Get to Know your Bureaucracies. Mapping Subnational Bureaucratic Bodies with Evidence from Official Household Surveys. Argentina2003-2011

MPSA 2015 KnowBureaucracies Moscovich

1

Get to Know your Bureaucracies. Mapping Subnational Bureaucratic Bodies with

Evidence from Official Household Surveys.Argentina2003-2011

Lorena Moscovich

Universidad de San Andrés*

Work in progress and preliminary version. Please, do not quote without permission.

I would appreciate to hear your comments and suggestions

[email protected]

Paper prepared for the Midwest Political Science Association Annual Meeting

Chicago, April 15-18, 2015

mailto:[email protected]


2

Introduction1

According to Cortazar Velarde, Lafuente and Sangines (2014) in 2012, on

average, 16 countries of Latin America have invested 5.8% of their GDP, around 40%

of tax revenues and 26% of public expenditure to pay salaries of bureaucrats.

Bureaucrats are crucial for state capacity; from national defense to policy success; from

the functioning of democratic process to property and citizen rights enforcement. The

characteristics and quality of bureaucratic human capital have a number of implications

for policy making, economic growth, democratic governance and citizens rights, among

others. However, beyond public policy analysis, bureaucracies do not seem to matter for

political scientists and comparative politics (Fukuyama 2012). Moreover we lack good

indicators of the characteristics and the quality of work that public servants perform

(Fukuyama 2013). Reliable indicators that allow us to classify bureaucratic bodies

across different political units and in time would help us to understand their influence

on the issues above mentioned.

Two simple claims structure this paper, the first one is that we can have a better

understanding of bureaucratic bodies if we know their characteristics. In other words if

we want to know more about bureaucracies we should ask bureaucrats to answer

questions such as what are your qualifications? How many hours do you work? Is this

your only job? Etc. The second is that we can use information from country statistics

with this aim. I see only advantages in complementing existing measures with

information from surveys made regularly by national statistics institutes using random

samples of bureaucrats and, such as household surveys. The data is available, standard,

public, free, official, and potentially useful for time series cross sectional comparisons

at national and subnational levels.

I propose a map of bureaucracies at the subnational level in, Argentina, a

federal country considering information available in the Permanent Household Survey

(Encuesta Permanente de Hogares) such as education, income, career, qualification,

permanence, the use of technology, hours worked, exclusivity, hierarchy and turnover.

1 I want to thank Ernesto Calvo for his insights and suggestions and to Federico Merke for comments

made on an earlier draft. This paper would not have been possible without the superb research assistance

of Marcos Salgado. Final results, omissions and/or errors are my sole responsibility.


3

This set of variables allowed me to cover the main features of bureaucracies. Mapping

the distances between the bureaucratic bodies in the provinces in relation to these

variables, without an “a priory” or value attached to the performance of these variables,

allowed me to describe them and also to identify some outcomes not necessarily

suggested by the weberian (and hardly real -Migdal 2011-) ideal of autonomous

bureaucracies with well-paid qualified professionals that guarantee the quality of policy

making. Results show some interesting findings. For instance, that under some

circumstances, a stable and exclusive position does not always make bureaucrats

perform better in comparison with other bureaucratic bodies with a higher turnover rate.

This happens when stability and exclusivity do not come hand in hand with

qualifications. Another finding is that tenured positions do not always shape bureaucrats

expectations, as a result a tenured position can neither foster long term horizons nor

improve policy making, unless appointments are programmatic.

Scholars in the field of comparative politics, public policy and development

studies, also students of bureaucracy and administrative state capacity, should be able to

count on systematic data produced by national statistics institutes, to use as a unique

source or combined with other measures of bureaucratic quality. In sum I suggest the

need of public official data to use as reliable indicators of bureaucracies. To do this I

first review some of the most important sources of indicators for bureaucratic quality

and I give an example of one of the multiple possible uses of this data.

What is Good Bureaucracy Good for?

The combination of coercion and consensus has been acknowledged as the

Weberian formula for successful legitimate domination. The third part of this formula is

administration. Weber‟s (1995) seminal analyses of bureaucracies sometimes are taken

in isolation and not connected with his studies on political legitimacy. Bureaucracies are

at the core of the studies on political order. Bureaucracies are the kind of administrative

bodies which accompany modern states and they are necessary for legitimate

democratic governments to reach all their citizens and social relations within their

territory. As Weber suggests, while politicians give the strategic direction and content to


4

policy making, bureaucrats ideally apply the best tools to help them achieve expected

results. Already in Weber‟s time, bureaucracies also affected public policy priorities.

Thus bureaucracies have always mattered for democracies, state capacity, policymaking

and political order.

Scholars on state capacity acknowledge the important role of bureaucracies

(Hendrix 2010). Bureaucratic capacity is one of the central components of

administrative state capacity (Back and Hadenius 2008, Soifer and Vom Haw 2008),

and of its institutional strength (Kurtz 2013, Levitsky and Murillo 2009).1 Authors like

Charron and Lapuente (2010) use the level of corruption and bureaucratic quality to

assess the performance of state capacity in both democratic and authoritarian regimes.

Knutsen, (in Cingolani 2013) links regime type and state capacity to growth; he studied

state capacity through policy implementation and bureaucratic professionalism.

For Evans (1995) professional bureaucracies are a precondition for a

developmental state. Their features are directly linked to economic growth (Olson et. al.

2000) and markets (Wibbels 2005). A meritocratic, well paid and stable bureaucracy

will be less prone to corruption and will have more incentives to provide state

infrastructure (En Bai and Jin Wei 2000). As a result, a country with such bureaucracy

will be preferred for investments and this country will be more likely to grow and

develop (Evans and Rauch 1999). Rothstein (2005) stresses that social trust is fostered

by good bureaucracies. When bureaucracies deliver policies on programmatic, public

and universal basis, citizens understand that discretion is low. It increases their trust in

government because they think they are more likely to have fair and equal opportunities.

Bureaucracies are important for policy success (Berkman et al 2008), from the

control of specific knowledge and the implementation of central government directions

(Agranoff 2001), to the daily contact in the delivery of goods and services needed to

reach citizens (Honig 2006). When bureaucracies fail, within the same political regime,

variations in bureaucratic performance may result in the asymmetric rule of law and the

uneven enforcement of citizen‟s rights (O‟Donnell 2007, Smulovitz 2010), civil

conflicts or even wars (Hendrix 2010). Professional bureaucracies last longer than

governments and are the safeguard of programmatic politics. This is due to the status of

bureaucrats‟ appointment and their management of specific knowledge. It is not the case


5

when patrimonialism rules and bureaucrats access positions through political or

personal favors (Mazzuca 2012).2

We know, like Migdal (2011) suggests, that there are no handbook states in

which bureaucracies have clear limits in society. He stresses the fuzzy borders between

state and society. Ideal bureaucracies do not exist, and when scholars only focus on their

Weberian features it hampers their capacity to understand the conflicting interactions

within government agencies (Arnold 1989) and between state and society actors. In fact

bureaucrats can hinge on society and also on politicians (and legislatures). For Sikkin

and Wolfson (1993) the autonomy from society was a precondition which allowed

Brazil to develop a more professional bureaucracy than Argentina.3Geddes (1996) also

stresses that bureaucratic autonomy is needed for policy success. On the other hand,

bureaucrats depend on politicians appointments and budget approval. However

politicians and bureaucrats can have conflicting interests. The same knowledge

supposed to be a distinctive quality of bureaucracy can results in an undesirable

autonomy when it is used in their favor. They may give government selective

information or ignore its instructions. Politicians give money and directions to

bureaucrats, but they actually lack information about the results of policies and depend

on their expertise (Bendor et. al. 1985).

