market price Y: 75.2% N.B. Beginning with period 9these taxes can change
practice practice
example period 1 period 2 period 1
Your sales good good good goodV W V W V W V W
. . . .
not sold units V A . . . .
. . . .
total V subsidy B . . . .
sold units C Eaverage sales price(in market prices) D F
total sales proceeds G=B+C*D+E*F
end cash phase 1 H=initial cash+G
Your purchases good good good goodX Y X Y X Y X Y
bought units I Javerage purchase price(in market prices) K L
total purchase costs M=I*K+J*L
end cash N=H-M
Your earnings good good good goodV X Y V X Y V X Y V X Y
number of units =A =I =Jtotal valueV, X, Y separately O P Q
earnings R=O+P+Q none none
B.2. REGISTRATION TABLES 109
Foreign consumers in Van Elswijk system (selection)
Your type: Type-IMarkets on which you can trade: V2, W1, X1, Y2Your exchange rate: 10 francs = 17 cents
Your initial endowment:
Inititial cash 1268Initial endowment V 105Initial endowment W 70
Subsidies you receive: Other relevant subsidies: Type-II traders receive
for every not sold unit V: 70 noneTaxes you have to pay: Other relevant taxes: Type-II traders pay
tax on market price X: 65.2%� � Not all Type-II traders pay taxnone on market price X
tax on market price V: 37.8% N.B. Beginning with period 9these taxes can change
practice practice
example period 1 period 2 period 1
Your sales good good good goodV W V W V W V W
. . . .
not sold units V A . . . .
. . . .
total V subsidy B . . . .
sold units C Eaverage sales price(in market prices) D F
total sales proceeds G=B+C*D+E*F
end cash phase 1 H=initial cash+G
Your purchases good good good goodX Y X Y X Y X Y
bought units I Javerage purchase price(in market prices) K L
total purchase costs M=I*K+J*L
end cash N=H-M
Your earnings good good good goodV X Y V X Y V X Y V X Y
number of units =A =I =Jtotal valueV, X, Y separately O P Q
earnings R=O+P+Q none none
110 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
Domestic X-producers in wage tax system (selection)
Your type: Type-IIMarkets on which you can trade: V1, W1, X1Your exchange rate: 10 francs = 28 cents
Your initial endowment:
Inititial cash 1223
Subsidies you receive: Other relevant subsidies: Type-I traders receive
none for every not sold unit V: 70Taxes you have to pay:
tax on market price V: 37.8%N.B. Beginning with period 9this tax can change
practice practice
example period 1 period 2 period 1
Your purchases good good good goodV W V W V W V W
bought units A Baverage purchase price(in inclusive prices) C D
total purchasing costs E=A*C+B*D
total production H
average costs I=E/H
Your sales good good good goodX X X X
sold units Javerage sales price(in inclusive prices) K
total sales proceeds L=J*K
Your earnings M=L-E none none
B.2. REGISTRATION TABLES 111
Domestic Y-producers in wage tax system (selection)
Your type: Type-IIMarkets on which you can trade: V1, W1, Y1Your exchange rate: 10 francs = 42 cents
Your initial endowment:
Inititial cash 815
Subsidies you receive: Other relevant subsidies: Type-I traders receive
none for every not sold unit V: 70Taxes you have to pay:
tax on market price V: 37.8%N.B. Beginning with period 9this tax can change
practice practice
example period 1 period 2 period 1
Your purchases good good good goodV W V W V W V W
bought units A Baverage purchase price(in inclusive prices) C D
total purchasing costs E=A*C+B*D
total production H
average costs I=E/H
Your sales good good good goodY Y Y Y
sold units Javerage sales price(in inclusive prices) K
total sales proceeds L=J*K
Your earnings M=L-E none none
112 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
Foreign X-producers in wage tax system (selection)
Your type: Type-IIMarkets on which you can trade: V2, W1, X1Your exchange rate: 10 francs = 4 cents
Your initial endowment:
Inititial cash 8557
Subsidies you receive: Other relevant subsidies: Type-I traders receive
none for every not sold unit V: 70Taxes you have to pay:
tax on market price V: 37.8%N.B. Beginning with period 9this tax can change
practice practice
example period 1 period 2 period 1
Your purchases good good good goodV W V W V W V W
bought units A Baverage purchase price(in inclusive prices) C D
total purchasing costs E=A*C+B*D
total production H
average costs I=E/H
Your sales good good good goodX X X X
sold units Javerage sales price(in inclusive prices) K
total sales proceeds L=J*K
Your earnings M=L-E none none
B.2. REGISTRATION TABLES 113
Foreign Y-producers in wage tax system (selection)
Your type: Type-IIMarkets on which you can trade: V2, W1, Y2Your exchange rate: 10 francs = 6 cents
Your initial endowment:
Inititial cash 5705
Subsidies you receive: Other relevant subsidies: Type-I traders receive
none for every not sold unit V: 70Taxes you have to pay:
