Planning R&D in a Post Centrally-Planned Economy: Assessing the Macroeconomic Effects in Poland CoPS Working Paper No. G-268, December 2016 The Centre of Policy Studies (CoPS), incorporating the IMPACT project, is a research centre at Victoria University devoted to quantitative analysis of issues relevant to economic policy. Address: Centre of Policy Studies, Victoria University, PO Box 14428, Melbourne, Victoria, 8001 home page: www.vu.edu.au/CoPS/ email: [email protected]Telephone +61 3 9919 1877 Katarzyna Zawalińska IRWiR, Polish Academy of Sciences Nhi Tran Centre of Policy Studies, Victoria University Adam Płoszaj EUROREG, University of Warsaw ISSN 1 031 9034 ISBN 978-1-921654-76-3
31
Embed
Planning R&D in a Post Centrally-Planned Economy ...
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
Planning R&D in a Post Centrally-Planned Economy: Assessing the Macroeconomic Effects in Poland
CoPS Working Paper No. G-268, December 2016
The Centre of Policy Studies (CoPS), incorporating the IMPACT project, is a research centre at Victoria University devoted to quantitative analysis of issues relevant to economic policy. Address: Centre of Policy Studies, Victoria University, PO Box 14428, Melbourne, Victoria, 8001 home page: www.vu.edu.au/CoPS/ email: [email protected] Telephone +61 3 9919 1877
Katarzyna Zawalińska
IRWiR, Polish Academy of Sciences
Nhi Tran
Centre of Policy Studies, Victoria University
Adam Płoszaj
EUROREG, University of Warsaw
ISSN 1 031 9034 ISBN 978-1-921654-76-3
1
Planning R&D in a post centrally-planned economy: assessing the
macroeconomic effects in Poland
Katarzyna Zawalińskaa *
, Nhi Tranb, Adam Płoszaj
c
a IRWiR, Polish Academy of Sciences; ul. Nowy Świat 72, 00-330 Warsaw, Poland; email:
[email protected] (corresponding author) b Centre of Policy Studies Victoria University; 10/300 Flinders Street, Melbourne, Australia;
email: [email protected] c EUROREG, University of Warsaw; ul. Krakowskie Przedmieście 30, 00-927, Warsaw,
6. Construction (section F), 7. Wholesale and retail trade (section G), 8. Transporting and storage (section H), 9.
Accommodation and food (section I), 10. Information and communication (section J), 11. Financial and
insurance activities (Section K), 12. Real estate activities (section L), 13. Professional, scientific and technical
activities (Section M without R&D), 14. Scientific Research and Development services (part of Section M), 15.
Administration (section N), 16. Public administration and defense; compulsory social security (section O), 17.
Education (section P), 18. Human health and social work (section Q), 19. Rest of services (sections R, S, T, U). 7 The household sector in the model includes non-profit institutions serving households (NPISH). This is these
institutions that use R&D services.
13
of Economic Analysis (Sliker, 2007; Bernat 2007), stock of knowledge is calculated using the
perpetual inventory method. That is, for each agent and each year of the simulation period,
𝐾𝑅&𝐷,𝑡 = 𝐾𝑅&𝐷,𝑡−1(1 − 𝐷𝑅&𝐷) + 𝐼𝑅&𝐷,𝑡 (1)
where 𝐾𝑅&𝐷,𝑡−1 and 𝐾𝑅&𝐷,𝑡 are stock of R&D at the begininng and end of year t respectively,
𝐷𝑅&𝐷 is R&D depreciation rate, and 𝐼𝑅&𝐷,𝑡 is R&D expenditure during year t.
Following the method used by Bernstein and Mamuneas (2006), we calculated the initial
R&D stock for the initial year (𝐾𝑅&𝐷,0) as the ratio of R&D expenditure in the year (𝐼𝑅&𝐷,𝑜) to
the sum of R&D depreciation rate and the average growth rate of R&D over some earlier
period (�̇�𝑅&𝐷).