Main Indicators Used for the Study of Bureaucracy

Researchers on bureaucracy sometimes use their own measures of bureaucratic

quality (some of them build databases), others use indicators from risk agencies,

foundations or international organizations, and a last group of scholars combines

different measures. As you will soon notice, sources on bureaucratic quality are usually

part of broader measures for different aims. These sources usually rely on more or less

standard surveys applied to non-randomized samples of experts (usually a few per

country). To avoid the potential bias implied by these sources, others gather very

different data (i.e. World Bank) and develop specific controls (i.e. Bertelsmann

2 For Gailmard and Patty (2007) there is a situation in which selective incentives distributed discretionally

by legislatures to bureaucrats can explain the choices to increase their expertise. 3 For a more recent comparison between Argentine and Brazilian bureaucracies see Souza 2015.


6

Stiftung's Transformation Index), while others use a variety of indirect indicators, such

as corruption, (Bai and Wei 2000, Dalsthröm and Lapuente 2012, Albornoz and Cabrera

2013) or the politicization of removal of central bank governors (Cingolani et al 2013).

A number of studies gather, comment on and use indicators of bureaucratic

quality (or at least include some of them). Institutional performance, measured trough

bureaucracies, corruption, policy performance or public sector management, is one of

the indicators of good governance included in some of the 52 measures reviewed by the

UNDP (2007). Berkman and her colleagues (2008) did extensive research, gathering

and reorganizing information of different sources to explore the conditions of success of

policy making processes, policy enforcement and implementation. They developed a

number of dimensions; one is the institutional dimension within which they create a

bureaucracy index made of a combination of selected questions from the

Columbia University State Capacity Survey and data from the Political Risk Service

Index.

Hanson and Sigman (2013) use 24 different sources of indicators on state

capacity. They work with data reduction and Bayesian techniques (so do Bersch et al

2013).They also explore the administrative dimension of state capacity understood in

terms of policy success and the quality of bureaucracy.4Cingolani (2013) wrote one of

the most complete literature reviews on state capacity with its different definitions and

operationalization. Also Cingolani reviews the most common issues with which it has

been analyzed, such as development, autonomy from society, etc. Hendrix (2010) also

uses 15 different operationalizations of state capacity (some of them gathered in the

administrative capacity of the state) and introduces a test of dimensionality using factor

analysis.

Van de Walle (2006) focuses on how bureaucratic quality has been addressed by

governance indicators. His study reveals that there are no indicators for bureaucratic

quality in all policy areas for most countries (particularly for OECD countries) since

cross country studies tend to focus on just one policy area. On the other hand he

critically describes main features and uses of the European Central Bank‟s working

paper „„Public Sector Efficiency: An International Comparison‟‟(2003), the World

4 Like Soifer (2013) they use the frequency of Census as proxy for the state capacity. Bersch, Praca and

Taylor (2013) use census data from the International Programs Center of the U.S. Census Bureau


7

Bank‟s „„Government Effectiveness‟‟ indicator, the „„Public Institutions Index‟‟, part of

the World Economic Forum‟s Growth Competitiveness Index, and the International

Institute for Management Development‟s (IMD‟s) World Competitiveness Yearbook‟s

„„Government Efficiency Ranking‟‟ (Van de Walle 2006). He also considers the

measures used by each document, such us the chapter of bureaucracies of the

International country Risk guide of the Political Risk Services or the Transparency

International Corruption Perceptions Index".

Although mainly used for the development studies, aid agencies and investment

decisions, Van de Walle states that the use of indicators for bureaucratic quality for

academic and public policy research, comparative politics, and public policy decisions

has been limited. Reasons underlying this limitation are: lack of specific data on

bureaucracy5, the subjective nature of the sources,

6doubts regarding sampling design

and the selection of sources, confusion regarding the validity of these indicators in

relation to what they study, and the normative implications of the definitions of the

quality of bureaucracy used. .

Many studies review and classify existing indicators used to describe and

measure bureaucratic capacity and its applications for different aims. These studies also

underscore some of their limitations. Alternative measures could complement and

enhance the explanatory power of existing indicators. Following I summarize some of

the advantages that come with the use of different sources, particularly official sources,

and with the use of more simple and parsimonious measures for the assessment of

bureaucracies.

Why yet another Indicator of Bureaucratic Quality?

5 Remember that these indicators are a subset of broader measures of something else, like governance,

state capacity, good government, etc. 6 The five different sources of measures on bureaucratic capacity reviewed by Savoia and Sen (2012) are

also based on expert surveys. It is worth noticing that one of them also gathers standardized data and it is

not intertwined with subjective measures. Fukuyama (2012) suggests that the “Quality of Government

Dataset” initiative (http://qog.pol.gu.se/data) conducted by professor Bo Rothstein from the University of

Gothenburg has promising perspectives in order to overcome some problems of existing measures.

http://qog.pol.gu.se/data


8

I suggest that measures of bureaucracy are indirectly and partially useful in

comparative politics. Main existing indicators of bureaucracies are part of broader

measures, such as governance, state capacity, institutional quality or investment risks,

are based on subjective sources (such as expert surveys that are neither random nor

representative). To solve this potential source of bias, other measures combine or use

different indexes and/or kinds of information. As a result their complexity impedes

understanding what exactly they describe. Sometimes measures do not cover all

countries or years7, and a few are made at a subnational level. Overall, most depend on

particular databases, from research projects and are only updated while the project lasts.

Other databases belong to risk agencies, sometimes unaffordable for researchers and

institutions from developing countries.8

An indicator to describe bureaucracies that can be used as a tool to assess their

influence should meet some conventional requirements for measures in social research:

- Non biased. The indicator or measure must accurately describe the features of existent

bureaucratic bodies. It must have a proper, randomized, large enough N, representative

sample of bureaucrats. This sample seems to be a better source than, or at least a good

companion for, expert surveys.

- Parsimonious. In the way that variables represent relevant features of bureaucratic

quality, such as education, appointment, hierarchy, but at the same time the resulting

indicators are transparent and understandable.

- Comparable. Apart from official data, household surveys of national statistics

institutes are usually standard and (with some limitations) comparable across districts

(at national and subnational level), also many are permanent increasing the temporal

scope of comparisons thanks to the use of longer time series of information. .

- Available. Public and accessible information allows both its use and replication. Of

course the quality of national statistics is diverse and also varies with time, but it is

more likely to be available for researchers. A good measure should not have to depend

on databases not only many of them of subjective nature, but also that do not always

7 As a result some researchers try to replicate these measures for their countries (see Souza 2015)

8 Last footnote applies


9

cover all countries, are not updated, or are expensive National statistics are usually

available, sometimes on request, free of charge (or cheaper).

