The Effect of Government Size on Economic Growth and Technical Change45312

The Effect of Government Size on Economic Growth and Technical Change Cuong Tat Do 1* , Anh Ngoc Thi Ngo 1 , Dinh Van Nguyen 2 1 Institute of Economics, Ho Chi Minh National Academy of P

The Effect of Government Size on Economic Growth and Technical Change

Cuong Tat Do 1* , Anh Ngoc Thi Ngo 1 , Dinh Van Nguyen 2

1 Institute of Economics, Ho Chi Minh National Academy of Politics, HCM city, Vietnam

2 VNU International School, Vietnam National University, Hanoi, Vietnam

* Correspondence: cuongdt@hcma.vn

Abstract: This paper examines the effects of government size on economic growth and technological

change at country level using Penn World Table version 9.0 with a focus on Asian nations Government size is measured by the share of government expenditure to gross domestic product and technological change is measured by total factor productivity at national level Using endogenous growth theory as major theoretical framework and employing country mixed-effect regression models, we find that the effects government size on economic growth and technological change are complex and nonlinear Indeed, the causal relationship between government size and the both economic growth and technological change shows a multiple equilibrium function As a consequence, variation of government size yields an unpredicted variation of economic growth and technological change It implies that once policy makers decide a change in the size of government, they should aware that the effects of the change on economic growth and technological change might not as their expectations

Keywords: Government size; economic growth; technological change; Asian nations

The paper supplies new and robust empirical evidence into these issues and contributes to contemporary literature in several dimensions First, instead of emphasizing

on the direct effects, this paper shows nonlinearity and heterogeneity in the effect of government size on economic growth Precisely, it investigates a threshold such that existed government size will change its effect to economic growth The main novelty of the study is

to employ a full polynomial function to analyze the effect of government size on economic growth Second, the paper assesses whether the government size affect to economic growth through its effect on total factor productivity (TFP) The question is very important because TFP growth has been shown to be the main driver of economic performance (Easterly and Levine (2002); Parente and Prescott (2002); Caselli (2005); Gómez-Sancho, Barcenilla-Visús, López-Pueyo, Mancebón, and Sanaú (2013); Kim, Wu, and Lin (2018)), which is a standard economic growth accounting dated back to Solow’s first effort Third, the paper analyses the effect of government size on technological progress Technological progress has been shown

as a change of capital-labor ratio which might be affected by government intervention,

measured by general government expenditure, on the labor and capital market

2 A brief literature review

Government can affect economic growth by its size and quality The effect of size and quality of government is the two distinctive branches of economic literature This paper focuses on the effect of government size rather than the both effects Studies on the effect of government size to economic growth emphasize the importance of the state’s absorption of society’s resources through its expenditure and taxation These studies find that a larger government expenditure is growth-impeding (for example: Barro (1991); Fölster and Henrekson (2001); Sala-i-Martin, Doppelhofer, and Miller (2004); Ghosh and Gregoriou (2008); Bergh and Karlsson (2010); Kim et al (2018)) They also point out the negative effect

of the expansion of the government size in fostering the rent-seeking behavior at the costs

of economically productive activities On the other hand, several studies find the opposite results (for example: Barro (1990); Grossman (1990); Barro and Sala-i-Martin (1992); Hansson and Henrekson (1994); Turnovsky and Fisher (1995); Kneller, Bleaney, and Gemmell (1999)) and contribute the positive effect to the existence of market failure and negative externalities

The inconclusive findings could arise the question of the existence of a threshold where government size might switch its effect on economic growth Barro (1990), Karras (1996), Asimakopoulos and Karavias (2016), and Kim et al (2018) show that the effect of government size on economic growth does have a threshold Once the size of government passes a particular threshold, the effect of government size on economic growth will change from positive to negative vice versa They also point out that the relationship between government size and economic growth is nonlinear with thresholds rather than linear

