Relationship between public expenditures and economic growth in budget constraint

Based on the panel data of 66 developing countries in the period of 1998 -2106, the study find that fiscal surplus enhances the growth effects of public unproductive expenditure and publ

Relationship between public expenditures and economic growth in budget constraint

TRAN TRUNG KIEN

University of Economics HCMC – kientt@ueh.edu.vn

Abstract

This paper investigates relationship between public expenditures and economic growth in budget constraint Public expenditures are divided into productive expenditure and unproductive expenditure The budget constraint is conditional on public debt and fiscal balance Based on the panel data of 66 developing countries in the period of 1998 -2106, the study find that fiscal surplus enhances the growth effects of public unproductive expenditure and public productive expenditure, while the growth effect of public debt is only significant in the case unproductive expenditure when

it exceeds 17.025% GDP

Keywords: economic growth; public expenditure; budget constraint

1 Introduction

Many theories explain the impact of public expenditure on economic growth Wagner law focuses on the causal link between public expenditure and growth (Al-Faris, 2002; Bagdigen & Cetintas, 2003) As Keynesians say, public expenditure is a component of growth equation (Keynes, 1937) Endogenous models consider public expenditure by the extension of AK model (Barro, 1990; Devarajan, Swaroop, & Zou, 1996; Mankiw, Romer,

& Weil, 1992)

Empirical evidence has been still ambiguous When some papers indicate the positive effect of public expenditure on growth (Grier & Tullock, 1989; Holmes & Hutton, 1990; Lin, 1994; Ram, 1995), others imply the opposite one (Easterly & Rebelo, 1993; Fölster & Henrekson, 2001; Landau, 1985; Marlow, 1988; Tanninen, 1999) Meanwhile, some recent studies conclude the nonlinear growth effects of public expenditure (Chen & Lee, 2005; Abounoori & Nademi, 2010; Herath, 2012; Altunc & Aydin, 2013)

Explaining the heterogeneity of the empirical results, some scholars point out the rationale of this one is the difference in public expenditure composition across the countries (Bayraktar & Moreno-Dodson, 2015; Gemmell, Misch, & Moreno-Dodson, 2012;

Martins & Veiga, 2014) A general increase in the scale of public expenditure leads to the apprehensiveness that increasing public spending would hurt the economy, causing negative externalities such as crowding-out effects, inefficient distribution However, the provision of public goods and public services like education, infrastructure and healthcare system are necessary for economic growth (Martins & Veiga, 2014) Some studies examining the growth effect of public expenditure with general measure often ignore trade-offs effects of specific public expenditure components If the budget constraint is constant, an increase in a component of public expenditure must be offset by the decrease

in other public expenditure components (Gemmell et al., 2012; Martins & Veiga, 2014)

Besides the disparity of public expenditure composition, the state of budget constraint

is also a rationale of the difference in growth effect of public expenditure As Gemmell et

al (2012) explain, budget constraint plays an important role in examining the impact of public expenditure on economic growth empirically because an increase in public spending generally requires funding, usually from tax increases However, the government can not forever raise taxes With limited resources, public expenditure financing comes not only from tax increases but also from debt financing, so public debt also plays an important role in the relationship between public expenditure and growth ( Zagler & Dürnecker, 2003)

In this study, we explore the role of public expenditure composition and budget constraint in public expenditure and economic growth nexus in the case of developing countries, both in linear and non-linear context Following Martins and Veiga (2014), there is a large difference between developed and developing countries, so we concentrate

on developing countries As far as we’re concerned, there are few of studies examining simultaneously the role of public expenditure components, fiscal balance and public debt

in public expenditure and economic growth nexus in developing countries, especially in non-linear context

This paper is designed as follows Section 2 summarises the literature review on public expenditure and growth nexus The empirical models are presented in Sect 3 and the results are displayed and analysed in Sect.4 Finally, Sect.5 is the conclusion

2 Literature review

Empirical studies researching on the relationship between public expenditure and growth are developed in many ways In general, based on research objectives, these studies are able to be classified into three groups:

Firstly, based on Wagner's Law, many studies aimed at exploring the causal relationship between public expenditure and economic growth in different cases (Akitoby, Clements, Gupta, & Inchauste, 2006; Bagdigen & Cetintas, 2003; Chang, Huang, & Yang, 2011; Devlin & Hansen, 2001; Erdil & Yetkiner, 2009; Ghorbani & Zarea, 2009; Huang, 2006; Islam, 2001; Lamartina & Zaghini, 2011; Magazzino, 2012; Wu, Tang, & Lin, 2010)