tax on market price V: 37.8%N.B. Beginning with period 9this tax can change
practice practice
example period 1 period 2 period 1
Your purchases good good good goodV W V W V W V W
bought units A Baverage purchase price(in inclusive prices) C D
total purchasing costs E=A*C+B*D
total production H
average costs I=E/H
Your sales good good good goodY Y Y Y
sold units Javerage sales price(in inclusive prices) K
total sales proceeds L=J*K
Your earnings M=L-E none none
114 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
Domestic X-producers in Van Elswijk system (selection)
Your type: Type-IIMarkets on which you can trade: V1, W1, X1Your exchange rate: 10 francs = 78 cents
Your initial endowment:
Inititial cash 1223
Subsidies you receive: Other relevant subsidies: Type-I traders receive
for every bought unit V: 70 for every not sold unit V: 70Taxes you have to pay:
tax on market price X: 65.2%�
�Not all Type-II traders pay atax on market price XN.B. Beginning with period 9this tax can change
practice practice
example period 1 period 2 period 1
Your purchases good good good goodV W V W V W V W
bought units A Baverage purchase price(in inclusive prices) C D
total purchasing costs E=A*C+B*D
total production H
average costs I=E/H
Your sales good good good goodX X X X
sold units Javerage sales price(in inclusive prices) K
total sales proceeds L=J*K
Your earnings M=L-E none none
B.2. REGISTRATION TABLES 115
Domestic Y-producers in Van Elswijk system (selection)
Your type: Type-IIMarkets on which you can trade: V1, W1, Y1Your exchange rate: 10 francs = 134 cents
Your initial endowment:
Inititial cash 815
Subsidies you receive: Other relevant subsidies: Type-I traders receive
for every bought unit V: 70 for every not sold unit V: 70Taxes you have to pay:
tax on market price Y: 75.2%N.B. Beginning with period 9this tax can change
practice practice
example period 1 period 2 period 1
Your purchases good good good goodV W V W V W V W
bought units A Baverage purchase price(in inclusive prices) C D
total purchasing costs E=A*C+B*D
total production H
average costs I=E/H
Your sales good good good goodY Y Y Y
sold units Javerage sales price(in inclusive prices) K
total sales proceeds L=J*K
Your earnings M=L-E none none
116 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
Foreign X-producers in Van Elswijk system (selection)
Your type: Type-IIMarkets on which you can trade: V2, W1, X1Your exchange rate: 10 francs = 4 cents
Your initial endowment:
Inititial cash 8557
Subsidies you receive: Other relevant subsidies: Type-I traders receive
none for every not sold unit V: 70Taxes you have to pay:
tax on market price V: 37.8%N.B. Beginning with period 9this tax can change
practice practice
example period 1 period 2 period 1
Your purchases good good good goodV W V W V W V W
bought units A Baverage purchase price(in inclusive prices) C D
total purchasing costs E=A*C+B*D
total production H
average costs I=E/H
Your sales good good good goodX X X X
sold units Javerage sales price(in inclusive prices) K
total sales proceeds L=J*K
Your earnings M=L-E none none
B.2. REGISTRATION TABLES 117
Foreign Y-producers in Van Elswijk system (selection)
Your type: Type-IIMarkets on which you can trade: V2, W1, Y2Your exchange rate: 10 francs = 6 cents
Your initial endowment:
Inititial cash 5705
Subsidies you receive: Other relevant subsidies: Type-I traders receive
none for every not sold unit V: 70Taxes you have to pay:
tax on market price V: 37.8%N.B. Beginning with period 9this tax can change
practice practice
example period 1 period 2 period 1
Your purchases good good good goodV W V W V W V W
bought units A Baverage purchase price(in inclusive prices) C D
total purchasing costs E=A*C+B*D
total production H
average costs I=E/H
Your sales good good good goodY Y Y Y
sold units Javerage sales price(in inclusive prices) K
total sales proceeds L=J*K
Your earnings M=L-E none none
118 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
B.3 Redemption value tables
Domestic consumers in wage tax system and Van Elswijk system
Redemption values table Type-I
N.B. The redemption value is 0 if you have not got at least 1 unit of all the goods (V, X and Y).
unit V unit V total X unit X total Y unit Y totalvalue value value value value value
124 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
B.5 Trading rules
RULES FOR TRADING
1. Traders bid or ask. Bidding involves carrying out a bid order (purchase proposal), consistingof (a) a bid price and (b) the number of units that one wishes to buy at this price. Asking involvescarrying out an ask order (sales proposal), consisting of (a) an ask price and (b) the numberof units that one wishes to sell at this price.