The above effect of government size on economic growth is some time called direct effect, and the other effect is indirect effect The major force behind economic growth is technological progress The different levels of technological progress between countries explain the different growth rates worldwide Government size channels its effect on economic growth through technological progress via two ways: (i) incentivize innovation in

firms and government bodies (Caselli (1999); Galor and Moav (2000); Aghion, Howitt, and Violante (2002)) and (ii) provide stable environment for economic growth through optimal spending programs (Cozzi & Impullitti, 2010)

3 Model

It is assumed that all nations have same form of production function following the standard proposed by Aghion and Howitt (1998) In all economies, every firms and households have same technology Therefore, the basic production function of these economies is assumed:

, = , , , = , , , (1) Where , is national income of country i in the time t, , is national capital of

country i in the

time t, , is national labor of country i in the time t, and , is a measure current

level of technology for a country called total factor productivity of country i in the time t;

, are marginal elasticity of capital and labor to national income respectively

Allowing for a changing in technology level, we have the following equation:

Because changes in total factor productivity cannot observe directly, it is measure indirectly We have data on growth of national income, capital and labor From these data and growth equation, we can compute the growth of total factor productivity as follows:

∆ , ,

∆ , ⁄ , is the change in output that cannot explain by the change of inputs Thus,

the growth in total factor productivity is estimated as a residual This residual is the remaining of output after we have accounted for the determinant of growth that we can measure Indeed, ∆ , ⁄ , is also called technological progress, or sometimes called

Solow's residual

Total factor productivity can change for several reasons The most notable reason is the change of knowledge accumulation The other reasons might be education and government regulation or government size that can affect to the change of technological progress as well If government regulations require firms to consume capital to reduce

environmental pollution or increase the work safety for workers, then the amount of capital might increase in out and it implies a decrease in technological progress

If we divide both side of equation (1) by L and assume that + = , where ∈{ < 1; > 1; = 1}, we have the following equation:

=1 , ,

, (5) Apply approximation method we have the following expression:

The interest of this paper is the analyzing the relationship between government size and economic growth and technological progress It is assumed that the changes size of government require government purchase more goods and services from markets, then it will likely boost the growth rate of economics On the other hand, the expenditure of government might result in a change in demand of accumulating more scientific knowledge

to serve the further increasing or expansion of economies

From (1), the following equation is introduced:

, = , , , , , = , , , , (7) The equation (7) different to (1) is on the term government size, measured by G Government expenditure is a element of aggregate demand, so increase in government expenditure will likely increase in aggregate demand if all the other variables are held constantly Practically, raising government expenditure not always leads to the increase in aggregate demand Therefore, in the equation (8) government size is treated as exogenous variable The exogenous variable might have several thresholds where it switches its effect

on economic growth and technological progress It implies that in empirical model, we should consider utilizing the two stages regression models

Applying rules of transformation from equation (1) to (6) on equation (7) we have a system of the following equations:

∆ , ,

∆ , ,

in different thresholds

Equation (10) allows us to measure the effect of both technical change and government size on the growth of labor productivity The difference of technical change variable in equation (10) to equation (9) is that the effect of government size on technical change is taken out Therefore, technical change and government size variables in equation (10) should not be correlated each other

4 Research method and data

4.1 Research method

To examine the relationship of government size with economic growth and technical change, we estimate the following equations based on the light of equation (3), (8) to (10):

∆ , ,

country-Equation (11) and (12) assume that there is a complex and short-run relationship between of government size with economic growth and technical change Necessary conditions to hold this assumption are that individual time series for∆ , ⁄ , , ∆ , ⁄ ,

and ∆ , ⁄ , are stationary and integrated of the same order and that ∆ , ⁄ , , ∆ , ⁄ ,

and ∆ , ⁄ , form a cointegrated system

In the presence of cointegration, the effect of government size to economic growth and technical change can be estimated by applying the group-mean dynamic OLS and fully modified OLS estimators suggested by Pedroni (2001) The both estimators allow for a greater flexibility in the occurrence of heterogeneous cointegrating vectors As suggested in Kim et al (2017), these estimators are robust to the omission of variables that are not parts

of cointegrating relationship Moreover, they can be shown as the expectation value of cointegrating vectors