Surveys and interviews are powerful tools to assess characteristics of

bureaucracies by qualitative means. Expert surveys account for opinions and

interpretations not facts.9 Expert interviews are better interpreted using qualitative

means. Qualitative approaches can contextualize testimonies giving them meaning;

allowing us to know experts‟ reasons or conclusions, and also their particular interest or

position in relation to the topic involved.10

For instance, in some countries experts from

certain universities can support the government or be in the opposition. This is not a

minor problem in the context in which political polarization deeply divides academia.11

When standardized questionnaires are used instead of surveys for cross-country

comparison, indicators may gain reliability because the reply coding is homogeneous.

However researchers who design these questionnaires can omit unknown aspects from

bureaucracies, such as informal practices which vary from country to country. Experts

surveys of any kind barely capture some practices at the local level in which bureaucrats

are hired and work, of course it would be impossible for them to be familiarized with

day to day work and the already mentioned very important informal practices,

particularly in developing countries. Of course if we ask bureaucrats instead of experts

these previously unknown aspects are more likely to become evident.12

This leads to the

last concern; even the best measures and more complete databases mostly have

information at the national level. In highly decentralized countries, such as Argentina,

subnational bureaucracies matter a lot for policy success and enforcement of citizens‟

9 This is such a basic problem of social sciences, we, researchers, work with subjects that have already

understood their world. In this way, as Giddens (1998) suggests, ours is a double hermeneutic. We try to

understand a reality that was already understood. This is not only the big difference between social and

natural sciences, but also a limitation to the scope of our conclusions. Because people also learn from

experience and can change their actions and meanings assigned to their experiences. Moreover they can

read our studies and perform in different ways. In this sense using only expert surveys result in trying to

build scientific knowledge from a reality previously understood by subjects, then understood by experts.

Drawing new conclusions only from expert sources is like working at this third level of understanding,

one level away from facts and reality. 10

Another option is to study state and bureaucracy using a comparative historical path dependence

approach (Arnold 1989). However when one wants to move to the comparative field, the use of these

approaches (and also quantitative ones for one paper or case) is limited.

11 Try to interpret standardized anonymous surveys in countries such as Venezuela and do not fail in the

attempt. 12

For instance, in my analysis I use objective and subjective data on bureaucratic stability whose results

are contradictory. Employees fear losing their jobs, this allows me to infer the uses of informal and

clientelistic appointments which make bureaucrats feel their positions are precarious even in provinces

with lower levels of turnover.


10

rights. Aggregating expert opinions and transforming them into data as if they were

facts can be a source of bias.

Of course this problem is widely acknowledged and some indexes develop

safeguards to avoid potential biases. One is to combine different standard surveys and

conduct interviews to control results (Rauch and Evans 2000). The other possible

solution is adding information to the expert surveys, such as the Quality of Government

Data Set and the Worldwide Governance Indicators (Kaufmann et al 2007) do. The

latter multiplies the indexes in order to reduce this danger, also they are transparent with

their sources, procedures and possible errors. Also the “dark side” of this strategy is that

the many kinds of data disconnect sources from the substantive contents of the variables

Van de Walle (2006) wonders what the aggregation of 31 different indicators actually

tells us ? What they measure is not clear.

Longer time series is another way to reduce errors, sources like the ICRG have

generated data monthly for a large number of countries since the 50‟s. However there

are normative concerns regarding the concept of bureaucratic quality implied in

investments risk evaluation. From this point of view, bureaucratic quality assesses the

bureaucrats ability to be isolated from political turbulence. Is this independence

desirable? Independent from politicians and dependent on business interests? Does it

give bureaucracies too much autonomy from their principals (legislatures or citizens)?

When bureaucrats manage expert knowledge in their fields they can use this asymmetry

in their own interest. A discussion must be given on the normative and practical

implications of some definitions of bureaucratic quality. Otherwise the result could be

giving bureaucracy more power than we should expect in democratic settings. These

questions remain open.

Even when we conclude that although not perfect, these indicators are the best

available they may have other problems. First as long as in order to gain accuracy,

measures became more complex , and researchers totally depend on the institutions that

develop and update them. Some of these measures are public, but others like the

belonging to risk agencies are expensive, limiting their access for independent

researchers or those from developing countries. Moreover when projects for some

reasons finish so do the updates and sometimes the availability of the information.13

I

13

Such as the IRIS project of Maryland University


11

suggest that we can have a better understanding of bureaucratic bodies if we learn their

characteristics by asking bureaucrats. The use of permanent and standardized household

surveys as source of information can both give an alternative indicator to know public

servants, likely, to be replicated and also become a complementary source to control the

validity of existent measures.

How could Official Statistics Improve our Understanding of Bureaucracies?

National statistics are far from perfect and objective sources of information. It is

known that some national statistics institutes for different reasons do not meet basic

quality standards. However most do and the standard and public information they

provide, can help us to improve our understanding of different groups and elites.14

So

far it seems that specialists in national statistics and political scientists live in a parallel

universe (Leuprecht and Goldstone 2013).15

For instance, Rodrigues-Silveira (2014)

highlights the usefulness of census, statistics and more generally a demographic

approach to explore changes in municipal local elites in Brazil.16

Bersch, Praca and

Taylor (2013) also use official information from the Brazilian federal state from a

variety of sources to study the autonomy and capacity of bureaucracies. The databases

they use contain public information on federal officials, updated regularly. However,

their work is based on a variety of sources difficult (but not impossible, of course) to be

replicated or comparable, particularly at the subnational level where the information is

more diverse. Again national standard surveys would help us to overcome this problem.

Food aid programs, public health accounts, migrations, researchers in different

fields, stress the need and validity of national permanent household surveys in order to

produce reliable data. For instance, since no global data exists on middle classes,

14

Domicilary surveys have many kind of errors, field, coding, sampling errors, but we do not expect

respondents to be biased in their replies, such as not giving accurate information regarding their years of

schooling or hours worked. Expert surveys can be systematically biased in a politically polarized country

or when respondent has a personal or private interest in the survey‟s results. 15

Although Leuprecht and Goldstone (2013) focus on political demography, understood as “the study of

the size, composition, and distribution of population in relation to both government and politics” (online,

no page number given) this claim also applies to the use of different systematic sources of information

produced by demographers and national statistics centers, like permanent household surveys. 16

National statistics themselves relate to state capacity, its success is a proxy for the reach of the state

(Soifer 2013, Hanson and Sigman 2013)


12

Banerjee and Duflo (2008) used household survey data from 13 countries to identify

patterns of consumptions and features of middle classes in developing countries. In

particular they are able to compare these surveys because countries included the Living

Standard Measurement Surveys (LSMS) in household surveys. Many studies have used

permanent official household surveys in order to measure income (Milanovic 2002) or

social inequality (Calvo and Moscovich 2013). In his comprehensive analysis to

household surveys, Deaton (1997) distinguishes a number of areas for which they can

be used. He zooms in on methodological concerns relevant in order to successfully

replicate household survey data in these fields, such as saving and consumption,

nutrition and intra-house allocation, welfare and inequality, among others.