𝐾𝑅&𝐷,𝑜 =𝐼𝑅&𝐷,𝑜
𝐷𝑅&𝐷 + �̇�𝑅&𝐷
(2)
The value for the initial R&D expenditure (𝐼𝑅&𝐷,𝑜) is obtained directly from the input-
output tables. We have adopted the value of 10% for the depreciation rate (𝐷𝑅&𝐷 ) which is
consistent with the value adopted by many studies on R&D (Garau and Lecca, 2015: Keller,
2004: Krammer, 2010). This value is higher than the average depreciation rate for physical
capital in POLTERM (6.2%). This is however consistent with the finding from studies
directly estimating R&D depreciation rates (see, for example, Bernstein and Mamuneas 2006)
that R&D stock depreciates faster than physical capital stock. The growth rate of R&D stock
(�̇�𝑅&𝐷) is the average growth rate between 2005 and 2010, which is 10.3% according to
OECD (2012).
The R&D stock in each region is calculated as the sum of R&D stocks by industry,
households and government in the region8.
We explicitly model the link between returns on R&D stock and total factor productivity
for each region. If we call KR&D and RoRR&D the stock and rate of return on R&D stock
respectively, then the change in GDP at factor costs (Y) in each region resulting from the
returns on R&D would be:
∆𝑌 = 𝑌 × �̇� = 𝐾𝑅&𝐷 × 𝑅𝑅&𝐷 (3)
where �̇� the growth rate of GDP at factor costs. Representing GDP as a function of
technology, labour and capital stock, i.e. Y=AF(L,K), the growth rate in GDP equals the
8 We do not count the exports of R&D services as investment in R&D stock. This is because R&D exports are
likely to add to the stock of knowledge by foreign countries, and not directly to the stock of knowledge in the
domestic economy.
14
growth rate of technology plus the share-weighted average of growth rates of labor and capital
stock. That is:
�̇� = �̇� + S𝐿�̇� + S𝐾�̇� (4)
where the dot above the variables represent their growth rates, and SL and SK are shares
of labor and capital stock respectively in GDP at factor costs. When the change in GDP is due
to the change in technology alone, we have �̇�=�̇�, and the growth rate in TFP can be calculated
from (3) as:
�̇� =𝐾𝑅&𝐷 × 𝑅𝑅&𝐷
𝑌
(5)
Estimates for the rate of return on R&D stock (𝑅𝑅&𝐷) in the literature vary from 8.0 to
170.0% (Brandsma and Kancs, 2015: Mairesse and Sassenou M., 1991: Pakes and
Schankerman, 1978). Kolasa (2008) estimated the private rate of return to R&D in Poland of
above 30%, but as Meriküll et al. (2013) pointed out, in transition economies the country-
level analysis may overestimate TFP growth and the role of R&D in it, as aggregate TFP
growth due to changes in industry structure may be erroneously ascribed to TFP growth due
to R&D. Therefore, we have adopted a more conservative value of 25%.
Studies linking R&D and productivity sometimes use the elasticity of TFP with regard to
R&D stock instead of the returns on R&D stock used in this paper. For comparison, we have
calculated the elasticity resulting from our method. The elasticity is not constant, because in
equation (5) above the numerator and denominator on the right hand side of the equation
would change in different ways. R&D capital stock movements depend on the level of R&D
investment, whereas GDP movements depend both on changes in the TFP itself and on
changes in capital stock and employment, which in turn depend on many factors apart from
changes in R&D stock. However, on average, our simulation results show that the elasticity
varies between 0.02 to 0.06, which is within the estimated range in the studies by
Świeczewska (2015) and Madden et al. (2001).
3.3. Simulation scenarios for R&D policies
The Europe 2020 strategy has set a target for R&D expenditures to GDP ratio for Poland
of 1.7% by 2020. To reach this level from the 2014 level of 1.07%, the ratio will need to
increase on average by 7.99% per annum over the period 2015 to 2020. Whilst in theory there
are a number of ways to achieve the overall target, in this paper we explore just two possible
policy instruments as these are the only two currently available in Poland. Policy 1 is to
15
increase government demand for R&D services (e.g. research programs and grants available
for public and private institutions). Policy 2 is to provide R&D tax relief for enterprises.