However, when all individual time series are stationary and cointegrated we should not utilize fully modified OLS as suggested in Kim et al (2018) In this case, we will apply mixed-effects regression models to address fixed and random effects of independent variables on dependent variable Therefore, the three above equations will be transformed

to the following equations for estimating purposes:

∆ , ,

= + ∆ ,

,

+ , + , + , (13)

∆ , ,

= + , + , + , (14)

In the equation (14) and (15), the new variable is "government size effect" while the estimated coefficients are fixed effects Fixed effects are stable within group of countries sharing similarities of geographical and climatic conditions or cultural traditions, while random effects are the variation among countries regard to time, countries and regions We can observe fixed effects while we cannot observe random effects

4.2 Data

To investigate the relationship of government size with economic growth and technical change, we consider a panel data of 183 developed and developing countries which cover period of 1970 to 2014 Government size is measured as growth of the share of government expenditure to GDP (following Fölster and Henrekson (2001), Bergh and Karlsson (2010)) and growth rate of government expenditure Penn World Table version 9.0 (hereafter PWT9) provides useful source of data on government expenditure as share of GDP; base on this information we can calculate the absolute value of government expenditure Growth rate of GDP is also estimated based on data taken from the PWT9

In this paper, economic growth is measured as growth of labor productivity and technical change is measured as growth of total factor productivity Total factor productivity

is estimated through equation (3), while labor productivity is measured as GDP per worker

Table 1 Descriptive statistics Growth of GDP

per labor

Growth of capital per

labor

Growth of Government size Panel A: Summary statistics

Max 1.986 0.813 0.964

Panel B: Panel unit root test and cointegrated test

Source: Authors' estimation

In the following figures, we show the relationship of government size with economic growth and technical change in the three large continents The pattern of the relationship between government size and technical change is similar among three continents (figure 2), while the relationship between government size and economic growth in Asian and Latin American Nations is different to European Nations

It also implies that the pattern of relationship between government size and technical change is consistence while the pattern of the relationship between government size and economic growth is not consistence through these nations Additionally, different effect of government size to economic growth reflects the fact that quality of governments on spending their money might affect to the performance of economic growth

The mixed pattern effect of government size on economic growth and technical change in figure 1 and 2 remind us that there should be a non-linear relationship of government size to economic growth and technical change The mixed pattern requires a more complex data analysis techniques rather than a simple data analysis framework

Figure 1 Relationship between government size and growth of GDP

Source: Authors' estimation

Figure 2 Relationship between government size and total factor productivity

Source: Authors' estimation

5 Empirical results

As can be seen from table 2 to 5, there are three models with different group of independent variables The model 1 consists of our interested variable while in the other two models we add more control variables to control some national characteristics In the model 2, we utilize two variables related to human education and health, whereas in the model 3 we add four variables including two variables related to human education and health and the other two proxied for economic status: openness and dependent Openness measure the ratio of export and import to GDP in each country while dependent variable measure the ratio of active workers to the rest of population

In this section, we will present two stories of empirical results based on two different set of proxied variables measuring government size The first story is about absolute value and the other story is on the side of relative values

5.1 A story of absolute value

From table 2, the effect of growth of capital per worker to economic growth is large and statistical significance at 1% With the appearance of government, the increasing ratio

of capital to labor by 1% will likely increase 0.632% economic growth The value of this variable is quite consistent in the three models, it varies from 0.604 to 0.632 It implies that

our selection of functional form is quite consistence with the other empirical studies utilizing endogenous growth theory as major theoretical framework

Table 2 The effect of government size on economic growth

Dependent variable: Growth of GDP per worker Growth of Capital per worker 0.632***

(0.015)

0.605***

(0.017)

0.604*** (0.017)

(0.002)

0.004** (0.002)

(0.0001)

-0.0002** (0.000)

Note:

Standard Errors are in parentheses;