The Socio-Economic Database for Latin America and the Caribbean is a joint

initiative from CEDLAS, of the Universidad de la Plata, and The World Bank which

gathers comparable information from permanent household surveys conducted by

national statistics institutes in 24 Latin-American countries

(http://sedlac.econo.unlp.edu.ar/eng/).Household surveys are particularly useful to

collect information in developing countries where there are not as many sources of

systematic and reliable data as in other countries. However research aiming to kick off

comparative endeavors should take into account this fact and consider using the same

sources for developed countries, even having alternative ones, in order to have standard

comparable information for cross-country comparisons .

Household surveys vary across countries however most include data on income,

employment, education and living conditions. Panizza and his coauthors (2001) warn us

about the difficulties arising from the differences in permanent household surveys in

order to study the public sector. However they manage to gather comparable

information from household surveys from 17 Latin-American countries to analyze how

salaries affect public sector quality. I already mentioned the initiative of the SEDLAC

which gathers and harmonizes household surveys from different countries or the use of

the LSC in the work of Banerjee and Duflo (2008). Governments themselves carry out

initiatives in order to make their data comparable at a regional level. For instance,

occupational categories (for both public and private employees) are standard and

comparable in the surveys from the Mercosur countries (INDEC 2011). With household

surveys you can know if a person is an employee, if he or she works as a public servant

and if this is his/her main occupation and how many hours he/she devotes to it. Length

http://sedlac.econo.unlp.edu.ar/eng/


13

of studies is included with other information. It allows us to know the bureaucratic

human capital. This basic information, for instance is going help me assess bureaucrats

preparation, motivation (Dalbo et al 2013) and their expectations, all of them crucial for

policy success.

An Example of the Use of Official National Data to Map Bureaucracies at the

Subnational Level

Using data reduction techniques I show how household surveys can help us to

improve our knowledge of the features of bureaucratic human capital. Permanent

household surveys gather individual quarterly data from Argentine survey respondents

in thirty two urban areas, I use data from a nine year period, 2003-2011. The sample is

composed of bureaucrats, those who replied that they work in the administration of the

state (excluding for instance teachers and doctors)17

.

Argentina is a federal country with twenty three provinces and an autonomous

federal district. In the city is based most of the federal administration. Since I focus on

provinces bureaucracies, I excluded this district in order to avoid the noise represented

by the disproportionate number of federal bureaucrats living in the city.18

Argentine

provinces can be gathered in three different groups: the poorer of the northeast, the less

populated though developed region of the Patagonia and the richer metropolitan

provinces from the center (Cicowiez 2004). Peripheral provinces have less competitive

political environments, higher governors‟ reelection rates, and fewer resources (Ardanaz

et al 2014). I used an original sample from a national permanent household survey

composed of 53358 public employees, collapsed at their median per province and year.

The variables used for this analysis are: Career, Education, Monthly Income (ratio),

Hours worked, Hourly Wage, Hourly Wage (ratio), Exclusivity, Permanence

17

For the variable “public servants” I use the proportion of public employees over total –private/public-

in the province. It does include all kinds of public servants. Except for this variable, the rest of the

analysis was made using a sample of bureaucrats, those who declared their main job to be working state

administration. 18

The replies do not differentiate the level of government the bureaucrats work at. However, since

household surveys are taken in urban areas where provincial administrations are located, respondents are

much more likely to be public employees at the provincial level rather than other levels.


14

(subjective), Beginner, Outgoing , Qualification and Technology.19

I also used two

more variables at the provincial level, one is the proportion of public servants over total

employees (private and public), and the other is bureaucrats, public employees in the

administration of the state over total employees (private and public). I also built sub-

samples to isolate the effect of certain variables in bureaucrats with positions of

authority (hierarchic or directions positions) from the rest of the bureaucrats.

Multidimensional scaling is a means of visualization of the data. It distributes

objects according to their relative distance in a space organized along two or more

dimensions.20

Imagine that you know distances between a number of cities but not their

location. , MDS would distribute each city and allow you to see the map. The objects

are located in a space in a way that the distances between them are preserved. When

instead of distances we have a number of variables, MDS identifies patterns of relations

among objects according to the values of the variables. For instance MDS has been used

to summarize information and cluster objects as a way of reducing data dimensionality

of many variables, it can be also used to model people perceptions according to an ex

post evaluative criteria of their replies or to classify them in many ways (i.e. politicians

perceptions regarding bureaucrats –Waterman 1998). Last MDS can also be used to test

theories and see if the data is consistent with some expected parameters of the spatial

model (Jacoby 2012)

MDS can be also used to map conceptual distances (Jacoby 2012). I know the

values of a set of variables which describe basic features of my objects: bureaucrats.

The objects I aim to map along certain dimensions are bureaucrats per province and

year. However, I do not know how these values relate, neither how similar or different

bureaucratic bodies are as a result of their within-province variation. The advantage of

MDS is that dimensions are determined by the data itself, there are not assumptions

regarding the relation among variables in a given bureaucratic body.21

Variables have

certain influence on each dimension. The combinations of the values of the variables for

each object locate them in a point in space. Since I work with two dimensions this space

19

Since variables used to assess bureaucratic characteristics have a different scale, first it was necessary to

standardize them. Find a complete list of variables in the appendix.

20 Distances result from objects according to their similarities and dissimilarities. Hendrix (2010),

Hendrix (2010), Hanson and Sigman (2013) and Bersch, Praca and Taylor ( 2013) used reduction data

techniques, factor analysis and Bayesian approaches. Only Bersch and her coauthors focus on

bureaucracies. 21

Torgeson quoted in Borg and Groensen (2005).


15

is represented bellow by a two axis graphic. 22

The assessment of the dimensions is an

ex-post operation, and outcome of the behavior of variables within each dimension,

which allows to unveiled patterns of relations and to describe bureaucratic bodies. And

this is what I am going to do next.

Subnational bureaucratic bodies are distributed in a space round two main

dimensions: bureaucrats‟ qualifications and job stability and exclusivity. The

distribution of the provinces in the space given by the dimensions of stability and

qualification follow a predictable pattern, with a couple of apparent surprises.23

Qualifications and stability seem to be both part of an efficient bureaucratic sector. We

can expect better performance in policymaking in a province with highly qualified

bureaucrats with stable positions in relation with others having just one of the two

conditions. Results show that only two provinces meet the two requirements. Not only

that, for the rest, qualifications and stability go in opposite directions. My findings show

that these dimensions do not necessarily go together, interestingly, bureaucrats‟

expectations over their permanence may not be an obvious outcome of their turnover

rate in a given province. Qualification accounts for more professional public sectors

than stability because the latter, stability, sometimes appears related to the lack of

opportunities in the provincial private labor market or, even worse, with clientelistc

biased and non-meritocratic criteria for appointments. More generally results show that

stability does not shape bureaucrats expectations (and this has a number of

consequences).

22

In this case dimensions along which provincial bureaucratic bodies are located are an outcome of the

distribution made by the MDS solution, in this case similarities and dissimilarities are measured in

Euclidean distances. Each dimension carries certain variables, which can be positively or negatively relate

to it. In the graphic bellow, dimension 2 is represented in the vertical axis, thus a variable whose value is

high and positive makes a province go left and upward. Dimension 1 is the horizontal axis, a positive and

high value of a variable within it makes a province move down and right in this spatial configuration.