To compare the efficacy of these policies, two simulations with POLTERM for the period
2011-2025 were run. Going beyond the 2020 deadline enables investigation not only of the
short-run but also the long-run impacts of the R&D targeting policies. The simulations are run
in two stages. First, a baseline (business-as-usual) forecast was produced for the 2011-2025
period. This forecast includes available forecasts for the economy, but excludes the effects of
the R&D targeting policies. The second stage involved the development of two policy
forecasts for the same period, one for each of the policies listed above. Each policy forecast
includes the shocks underpinning the aforementioned baseline forecast, but with the addition
of a shock describing the R&D targeting policy. We report results for each of the policies as
time paths of percentage deviations in the values of variables in each policy forecast away
from their values in the baseline forecast.
In each of the policy simulations we exogenize the R&D/GDP ratio and shock it with the
values calculated for each year during the period 2015-2020 at a constant growth rate of
7.94% discussed above so as to reach the target of 1.7% by 2020, and then maintain it at that
level by 2025 (see Fig. 2). We endogenize the variable representing the policy under
consideration so as to force the model to find the changes in the policy that are required to
reach the R&D/GDP target. Specifically, for Policy 1 we endogenize government demand for
R&D services. For Policy 2 we endogenize the consumption tax on R&D uses by domestic
businesses. Changes in those policy instruments are reported in Fig 2.
Simulation results show that by 2020 the government would have to increase its level of
expenditures on domestic R&D by 300% compared with baseline. The share of R&D in total
government consumption expenditures increases from 1.1% in 2014 to 4% in 2020.
Afterwards, the expenditures change at a relatively low rate (see Fig.2).
In the scenario where the government provide tax relief to businesses, simulation results
show that the rate of tax relief on the production of R&D services will have to be from 98.9%
in 2015 up to 100% by 2020 and then stay at that level in order to reach and maintain the
target.
16
Fig. 2. Changes in R&D/GDP ratio and in policy instruments (Level for R&D/GDP target, %
deviations from baseline for government expenditures; and deviations in the rate of tax relief for R&D
expenditure)
Source: Authors’ own calculations based on POLTERM model.
It should be noted that in this paper we target R&D at the national level, that is, we
assume that the policy instruments would change at the same rate for all regions. Other
assumptions in the policy simulations are as follows. As in the baseline, we assume that
government consumption moves with private consumption. The composition of government
consumption, i.e. the share of different commodities in the government consumption basket,
remains unchanged, except when the consumption of R&D is shocked. We also assume that
aggregate final consumption moves with GDP. This means an increase in government
expenditures occurs at the expense of private consumption. This assumption is used to prevent
the economy from building up its foreign debts. We also assume that aggregate employment
remains unchanged at the baseline level. This is because employment is already growing
strongly at over 1.5% p.a. in the baseline, higher than the population growth rate, and hence
there is little scope for an increase in employment. We also expect that net migration from
other EU countries due to R&D expansion will be negligible, because all EU countries will be
implementing policies to increase their R&D expenditure as well.
0
50
100
150
200
250
300
350
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
20
14
20
15
20
16
20
17
20
18
20
19
20
20
20
21
20
22
20
23
20
24
20
25
R&
D t
arge
t,
Rat
e o
f ta
x re
lief
on
R&
D e
xpe
nd
itu
re
Go
vern
me
nt
exp
en
dit
ure
on
do
me
stic
R&
D
R&D/GDP target
Rate of tax relief on R&D expenditure
Government expenditure on R&D
17
4. Results
This section discusses the results from our policy simulations. First we discuss the main
impacts of each of the alternative policies, then we compare them in terms of their impacts on
GDP and government budget.