2. Traders can also buy or sell. Buying involves carrying out a buy order, consisting of the numberof units that one wishes to buy at the outstanding (current) ask price. Selling involves carryingout a sell order, consisting of the number of units that one wishes to sell at the outstanding(current) bid price.
3. Only the highest bid price and the lowest ask price are valid on the market irrespective of thenumber of units that one wishes to buy (demand) or sell (supply), respectively, at this price.
4. Transactions can involve either the total amount supplied or demanded on the market or part ofit.
5. If the total amount supplied on the market is bought, then the market is open for new askorders as well as new bid orders, and any price is permitted. If the total amount demanded onthe market is sold, then the market is open for new ask orders as well as new bid orders, andany price is permitted.
6. If only part of the amount demanded is sold, then the amount left remains on the marketat the current price. If only part of the amount supplied is bought, then the amount leftremains on the market at the current price.
7. If a buy order is for more than the amount supplied on the market, then the buying traderreceives the amount supplied on the market. The remaining units are regarded as a new purchaseproposal at the current price. If a selling order is for more than the amount demanded onthe market, then the trader selling sells the amount demanded on the market. The remainingunits are regarded as a new sales proposal at the current price.
8. There is another way to buy. If a newly introduced bid price is higher than the current askprice, trading automatically takes place at the current ask price. If such a purchase proposal isfor more than the current amount, then the amount remaining is dealt with as a new purchaseproposal at the matching price.There is another way to sell. If a newly introduced ask price is lower than the current bid price,trading automatically takes place at the current bid price. If such a sales proposal is for morethan the current amount, then the amount remaining is dealt with as a new sales proposal at thematching price.
B.6. COMPUTER SCREENS 125
B.6 Computer Screens
126 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
B.6. COMPUTER SCREENS 127
128 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
B.6. COMPUTER SCREENS 129
130 APPENDIX B. INSTRUCTIONS, TRADING RULES, COMPUTER SCREENS
Appendix C
Complete summary of the resultsof all sessions and all periods
where yit is the dependent variable, B1i (i = 1; 2; 3) the session intercepts, B2 the commonasymptote, Di (i = 1; 2; 3) session dummies, t the trading period (t = 1; :::; 8), and u an errorterm.
Table D.1: Convergence results for constant tax periods un-der wage tax system
166 APPENDIX D. CONVERGENCE ANALYSIS FOR THE WAGE TAX SYSTEM
Appendix E
Detailed results of hypotheses
testing
167
168 APPENDIX E. DETAILED RESULTS OF HYPOTHESES TESTING
Hypothesis 1: Capital ight
Null hypothesis: Implementing the Van Elswijk system will not lead to capital ight.VE hypothesis: Same as the null hypothesis.SC hypothesis: Capital ight will take place.
Hypothesis 1.1 Capital inputt-test:di�erence = value system A� value system BNull : di�erence � 0;VE : di�erence � 0; SC : di�erence > 0:
trading period 0.0221** -0.0058* -0.0054(0.008) (0.003) (0.004)
System B -0.0384* -0.0816**(0.020) (0.015)
wage tax rate 0.2569**home country (0.045)wage tax rate -0.0655*foreign country (0.036)tax rate product X 0.1620**
(0.046)N 72 162 162R2 0.1284 0.1771 0.3532
Note: * signi�cant at 10%, ** signi�cant at 5%; standard errors in parentheses.
170 APPENDIX E. DETAILED RESULTS OF HYPOTHESES TESTING
Hypothesis 2: Labour employment
Null hypothesis: Implementing the Van Elswijk system will not lead to a change in total labouremployment.VE hypothesis: Labour employment increases.SC hypothesis: Labour employment decreases.
t-test:di�erence = value system A� value system BNull : di�erence = 0;VE : di�erence < 0; SC : di�erence > 0:
trading period 0.0067 0.0467 0.0727(0.244) (0.081) (0.0891)
System B 0.7 1.7852**(0.561) (0.408)
wage tax rate -4.2588**home country (1.060)wage tax rate 0.3094foreign country (1.267)tax rate product X -1.6193
(1.617)N 120 270 270R2 0.0131 0.0682 0.1192
Note: * signi�cant at 10%, ** signi�cant at 5%; standard errors in parentheses.