23 The data was collapsed by province and year, for illustrative purposes. In the graphic I just show the

median per province of all the previous observations. I would like to be cautious in the interpretation of

these first results, particularly with the unexpected results of some provinces like Tucuman, Neuquén or

Chubut. I have proved different alternatives to visualized data, using the median instead of mean,

removing the province of San Luis in order to see the results without this outlier, any alternative proven

had very similar results, both in the position of the provinces in each dimension and in the loading of each

variable within the dimensions. However more fine tuning and rescaling may be necessary in order to see

more clearly and to evaluate the distances of the provinces grouped in the center of the graphic.


16

The first dimension is related to qualifications. Provinces located closer to this

dimension have more skilled, better paid and professional bureaucrats. They are

qualified and use technology (i.e. computers). These bureaucrats may be temporal or

tenured but they expect to continue working in the provincial public sector. It is

important to notice that I used subjective and objective measures for job permanence.

The first was permanence itself. This variable gathers the reply to the question “¿does

Table 1: Bureaucratic Human Capital. Loading of Variables, MDS, First Two Dimensions.

d1

Qualification

D2

Stability

Public servants -0.123 0.347

Bureaucrats -0.144 0.311

Exclusivity -0.0249 0.289

Exclusivity bureaucrats employees -0.0144 0.270

Beginners 0.108 -0.250

Beginners/ bureaucrats employees 0.0999 -0.237

Outgoing 0.108 -0.232

Income/ratio 0.282 0.231

Income/ bureaucrats employees 0.285 0.225

Outgoing/ bureaucrats employees 0.0971 -0.218

Income/ bureaucrats hierarchical 0.172 0.186

Income/ratio 0.263 0.181

Hourly Wage / bureaucrats employees 0.259 0.167

Qualification/ bureaucrats employees 0.282 -0.166

Exclusivity/ bureaucrats hierarchical -0.0369 0.165

Hourly Wage/ bureaucrats hierarchical 0.155 0.151

Education 0.241 -0.142

Education/ bureaucrats employees 0.230 -0.142

Qualification/ bureaucrats hierarchical 0.155 0.124

Hours worked 0.239 0.123

Hours worked/_ bureaucrats employees 0.230 0.112

Qualification 0.300 -0.110

Outgoing/ bureaucrats hierarchical 0.0544 -0.106

Beginners/ bureaucrats hierarchical 0.0400 -0.101

Hours worked/ bureaucrats hierarchical 0.0905 0.0816

Education/ bureaucrats hierarchical 0.114 -0.0616

Permanence (subjective)/ bureaucrats

employees 0.214 0.0292

Permanence (subjective) 0.213 0.0264

Career/ bureaucrats hierarchical 0.00596 -0.0165

Permanence (subjective)_ bureaucrats

hierarchical 0.0370 -0.0141

Career/ bureaucrats employees 0.0420 -0.0124

Technology 0.176 -0.0117

Career 0.0489 -0.0107

Note: Results describe the MDS loadings for the first two dimensions.


17

your job have a time limit?” In this first variable, replies on job permanence are based

on people‟s perceptions regarding their chances to continue working in their position.

For this reason this is a subjective measure of permanence because the perception may

not necessarily mean that they have a tenured position. A person who gives services on

regular bases to the state, even having consecutive temporary contracts, as it is the rule

in some districts in Argentina, can feel that his/her work has no time limit (no matter if

formally it is a temporary job). Due to procedurals impediments and/or budgetary

constraints to hire tenured employees, many agencies at the federal or provincial level

use consultancy or freelancing schemes in a way that actually cover the status of

employees. On the other hand I use an objective measure form permanence composed

of two variables with information about turnover. The beginners are the ones who were

in the first quarter of their public post at the moment of being surveyed. Outgoing

variable shows the employees that were just leaving their jobs at the moment of

replying. Higher and positive values of these two variables mean more weight of

turnover in a given dimension (this is the case of dimension 1).

The dimension 1 of qualification gathers non-exclusive employees, who also

rotate more, but with expectations of longer permanence. Since they have better

qualifications and they live in richer provinces (see below), bureaucrats in the

qualification dimension have more choices and usually their public employment is not

their only job. Exclusivity negatively relates to this dimension. Employees of provinces

closer to dimension 1, can distribute their labor day in different positions and charge

hourly for their services, average salary per hour is as important for them as their

monthly income. They can either have more than one job in the public or private sector,

and can also leave and enter the provincial administration, it may explain why both

turnover and subjective expectancy over job permanence are both positively related to

dimension 1.24

The apparent contradictory values of turnover, time limit and exclusivity account

for the particular features of Latin-American bureaucratic sector in which informal

24

Another interesting finding is the lack of influence of careers, time spent in the position, for both

dimensions. Since dimensions explain 65% of spatial distribution of bureaucracies (see table 1) it can be

inferred that career is captured by a third dimension. Even in this case the fact that it is not important for

the two first dimensions is a fact worth noticing. For different reasons I only worked with two

dimensions, there was no clear elbow where dimensions drop, two dimensions are easier to visualize and

analyze and since the first two explain the 65 % of the spatial distribution I decided to work with just two

of them for the sake of parsimony.


18

practices and no clear meritocratic procedures for recruitment are the rule (Zuvanic, et

al. 2010). Evans and Rauch suggest that expectancy over permanence is important

because stability makes bureaucrats less prone to corruption and allows them to, plan on

long term basis. The global result of this long term planning and the security regarding

stability is that they are more prone to deliver better services and to develop more

public infrastructure. A district with such bureaucratic body is more attractive for

investment and more likely to grow. But when Evans and Rauch talk of stability they

mean that bureaucrats improve their performance as they work with longer time

horizons because they expect to last in their position. Overall if we replace stability with

expectations of stability (the subjective measure of permanence) the results would be

quite similar, since expectations shape the structure of incentives for public players.

Job stability and exclusivity are main factors for a bureaucracy up and left,

closer to axis 2. Both beginners and outgoing, the ones who are either leaving or

entering or leaving the public sector, load negatively in this dimension. A province

upper and more close to the left in dimension 2 is one that has less turnover of

employees, but more turnover of those in higher positions (it makes sense because they

are more qualified and they follow the pattern of turnover described for dimension 1).

Then provinces closer to this dimension have more objective stability of employees

(90% of the sample) and higher turnover in the personnel with hierarchy.

Of course in extreme turnover can hamper the quality of policies, however

results show that in dimension 2 more stable bureaucrats do not always guarantee better

ones. First it is because, dimension 2 not only accounts for lower levels of turnover but

also for lower levels of education, qualification and hierarchy. In this context a stable

bureaucracy does not necessarily mean that has more expertise. Stability does not

guarantee quality by itself, if it does not come with preparation, qualification and the

use of basic technology (which is not the case of dimension2).