4.1. R&D target achieved by government R&D expenditures (Policy 1)
The first impact of an increase in government expenditures on R&D is to increase demand for
R&D services. This raises the price and output of R&D. As a back-of-the-envelope (BOTE)
calculation, by 2020 government consumption of domestic R&D increases by 268% relative
to the baseline. In 2014 the government consumes 20% of R&D output. The 268% increase in
government consumption alone would cause R&D output to rise by about 53.6%. This almost
entirely explains the 54.1% increase in R&D output.9 Employment in the sector increases,
while higher rates of return triggers investment, leading eventually to an increase in capital
stock. The dominant impact of the policy is the increase in R&D stock, which leads to an
increase in productivity for the whole economy. This, in turn, causes GDP to rise. By 2020
the deviation of real GDP from baseline is 0.38%. Fig. 3 reports simulation results for GDP
income components. As can be seen from the figure, the increase in real GDP is mainly due to
the increase in productivity induced by R&D stock accumulation.
During the 2015 to 2020 period the increase in real GDP is slightly lower than that of
productivity. This is because there is a slight negative deviation in capital stock due to two
main reasons. First, with a constant average propensity to consume, defined as a ratio of
nominal final consumption (C+G) to nominal GDP, the rise in government consumption due
to its increased expenditures on R&D causes a fall in private consumption. As private
consumption is more capital intensive than government consumption, a fall in private
consumption causes capital stock to fall relative to baseline. Second, we have assumed that
aggregate employment stays at the baseline level. The expansion of the R&D sector attracts
labor away from other industries, causing marginal products of capital in these industries to
fall. This lowers their rates of return, and hence their investment, and subsequent capital stock
fall. However, as can be seen from
Fig. 3, these effects on capital stock are very small.
9 As discussed earlier, simulation results for variables are deviations from their values in the baseline forecast.
For brevity, we sometimes use a “rise” or an “increase” to indicate a positive deviation from baseline, and a
“fall” or a “decline” to indicate a negative deviation from baseline.
18
In the long run, after the R&D target has been reached, the above effects become much
smaller, and the improvement in productivity raises the marginal product of capital stock and
labour in all sectors, causing aggregate capital stock to increase. Together, the increase in
capital stock and productivity, as well as in the wagebill-weighted employment, causes real
GDP to continue to grow, reaching 3.1% higher than the baseline by 2025.
Fig. 3. Policy 1 – GDP income components, government expenditures on R&D and
R&D output (% deviations from baseline)
Source: Authors’ own calculations based on POLTERM model.
4.2. R&D target achieved by the provision of tax relief on R&D expenditure
(Policy 2)
In this section we explore the impacts of a tax relief on the expenditure on R&D by
domestic businesses. As discussed earlier, simulation results show that the tax relief rates will
have to increase to approximately 100% by 2020 to reach the R&D intensity target. Fig. 4
reports the percentage deviations of R&D output and GDP and its income components from
their baseline values.
-50.00
0.00
50.00
100.00
150.00
200.00
250.00
300.00
350.00
-0.5
0
0.5
1
1.5
2
2.5
3
3.5
20
14
20
15
20
16
20
17
20
18
20
19
20
20
20
21
20
22
20
23
20
24
20
25
Go
vern
me
tn e
xpe
nd
itu
re o
n R
&D
, R&
D
ou
tpu
t
Re
al G
DP
, C
apit
al s
tock
, Te
chn
olo
gy
Real GDP Employment
Capital stock Technology
output of non-R&D sectors R&D output
19
Fig. 4. Policy 2 – R&D output and contributions of GDP income components to GDP
(% deviations from baseline)
Source: Authors’ calculations based on POLTERM model.
The first impact of the tax relief policy is to reduce the costs of R&D investment for
industry. This increases industry demand for R&D services. This explains the large positive
deviations in R&D output, which reaches 55.1% above the baseline by 2020. However, in the
first year real GDP experiences a small negative deviation of -0.06% from baseline. This is
mainly due to the decline in indirect tax revenues caused by the large tax relief. In 2015,
indirect tax revenues falls 8.8% compared with baseline. Indirect taxes contribute about 12%
of Poland’s GDP. The fall of 8.8% in tax revenues should have led to a 1.1% decline in real
GDP. However, real GDP declines by only 0.06%. This is due to the positive contribution
from productivity, capital stock and labor. The increase in R&D expenditure increases the
stock of knowledge and generates improvements in productivity. Productivity contributes
directly to GDP, but also improves the marginal products of capital stock and labor, so capital
stock increases. As employment levels are assumed to stay at the baseline level, its
contribution to GDP growth occurs via the increase in the real wage, but the contribution is
small. From 2016 onwards the positive contributions from technology, capital stock and
employment become larger than the negative impact of indirect tax revenues. Real GDP
deviations turn positive, reaching 0.85% higher than baseline by 2020 and 3.72% by 2025
respectively.