In particular, the most commonly used method is the Granger causality test for time series and panel data However, focusing on the causal link between public expenditure and economic growth, the role of other growth factors as well as the interactions between public expenditure and others have not been yet analysed thoroughly

Secondly, some studies focus on analysing the short-term impact of public expenditure

on economic growth These studies have also been developed in various ways, such as analysing the impact of a shock of public expenditure on macroeconomic factors (Beetsma

& Giuliodori, 2011; Beetsma, Giuliodori, & Klaassen, 2008; Burriel et al., 2010; Ramey, 2011) or exploring the public expenditure multiplier in each case (Auerbach & Gorodnichenko, 2014; Caldara & Kamps, 2008; Ramey & Zubairy, 2014) The empirical methods used commonly in these studies are VAR methods for time series and panel data

As usual, they use the Keynesian models to examine demand-side effects resulting from price rigidities and credit-constrained consumers and analyse investment and consumption responses explicitly (Gemmell et al., 2012) However, as Gemmell et al (2012) discussed, these studies don’t have a tendency to separate productive and unproductive public expenditure: public expenditure isn't analysed as a component of the production of private sector output, so one does not have an effect on the productivity of private sector inputs Consequently, its effect predicted by these models is determined mainly from other transmission channels that make it difficult to explore the effect of public expenditure components

Thirdly, based on endogenous growth theory, many studies are concerned with the effect of public expenditure on economic growth in the long-term In the early period, some studies explore this effect with general public expenditure measure (Agell, Lindh, & Ohlsson, 1997; Bairam, 1990; Fölster & Henrekson, 1999; Ghali, 1999; P Hansen, 1994; Landau, 1985; Sattar, 1993; Sheehey, 1993) They usually ignore the difference in public expenditure composition and the budget constraint so the empirical results are controversial Therefore, several recent papers focus on the role of public expenditure composition or budget constraint in this nexus (Benos, 2009; Bleaney, Gemmell, & Kneller, 2001; Bose, Haque, & Osborn, 2007; Fan, Yu, & Saurkar, 2008; Romero-Avila &

Strauch, 2008; Teles & Mussolini, 2011) However, these empirical studies are mainly implemented in the linear context

3 Empirical model

As analysed, the empirical model of this study is based on Barro framework In extension, we categorise public expenditure into productive expenditure and unproductive expenditure (Devarajan et al., 1996; Kneller, Bleaney, & Gemmell, 1999) and analyse the role fiscal balance and public debt (Gemmell et al., 2012; Teles & Mussolini, 2011) in the empirical model In particular, productive expenditure is public expenditure components used for education, health care, transport and communication (Devarajan et al., 1996; Kneller, Bleaney, & Gemmell, 1999) and unproductive expenditure

is extant public expenditure components Following Barro (1990), Mankiw et al (1992) Devarajan et al (1996) and Cooray (2009), the basic empirical model like that:

d"#$%& = α)*+ β."#y)*0.+ β1"#k)*+ β3"#h)*+ β5"#g.)* + β7"#g1)*+ β8Z)*+

ηi ~ i.i.d (0, ση ) ; εit ~ i.i.d (0, σε ) ; E(ηiεit ) = 0

GPP per capita growth rate (dlny)* = lny)*− lny)*0.) = The first difference of log of GDP per capita in each country

Private investment (lnkit) = Log of private investment to GDP in each country

Human capital (lnhit) = Log of human capital indicator

Productive expenditure (lng1) = Log of productive public expenditure components (education, health care, transport and communication) to GDP

Unproductive expenditure (lng2) = Log of other public expenditure components (except for productive public expenditure components) to GDP

Zit is a collective of control variables, such as public debt (dit) and trade openness (tradeit) All of them are in logarithm The choice of control variables is based on previous papers (Aizenman, Pinto, & Radziwill, 2007; Baldwin, 2004; Checherita-Westphal & Rother, 2012; Dowrick & Golley, 2004; Elmendorf & Mankiw, 1999; Grossman & Helpman, 1992; Harrison, 1996; Jin, 2000; Rodríguez, 2006; Saint-Paul, 1992)