Moreover, in the expectations realm, these stable and exclusive bureaucratic

bodies that move up and left through axis 2, curiously, are also the more uncertain

regarding their job permanence. Why do not bureaucrats of dimension 2 expect to last in

their jobs? Because they are appointed on a clientelistic basis and they know they can be


19

removed as easily as they got there, even really they tend to remain employed. .25

First it

should be noticed that another variable that locates a provincial public sector up and

left, dimension 2 defined by stability and exclusivity, is the proportion of public over

total employees. The higher the share of public employees in the labor market in a given

province, the closer this province to this dimension. It can be inferred, and the spatial

distribution of the provinces support this guess, that not only do people depend on

public employment because local economies are less diversified and offer no choices,

but also because they are appointed on political (non-programmatic and procedural)

basis (Calvo and Murillo 2004, Gibson et al 2004). In these cases, expectation regarding

job stability may be weaker since he/she “owes” that position to a politician or broker.

Other reasons for this perception may be associated with the condition of being a

beneficiary of a social program. In poorer provinces some positions in the states are

given to workfare beneficiaries who have to work a number of hours as a condition to

keep getting the benefit. This is a precarious situation that can nevertheless last in time..

What are the consequences of this lack of expectations? This sense of instability

promotes corrupt practices since bureaucrats take all they can while it lasts, also they

have shorter time horizons affecting long term planning both for public policies and

particularly for the development of public infrastructure. As suggested before a more

corrupt bureaucracy and the lack of infrastructure makes a district less attractive for

investment thus hampering districts‟ growth (Evans and Rauch 1999).

25

Calvo and Murillo (2013) show how expectations to get a position in the state is relation to , ideology

and closeness to political networks.


20

Cordoba, Tierra del Fuego, and Santa Cruz are the provinces whose bureaucrats

are more qualified and educated, they use technology and expect to remain in their

positions. In Cordoba higher levels of turnover and having more than one job is the rule.

Tierra del Fuego is also up in dimension 2, so we can expect that its bureaucrats also

work exclusively in the public sector and their turnover is low, even when they have

choices in the private market. Along with Tierra del Fuego, Santa Cruz is also located

up in dimension 2 and right in relation with dimension 1, combining qualification and

stability/exclusivity. Both provinces perform better in relation to the rest.

Tucuman, Mendoza and Salta bureaucratic bodies are also closer to the right

along the dimension of qualification in detriment of stability. Buenos Aires and Santa

Fe experience lower levels of turnover in relation to Cordoba, but also they are less

educated. The fact that the three provinces are richer and have highly diversified labor

Graph 1: Distribution of MDS Bureaucratic Quality Scores by province, 2001-2011.

Note: The first dimension captures the overall level of qualification of the public sector, with greater values

describing higher levels of bureaucratic qualification. The second dimension describes the level of job

security and exclusivity, with higher values describing public sectors with higher shares of full time workers

and lower turnover rates.

Source: Own elaboration with data from the Household Survey Argentina 2003-2011. For illustrative

purposes the graphic shows mean per provinces, for the whole period

BUENOS AIRES

CATAMARCA

CHACO

CHUBUT

CORDOBA

CORRIENTES

ENTRE RIOS

FORMOSA

JUJUY

LA PAMPA

LA RIOJA

MENDOZA

MISIONES

NEUQUEN

RIO NEGRO

SALTA

SAN JUAN

SAN LUIS

SANTA CRUZ

SANTA FE

SANTIAGO DEL ESTERO

TIERRA DEL FUEGO

TUCUMAN

-20

24

Dim

ensi

on

2: s

tab

ilit

y/ex

clu

siv

enes

s

-6 -4 -2 0 2 4Dimension 1: qualification


21

markets, allow bureaucrats to have more than one job and enter and exit the public

sector even when they do not expect the job to end in the short term.

The metropolitan provinces of Cordoba, Santa Fe, Mendoza and Buenos Aires

are grouped closer to the qualification dimension. One first curiosity is that Tucuman

joins this group of metropolitan districts and is far from peripheral provinces (Gibson

and Calvo 2000). The group of peripheral provinces also includes Santiago del Estero,

Jujuy, La Rioja and San Juan, which, as expected, have less educated and qualified

bureaucrats, who rotate less, but whose time horizons are shorter due to their pattern of

appointment. Entre Rios, Catamarca and Corrientes are at the right of these provinces,

closer to the qualification dimension, but still farther than Neuquén, Chubut, Chaco, Rio

Negro and Misiones. Other outlier is San Luis, its position is closer to Neuquén

Formosa, Rio Negro and La Pampa in relation to dimension 2. They share relative

higher turnover, more public employees in relation to the private sector, and they only

have one job. However San Luis is the only province in the extreme of the upper left

quadrant, far from the dimension of qualification.

Final remarks

I claim that national official data from household surveys, and also, why not,

from census, are not only accurate but necessary complement to existing measures of

bureaucratic human capital. Most used measures are, generally, just one particular

indicator within a wider measure (i.e. governance, risk investment, state capacity,

institutional performance, etc.) and mainly based on surveys made using neither -

randomized nor representative samples of country experts. The use of official statistics

gives a standard, public, free, regular and comparable (at national and subnational level)

information to extend the scope of our conclusions in the assessment of bureaucracies

and bureaucratic practices.

To illustrate my arguments I used data on permanent household surveys made by

the national statistics institute in Argentina to map subnational bureaucracies. The map

of bureaucratic sectors help me to do three things: describe what variables matter in

order to describe the composition of bureaucratic bodies, pattern their relations (and


22

showing that they do not always meet the expectations suggested by the literature ) and

show the suitability of official data used to do so. Bureaucratic bodies were located

along two dimensions, one related to their stability (lack of turnover) and the other to

qualification. More qualified bureaucrats are also more educated and less dependent on

their position. Being up and left along the axis of the stability/exclusivity dimension,

does not mean a better performance in policy making, because stability does not shape

expectations. Expectations are more conditioned by patronage and the sense that one

can lose the job at any moment (even when in these bureaucratic bodies turnover is

lower).

In the era of big data this paper makes it a point to come down to basics to take

advantage of existent classical standard decimononic official national statistics data. It

uses a tool to describe provincial public sectors to give an example of the application of

this official data. However, at this point I cannot advance in the understanding of all

mechanisms underlying the location of provincial bureaucratic bodies in relation to the

configurational space composed by stability and qualification. The next step is to use

the dimensions found in order to compare the Argentine case with others. Also to assess

how the features of bureaucratic bodies relate to governors turnover, policy success and

provincial state capacity. It is my desire that this proposal represents a meaningful

contribution for the understanding of the main factors shaping bureaucracy and that it

can offer new tools for practitioners and scholars in the field of comparative politics and

public policy research assessing bureaucratic human capital.

References

Agranoff, R. (2001). "Managing Within the Matrix: Do Collaborative

Intergovernmental. Relations Exist?." Publius 31(1): 31:56.

Albornoz, F. and A. Cabrales (2013). "Decentralization, Political Competition and

Corruption." Journal of Development Economics 105: 103-111.

Ardanaz, M., M. Leiras, et al. (2014). "The Politics of Federalism in Argentina and its

Effects on Governance and Accountability " World Development 53: 26-45.

Arnold, W. (1989). "Bureaucratic Politics, State Capacity, and Taiwan's Automobile

Industrial Policy." Modern China 15(2): 178-214.