*, **, and ***: statistical significance at 10%, 5% and 1% respectively

Source: Authors' estimation

The effect of growth of government size (measuring directly from absolute value of government expenditure) on economic growth is quite consistent through three models as well as growth of capital per active worker All estimated coefficients of government size are statistically significant at 1% It implies that these empirical results should be considered The effect of linear term of government size is changed from 0.097 to 0.124 when we add more control variables The changing of quadratic term of government size varied from 0.018 to 0.091, and the changing of cubic term of government size is from 0.001 to 0.015

Model 1 shows us that without any control variables, the effect of government size

on economic growth is low, while the other models with control variables bring to us higher effect of government size on the growth of economics The meaning of this situation is that control variables might take some more information from error terms that government size cannot take it out by itself From empirical result of model 3, if government decides to increase it consumption by 1% it will likely increase economic growth by approximately

0.124% In a longer term, increase 1% of government size will likely increase economic growth by around 0.21% Therefore, the total effect of changing in government size and ratio

of capital to active worker will closely to 1 It means that in this case, utilizing endogenous growth theory might explain the result better than neo-classical economic growth

Table 3 The effect of government size on technical change

Dependent variable: Growth of TFP

(0.015)

-0.022 (0.017)

-0.023 (0.017)

(0.002)

0.004** (0.002)

(0.000)

-0.0002** (0.0001)

0.0001 (0.006)

Note:

Standard Errors are in parentheses;

*, **, and ***: statistical significance at 10%, 5% and 1% respectively

Source: Authors' estimation

The changing of ratio of capital to labor does not have any effect on the growth of total factor productivity as our expectation It is true because total factor productivity is taken as a residual of growth regression model, so under the assumption that residual of regression must not have any relationship with regressors This situation is hold for all three models in the table 3

Growth of government size does have effect on the technical change of the economy All three estimated coefficients are statistically significant at 1% The value of estimated coefficients of linear term of government size varied from 0.097 to 0.124 The value of the quadratic term of government size varied from 0.018 to 0.091 and for the cubic term of government size, the changing varies from 0.001 to 0.015 The pattern of effects of government size on technical change is quite similar to these effects on economic growth

This result has not surprised us because technical change and economic growth should have same pattern and statistical distribution Thus, the effect of the growth of government size

on the growth of economics and technical change should have the same patterns

Without any control variables, increase government size by 1% will likely lead to an increase of approximately 0.12% in the changing of technology Under the co-effect with related human variables, the effect of government size on technical change is increase from 0.12% to almost 0.21% In the full model, the effect increase to nearly 0.22% The more control variables we add in the estimation model the more information we can take from error term, and it likely help to raise the effect of the growth government size to technical change This result implies that the effect of growth of government size on technical change is somehow scaled up through the interaction with control variables

The empirical results do not tell us much information about the thresholds that the effect of government size on economic growth and technical change might be switch from positive to negative because functional form is non-linear Based on information about estimated coefficients of the effect of government size on economic growth and technical change in table 2 and 3, we demonstrate the graph of these relationship

As can be seen from figure 3, the dash curve has only one threshold where the effect

of government size on economic growth, while the solid line express two thresholds where the effect of government size on technical change will be switch from positive to negative and vice versa At the beginning of analyzing period, the effect of the growth government size on economic growth and technical change is below 0 and quite similar Then, the pattern

of these effects is diverged and showed differently Although economic growth and technical change have similar pattern, the effect of growth of government size to them is not similar

Figure 3 Effect of government size on economic growth and technical change

Source: Authors' estimations

5.2 Another story of relative values

From table 4, from model 7 to 9, the relationship between growth of capital per worker and growth of GDP per worker is stable and statistical significance at 1% This relationship should be consistence with model 1, 2, and 3 because we are using the same variables The more important thing is that under the new selection of government size variables, the effect of growth of capital per worker to growth of GDP per worker is similar

It again show us that the functional form we select is appropriate for this analysis

Table 4 The effect of government magnitude on economic growth

Dependent variable: Growth of GDP per worker Government size is measured in changing share of government expenditure to GDP

Growth of Capital per worker 0.647***

Effect of Government size on economic growth

Effect of Government size on Technical Change

Model 7 Model 8 Model 9 Dependent variable: Growth of GDP per worker

Government size is measured in changing share of government expenditure to GDP

(0.0001)