-40
-20
0
20
40
60
80
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
20
14
20
15
20
16
20
17
20
18
20
19
20
20
20
21
20
22
20
23
20
24
20
25 R&
D o
utp
ut
Re
al G
DP
, C
apit
al s
tock
, lab
ou
r, T
ech
no
logy
, ta
x
Real GDP Contribution from labour
Contribution from capital stock Contribution from technology
Contribution of commodity tax R&D output
Contribution from production tax
20
At the industry level, the results are quite similar to those found with Policy 1. The R&D
sector expands the most (see Fig. 4), with output deviating 55% by 2020 and 70% by 2025
from the baseline. However, the impacts on other sectors are more positive than those in
Policy 1. In aggregate, only in 2016 do they experience a small contraction of -0.06% in their
outputs. From 2017 onwards the sectors grow, reaching deviations of 1.06% and 3.95% above
baseline in 2020 and 2025, respectively. The reason for the more positive outcomes for other
industries under Policy 2 is that, although they also experience resource-allocation impacts
from the fast growth of the R&D sector as in Policy 1, they do not experience the negative
impacts of an increase in R&D price as in Policy 1. On the contrary, they experience a fall in
the price of R&D due to the tax relief policy.
4.3. Achieving R&D target by different policy instruments
Fig. 5 compares the impacts on real GDP and the impacts on government budget under
the two policies. It is clear that the policy of tax relief leads to a higher positive GDP
deviation from the baseline (0.85% in 2020 and 3.71% in 2025) than the policy of increasing
government expenditures on R&D, where deviation from the baseline is respectively 0.38% in
2020 and 3.10% in 2025. However, the tax relief policy is also much more costly to the
budget. In the government expenditures scenario the change in the value of government
expenditures on R&D compared with that in the baseline increases almost linearly from
PLN10
2bn in 2015 to PLN 12.6bn in 2020, and then up to PLN 16.4bn in 2025 at 2014 prices.
The total present value of the increase in government expenditures, discounted at 5% pa,
is PLN 33.9bn for the period 2015-2020. At the same time, in the tax relief scenario the
budgetary outlays increases from PLN 17.9bn in 2015 to PLN 32.8bn in 2020 and to PLN
40.9bn in 2025. The present value of the subsidies, discounted at 5% pa, is PLN 129.9bn for
the period 2015-2020.
So the tax relief policy is superior to the government expenditure policy in terms of the
impacts on GDP, both in the short run and the long run, but it is inferior in terms of the direct
costs on the government budget. So in comparing the two policies, the question arises which
of them is more efficient in terms of achieving the same R&D target with a lower output/input
ratio, where output is measured as real GDP growth (above the baseline) and input as the
budgetary cost of achieving it. In other words, which of the two policies buys the additional
increase in GDP growth for less. Tab. 1 compares the ratio for the two policies over the time
period considered in this study.
10
PLN, is the abbreviation for the Polish national currency the Polish Zloty, where 4 PLN ≈ 1 EUR.
21
Fig. 5. Real GDP (% deviation from baseline) and government budget outlays (PLN
billion change from baseline) under alternative policies
Source: Authors’ calculations based on POLTERM model.
Table 1 Approximate efficiency of R&D policies
Source: Authors’ own calculations
Tab.2 shows that the efficiency ratio has been changing over time for both policies. At the
beginning of the period the tax relief policy had a slightly higher efficiency. By 2020,
however, the efficiency of the two policies was equal (0.03) and beyond 2020 (in the long
run) the government expenditure policy brings greater value for money. By 2025 the
efficiency ratio is quite markedly higher for the government expenditure policy being 0.19