23

Bäck, H. and Hadenius (2008). "Democracy and State Capacity: Exploring a J-Shaped

Relationship." Governance: An International Journal of Policy, Administration, and

Institutions, 21(1): 1-24.

Banerjee, A. and E. Duflo (2008). "What is Middle Class about the Middle Classes

around the World?" The Journal of Economic Perspectives 22(2): 3-28.

Bendor, J., S. Taylor, et al. (1985). "Bureaucratic Expertise versus Legislative

Authority: A Model of Deception and Monitoring in Budgeting." The American

Political Science Review 74(4): 1041-1060.

Berkman, H., C. Scartascini, et al. (2008). Political Institutions, State Capabilities, and

Public Policy: An International Dataset. Washington DC, IDB

Bersch, K., S. Praça, et al. (2013). State Capacity and Bureaucratic Autonomy Within

National States: Mapping the Archipelago of Excellence in Brazil. Latin American

Studies Association Conference. Washington.

Borg, I. and P. J. F. Groenen (2005). Modern Multidimensional Scaling. Theory and

Applications. New York, Springer.

Calvo, E. and L. Moscovich (2013). No Protest, No Redistribution: Inequality, Social

Conflict, and the allocation of Cash Transfers in the Argentine Provinces. LASA.

Washington DC.

Calvo, E. and M. V. Murillo (2013). "When Parties Meet Voters: Assessing Political

Linkages through Partisan Networks and Distributive Expectations in Argentina and

Chile." Comparative Political Studies 46(4): 851-882.

(2004). "Who delivers? Partisan Clients in the Argentine Electoral Market."

American Journal of Political Science 48(4): 742-747.

Charron, N. and V. Lapuente (2010). "Does Democracy Produce Quality of

Government?" European Journal of Political Research 49: 443–470.

Cicowiez, M. (2004). Caracterización Economico-Social de las Provincias Argentinas.

Disparidades regionales y federalismo fiscal A. Porto. La Plata, Universidad de La

Plata.

Cingolani, L. (2013). The State of State Capacity: a review of concepts, evidence and

measures Working Paper Series on Institutions and Economic Growth. Maastricht,

UNU-MERIT.

Cingolani, L., K. Thomsson, et al. (2013). Minding Weber more than ever? The impacts

of State Capacity and Bureaucratic Autonomy on development goals. Working Paper

Series on Institutions and Economic Growth. Maastricht, UNU-MERIT.


24

Cortázar Velarde, J. C., M. Lafuente, et al., Eds. (2014). Al servicio del ciudadano una

década de reformas del servicio civil en América Latina (2004–13) Washington, dc,

IDB.

Dahlström, C. and V. Lapuente (2012). Weberian bureaucracy and corruption

prevention. DPADM Expert Group Meeting on Prevention of Corruption in Public

Administration, New York.

Dal Bo, E., F. Finan, et al. (2013 ). "Strengthening State Capabilities: The Role Of

Financial Incentives in the Call to Public Service." The Quarterly Journal of Economics

(2013) 128 (3): 128(3): 1169-1218.

Deaton, A. (1997). The Analysis of Household Surveys. Baltimore, World Bank and

The John Hopkins University Press.

En Bai, C. and S. Jin Wei (2000) Quality of Bureaucracy and Open-Economy Macro

Policies. National Bureau of Economic Research

Evans, P. (1995). Embedded Autonomy Princenton Princenton University Press.

Evans, P. and Rauch (1999). "Bureaucracy and Growth: A Cross-National Analysis of

the Effects of "Weberian" State Structures on Economic Growth." American

Sociological Review 64(5): 748-765.

Fukuyama, F. (2013). What Is Governance? Working Paper. Washington, DC, Center

for Global Development.

(2012). "The Strange Absence of the State in Political Science." Retrieved from

http://www.the-american-interest.com/2012/10/02/the-strange-absence-of-the-state-in-

political-science/ Accessed on April 2015

Gailmard, S. and J. Patty (2007). "Slackers and Zealots: Civil Service, Policy

Discretion, and Bureaucratic Expertise." American Journal of Political Science 51(4):

873–889

Geddes, B. (1994). Politicians Dilemma. Building State Capacity in Latin America

Berkeley University of California

Gibson, E. and E. Calvo (2000). "Federalism and Low-Maintenance Constituencies:

Territorial Dimensions of Economic Reforms in Argentina " Studies in Comparative

International Development 35(3): 32-55.

Gibson, E., E. Calvo, et al. (2004). Reallocative Federalism Legislative

Overrepresentation and Public Spending in the Western Hemisphere. Federalism and

Democracy in Latin America. E. Gibson. Baltimore, John Hopkins University Press.

Hanson, J. and R. Sigman (2013). Leviathan‟s Latent Dimensions: Measuring State

Capacity for Comparative Political Research. Syracuse.

http://www.the-american-interest.com/2012/10/02/the-strange-absence-of-the-state-in-political-science/

http://www.the-american-interest.com/2012/10/02/the-strange-absence-of-the-state-in-political-science/


25

Hendrix, C. (2010). "Measuring State Capacity: Theoretical and Empirical Implications

for the Study of Civil Conflict." Journal of Peace Research 47(3): 273-285

Honig, M. (2006). "Street-Level Bureaucracy Revisited: Frontline District Central-

Office Administrators as Boundary Spanners in Education Policy Implementation."

Educational Evaluation and Policy Analysis 28(4): 357–383.

INDEC (2011). Clasificación de Actividades Económicas para Encuestas

Sociodemográficas del MERCOSUR. Buenos Aires, INDEC.

Jacoby, W. (2012). Multidimensional Scaling: An Introduction Workshop In Methods

Indiana University. Indiana, Workshop In Methods Indiana University.

Kaufmann, D., A. Kraay, et al. (2007). The Worldwide Governance Indicators Project:

Answering the Critics. Washington, dc, World Bank.

Kurtz, M. (2013). Latin America State Building in Comparative Perspective. Social

Foundations of Institutional Order. Cambridge, Cambridge University Press.

Leuprecht, C. and J. Goldstone (2013) Political Demography. E-International Relations

retrieved from: http://www.e-ir.info/2013/02/23/political-demography/ acceded April

2015

Levitsky, S. and M. V. Murillo (2005). Building Castles in the Sand? The Politics of

Institutional Weakness in Argentina. Argentine Democracy. The Politics of Institutional

Weakness. S. Levitsky and M. V. Murillo. Pennsilvania, Pennsylvania State University

Press.

Mazzuca, S. (2012). Access to Power versus Exercise of Power. La calidad de la

democracia: perspectivas desde América Latina. G. Munck and S. Mantilla. Quito.

Migdal, J. (2011). Estados débiles, Estados fuertes. FCE, México.

Milanovic, B. (2002). "True World Income Distribution, 1988 And 1993: First

Calculation Based On Household Surveys Alone." The Economic Journal 112: 51-92.

O'Donnell, G. (2007). Las poliarquías y la (in)efectividad de la ley en América Latina

Disonancias. Criticas Democráticas a la Democracia. . G. O'Donnell. Buenos Aires

Prometeo.

Olson, M., N. Sarna, et al. (2000). "Governance and Growth: A Simple Hypothesis

Explaining Cross-Country Differences in Productivity Growth Public Choice 102(3/4):

341-364.