-0.0003** (0.0001)

Note:

Standard Errors are in parentheses;

*, **, and ***: statistical significance at 10%, 5% and 1% respectively

Source: Authors' estimation

In this section, we measure government size as a change in the share of government expenditure to GDP All estimated coefficients related to government size are negative and statistical significance at 1% Negative effect of changing in the share of government expenditure to GDP to economic growth implies that the relative size of government to its economy should be restricted Government cannot expand their size over time Instead, in specific periods, relative government size should be downsized and for some period the size

of government should be enlarge The enlargement of government size in terms of expenditure might squeeze the size of the firms and the other in the economy

The linear term of relative government size has estimated value ranging from -0.072

to 0.076 Quadratic term of relative government size has estimated value ranging from 0.008 to -0.074 and cubic term of relative government size has estimated value ranging from -0.0002 to -0.024 Therefore, if the share of government size to GDP increase by 1%, it will likely decrease economic growth by at least 0.08% with a cap is approximately -0.15%

-The statistical significance of quadratic and cubic terms of government size also confirm the long-run effect of the changing size of the government when we select the other way to measure government size

Table 5 The effect of government magnitude on technical change

Dependent variable: Growth of TFP Government size is measured in changing share of government expenditure to GDP

(0.015)

-0.002 (0.017)

-0.0001 (0.017)

Model 10 Model 11 Model 12 Dependent variable: Growth of TFP

Government size is measured in changing share of government expenditure to GDP

(0.002)

0.001 (0.002)

(0.0001)

-0.0003** (0.0001)

Note:

Standard Errors are in parentheses;

*, **, and ***: statistical significance at 10%, 5% and 1% respectively

Source: Authors' estimation

As can be seen from table 5, estimated coefficient of growth of capital per worker is not statistical significance as our expectation The effect of changing the share of government expenditure to GDP to the growth of TFP has similar pattern to the effect in economic growth equation All estimated coefficients related to the term of government size are statistical significance at 1%

The linear term of government size affect to the growth of TFP within the range from -0.072 to -0.076 The quadratic term of government size has a ranged effect from -0.008 to -0.074, and the cubic term of government size has a ranged effect from -0.0002 to -0.024 Therefore, increase the share of government size to GDP by 1%, it will likely decrease the growth of TFP by at least 0.08% to approximately 0.15% The increasing the size of government in this sample will likely lead to a squeeze of technical change Increasing government size, thus, is not always good for the economy

As can be seen from figure 4, the shape of the two curves is quite similar It is different to the situation in figure 3 The both curves has two thresholds When we look at the dash-curve, the two thresholds are above 0, while the solid curve has one threshold under 0 and the other is above However, after the second threshold, the two curves show

us a sharp down It implies that for the effect of the changing the share of government on economic growth the second switch is critical If government continuously increase its size,

the largely negative effect on economic growth and technical change will likely appear It can be understood as a punishment to government when it becomes too large to control

Figure 4 Effect of changing share of government expenditure to GDP on economic growth and

technical change Source: Authors' estimation

5.3 Measuring the effect of government size and innovative activities to economic growth

Going through all empirical evidence so far, we understand that government size will affect directly and indirectly, through technical change, to economic growth Therefore,

in this section we will analyze the effect of technical change and government size to economic growth in only one equation To do so, we will take out the effect of government size to technical change Then, we put the two variables into one equation to estimate their effect to economic growth The empirical results will be presented in the following table

Our main interest here is about the effect of government size to economic growth All estimated coefficients related to government size is statistically significant at 1% and has negative values The estimated coefficient of linear term of government size has a range from -0.063 to -0.059 For the quadratic term, the effect has a range from -0.007 to -0.017, and for cubic term, the effect has a range from -0.0002 to -0.005 Increase government size by 1%,

it will likely decrease economic growth by at least -0.07% to approximately -0.066% When government size is measured as government expenditure rather than government investment, the increase government expenditure will likely limit the economic growth rather than to boost economic growth as government's expectation