171
Hypothesis 3: Net wage
Null hypothesis: Implementing the Van Elswijk system will not lead to a change in net wages.VE hypothesis: Net wages do not decrease.SC hypothesis: Net wages decrease.
t-test:di�erence = value system A� value system BNull : di�erence = 0;VE : di�erence � 0; SC : di�erence > 0:
trading period 0.0028* -0.0000 0.0023**(0.001) (0.001) (0.001)
System B -0.0069** 0.0169**(0.003) (0.003)
wage tax rate -0.0252 **home country (0.010)wage tax rate -0.0577**foreign country (0.013)tax rate product X 0.02806*
(0.016)N 320 736 736R2 0.0251 0.0519 0.2239
Note: * signi�cant at 10%, ** signi�cant at 5%; standard errors in parentheses.
172 APPENDIX E. DETAILED RESULTS OF HYPOTHESES TESTING
Hypothesis 4: Shifts in production
Null hypothesis: Implementing the Van Elswijk system will not lead to a shift of the relativeproduction levels of the sectors.VE hypothesis: Conditional.SC hypothesis: The capital intensive sector will get smaller in relation to the labour intensivesector.
t-test:di�erence = value system A� value system BNull : di�erence = 0;VE : conditional; SC : di�erence < 0:
trading period -0.0001 0.0014 0.0012(0.007) (0.002) (0.003)
System B 0.0074 0.0155(0.017) (0.012)
wage tax rate -0.0339home country (0.032)wage tax rate 0.0123foreign country (0.039)tax rate product X -0.0159
(0.046)N 72 162 162R2 0.0028 0.0118 0.0147
Note: * signi�cant at 10%, ** signi�cant at 5%; standard errors in parentheses.
173
Hypothesis 5: Welfare
Null hypothesis: Implementing the Van Elswijk system will not lead to a change in welfare.VE hypothesis: Implementing leads to an increase in welfare.SC hypothesis: Implementing leads to a decrease in welfare.
Hypothesis 5.1 Utilityt-test:di�erence = value system A� value system BNull : di�erence = 0;VE : di�erence < 0; SC : di�erence > 0:
trading period 13.283 -3.203 -7.516(19.72) (6.84) (10.10)
System B 298.011** 201.29**(47.18) (35.78)
wage tax rate -302.936**home country (153.48)wage tax rate 216.286foreign country (188.49)tax rate product X -54.680
(218.93)N 192 432 432R2 0.1758 0.0691 0.0677
Note: * signi�cant at 10%, ** signi�cant at 5%; standard errors in parentheses.
176 APPENDIX E. DETAILED RESULTS OF HYPOTHESES TESTING
Hypothesis 5.4 Gross national product (GNP)
t-test:di�erence = value system A� value system BNull : di�erence = 0;VE : di�erence < 0; SC : di�erence > 0:
GNP signi�cance
period system A system B di�erence<0 di�erence>0
5 to 8 19.212 22.236 ** ns13 to 16 18.073 23.933 ** ns
Note: \ns" not signi�cant, * signi�cant at 10%, ** signi�cant at 5%.
Appendix F
Production eÆciency and
FGLS-estimates
177
178 APPENDIX F. PRODUCTION EFFICIENCY AND FGLS-ESTIMATES
Production eÆciency
The values in the tables give the number of units produced, per trading period, experimentalsession, and producer type, divided by the number of units that could have been produced ifthis occurred eÆciently. That is, if each producer had produced the same amount.
wherek h = home country; h = foreign countryz production sector z = x; ys sessiont period
rs;t nominal capital pricewk;s;t(1 + �w;k;s;t)� Tk;s;t nominal gross (after tax/subsidy) wagewk;s;t nominal wage�w;k;s;t wage tax rateTk;s;t employment subsidy per employed unit of labourLkz;s;t number of units of labourKkz;s;t number of units of capital
ln(Lkz;s;t=Kkz;s;t) natural logarithm of labour-capital ratioln rs;t natural logarithm of capital priceln(wk;s;t(1 + �w;k;s;t)� Tk;s;t) natural logarithm of gross wage
Table F.8: FGLS-regressions for system B
home country X home country Y foreign country X foreign country Y