Panizza, U., R. d. Tella, et al. (2001). "Public Sector Wages and Bureaucratic Quality:

Evidence from Latin America" Economía 2(1): 97-151.

Rauch, J. and P. Evans (2000). "Bureaucratic Structure and Bureucratic Performance in

Less Developed Countries." Journal of Public Economics 75: 49-71.


26

Rodrigues-Silveira, R. (2014). A demografia das elites políticas locais: usando os dados

do censo para avaliar a estrutura e as mudanças nos grupos políticos dominantes no

nível local. In Eleições, partidos e representação política nos municípios brasileiros. M.

Mendes Rocha and M. T. Miceli Kerbauy. Juiz de Fora, UFJF.

Savoia, A. and K. Sen (2012). Measurement and evolution of state capacity: exploring a

lesser known aspect of governance. Manchester, ESID.

SEDLAC (2014) Database. Available in http://sedlac.econo.unlp.edu.ar/eng/. Accessed

April 2015

Sikkink, K. and L. Wolfson (1993). "Las capacidades y la autonomia del Estado en

Brazil y la Argentina. un enfoque neoinstitucionalista." Desarrollo Económico 32(128):

543-574.

Smulovitz, C. (2012). Public Defense and Access to Justice in a Federal Context: Who

Gets What, and How in the Argentinean Provinces?. APSA.

Soifer, H. (2013). "State Power and the Economic Origins of Democracy." Studies in

Comparative International Development 48(1): 1-22.

Soifer, H. and M. vom Hau (2008). "Unpacking the Strength of the State: The Utility of

State Infrastructural Power." Studies in Comparative International Development 43:

219–230.

Souza, C. (2015). Capacidade burocrática no brasil e na argentina: quando a política faz

a diferença1 celina souza2. Texto para discussão Brasília : Rio de Janeiro, Instituto de

Pesquisa Econômica.

UNDP (2007). Governance Indicators: A Users‟ Guide. New York, UNDP.

Van de Walle, S. (2006). "The State of the World‟s Bureaucracies." Journal of

Comparative Policy Analysis, 8(4): 437-448.

Waterman, R., A. Rouse, et al. (1998). "The Venues of Influence: A New Theory of

Political Control of the Bureaucracy." Journal of Public Administration Research and

Theory 8(1): 13-38.

Weber, M. (1995). Economía y Sociedad. Mexico d. f, Fondo de cultura económica.

Wibbels, E. (2005). Federalism and the Market. Intergovernmental Conflict and

Economic Reform in the Developing World. Cambridge, Cambridge University Press.

http://sedlac.econo.unlp.edu.ar/eng/


27

Appendix

Multidimensional Scaling for the assessment of Bureaucratic Bodies of Argentina Provinces. Dimension

1 Qualification, Dimension 2 Stability/Exclusivity

Source: Own elaboration. MDS solution used in Euclidean distances, with information of 747

observations for 14 variables collapsed per province and year, from an original sample of 53338

bureaucrats interviewed, included in the Permanent Household Survey, Argentina from 2003 to 2011. .

10 652.3445 3.50 78.63 1.29 95.36

9 663.84043 3.56 75.13 1.33 94.07

8 726.79445 3.90 71.57 1.60 92.73

7 833.59235 4.47 67.67 2.10 91.13

6 1032.1646 5.53 63.20 3.23 89.03

5 1100.5824 5.90 57.67 3.67 85.81

4 1270.9555 6.82 51.77 4.89 82.14

3 1995.3256 10.70 44.95 12.05 77.25

2 2438.8404 13.08 34.25 18.01 65.20

1 3948.3194 21.17 21.17 47.19 47.19

Dimension Eigenvalue Percent Cumul. Percent Cumul.

abs(eigenvalue) (eigenvalue)^2

Retained dimensions = 2 Mardia fit measure 2 = 0.6520

Eigenvalues > 0 = 33 Mardia fit measure 1 = 0.3425

Number of obs = 630

dissimilarity: L2, computed on 33 variables

Classical metric multidimensional scaling

. mds std1*, id(id)


28

Variable Definition

Career An average of an ordinal scale for26

time spent in the same public position

1 = less than a month

2= 1 to 3 months

3= 3 to 6 months

4= 6 to 12 months

5= 1 to 5 years 6= more than 5 years

Education An average of an ordinal scale for educational status

0= no education

1= primary (incomplete)

2= primary school

3= high school (incomplete)

4= high school

5= college (incomplete)

6=bachelor degree

Income (ratio) Average per province and year of employee monthly income, over the mean income of

all public employees for this trimester

Hours worked Average per province and year of number of hours worked in the last week in the

person‟s main occupation (bureaucrats post)

Hourly Wage Average per province and year of income over hours worked

Hourly Wage (ratio) Person‟s hourly wage over the mean of all public employees hourly wage in this

trimester

Exclusivity Average per province and year of a dichotomous variable taking the value of 1 if the

position in the public sector is the person‟s sole employment or 0 otherwise

Permanence

(Subjective)

Average per province and year of a dichotomous variable taking the value of 1 if the

person replies that this job has no time limit or 0 otherwise.

Beginners Average per province and year of a dichotomous variable taking the value of 1 if

during the former trimester the person did not work in the public sector or 0 otherwise.

Outgoing Average per province and year of a dichotomous variable taking the value of 1 if in the

next semester the person does not work in the public sector or 0 otherwise.

Qualification Average per province and year of an ordinal scale for job qualifications

1=not qualified

2=operative

3=technician

4=professional

Technology Average per province and year of a dichotomous variable taking the value of 1 if the

person woks with some kind of computerized system or 0 otherwise.

Public servants Average per province and year of the ratio of people working in the estate in all

positions ( also doctors, teachers, etc.) over the total province employees (private or

public).

Bureaucrats Average per province and year of the ratio of people working in the administration of

the state over the total of province employees (private or public).

26

Averages of ordinal variables make sense as long as higher values represent, somehow, more of this attribute, for instance higher category 4, means more job qualifications than category 1, the same happens with education, a number 6 refers to a higher educational status.


29

Descriptive statistics

Mean Median Min Max StDv Obs

Hourly Wage

(ratio) 0,97 0,92 0,39 3,83 0,31 747

Income (ratio) 0,97 0,92 0,40 2,19 0,31 747

Exclusivity 0,88 0,89 0,67 1,00 0,06 747

Permanence

(Subjective) 0,82 0,84 0,16 1,00 0,13 747

Hours 34,28 34,28 25,28 44,32 2,94 747

Education 4,05 4,05 2,86 5,12 0,37 747

Career 5,44 5,47 3,07 5,94 0,23 747

Beginners 0,02 0,00 0,00 0,22 0,03 747

Outgoing 0,02 0,00 0,00 0,31 0,03 747

Technology 0,44 0,48 0,00 0,84 0,18 747

Qualification 2,30 2,30 1,42 2,98 0,19 747

Data from Household Surveys. Public employees in the administration of the state, per province and year.

Argentina 2003-2011

Get to Know your Bureaucracies. Mapping Subnational Bureaucratic Bodies with Evidence from Official Household Surveys. Argentina2003-2011

Documents