Similar to Teles and Mussolini (2011), we use interaction variables to examine the role

of fiscal balance (surplusit) and public debt (dit) Moreover, we include interaction

variables not only between productive expenditure with fiscal balance and public debt but also unproductive expenditure with ones:

d"#$%& = α)*+ β."#y)*0.+ β1"#k)*+ β3"#h)*+ β5"#g.)* + β7"#g1)*+ β8Z)*+

dln$%& = α)*+ β.lny)*0.+ β1lnk)*+ β3"#h)*+ β5"#g.)* + β7"#g1)*+ β8Z)*+

dln$%& = α)*+ β.lny)*0.+ β1lnk)*+ β3"#h)*+ β5"#g.)* + β7"#g1)*+ β8Z)*+

β@g.)*∗ d)*+ βBg.)* ∗ surplus)*+ η)+ ε)* (4) d"#$%& = α)*+ β."#y)*0.+ β1lnk)*+ β3"#h)*+ β5"#g.)*+ β7"#g1)*+ β8Z)*+

dln$%& = α)*+ β.lny)*0.+ β1lnk)*+ β3"#h)*+ β5"#g.)* + β7"#g1)*+ β8Z)*+

dln$%& = α)*+ β.lny)*0.+ β1lnk)*+ β3"#h)*+ β5"#g.)* + β7"#g1)*+ β8Z)*+

To estimate the proposed models, we use a panel data with the cross-section dimension of up to 66 developing countries and the time dimension covering the period 1998-2016 The majority of data are collected from International Monetary Fund (IMF), except for public expenditure components and human capital indicator Public expenditure components are gathered from Statistics on Public Expenditures for Economic Development (SPEED) của International Food Policy Research Institute (IFPRI) and human capital is from Penn World Table 9.0 (PWT 9.0) of Groningen Growth and Development Centre- Groningen University

Table 1

Descriptive statistics

GDP per capita (y) 5151.858 7088.306 111.531 54484.3

Productive expenditure (g1) 6.609 3.621 0.181 20.681 Unproductive expenditure (g2) 21.413 8.063 3.730 58.144

Source: IMF (2017); PWT 9.0, IFPRI

We employ generalised method of moments (GMM) approach to estimate these models in the case of 66 developing countries 1998-2016 Following Arellano and Bond (1991), this approach may be useful to estimate a dynamic model from a panel data consistently and efficiently However, GMM approach requires careful consideration in the selection of instruments and regressors in each equation An equation may be over-identified, exactly identified or under-over-identified, depending on the numbers of instruments There is no instruction to decide how many instruments are suitable (Roodman 2009) Roodman (2009) suggests a rule of thumb that instruments should not be larger than the number of cross-sections Therefore, this study implements Arellano-Bond difference GMM as the difference GMM needs fewer instruments than system GMM

In addition, we apply the Hansen’s threshold estimation to examine the non-linear growth effects of public expenditure components and explore the threshold values of these public expenditure components Although this approach is based on fixed-effect estimation, a bootstrap technique with 500 replications is employed that makes the empirical results more efficient In extension, we use fixed-effect estimation to analyse the role of fiscal balance and public debt in the public expenditure and economic growth nexus in each threshold regime1

4 Result

4.1 In linear context

The empirical results of the model (1), (2), (3), (4), (5), (6) and (7) are respectively presented in Table 1 The results show that the effects of most variables are similar to the theoretical expectation The negative effect of lnyt-1 reflects the convergence theory, the poor countries have faster growth rate to catch up with the rich countries Private investment and human capital have positive effects on economic growth that reflects the endogenous growth theory

Meanwhile, trade openness, although the significance of its effect is not clear, has a negative impact on economic growth This result is similar to Yanikkaya (2003), also in the case of developing countries Yanikkaya (2003) points out trade barriers have the positive impact on economic growth in developing countries Explaining this result,

1

The GMM approach is ineficient because of the small number of cross-sections in each sub-group

Yanikkaya (2003) implies that trade barriers such as tariffs would help protect young industries in developing countries as well as implement strategic trade policies

Similar to Presbitero (2012); Bal and Rath (2014); Zouhaier and Fatma (2014), the empirical results indicate a significant negative impact of public debt on economic growth

in developing countries Presbitero (2012) argues that developing countries borrow and use loans inefficiently that enhances the negative impact of public debt on investment and causes more policy fluctuations Moreover, in developing countries, the cost of over-borrowing outweighs the benefit of having more resources

Like previous studies, in this case, the productive expenditure components (education, health care, transport and communication) have positive effects on economic growth while unproductive components have the opposite effects (Barro, 1990; Devarajan et al., 1996; Fan et al., 2008; Kneller et al., 1999) Barro’s model implies that an increase in resources used for unproductive government spending is associated with lower economic growth while productive government spending boosts economic growth As Fan et al (2008) argue, developing countries should scale down their public expenditure in unproductive sectors such as military, and reduce exaggerated subsidies in irrigation, fertiliser, pesticides and energy

Meanwhile, the interactions between public expenditure components and fiscal balance have positive significant effects on economic growth According to Teles and Mussolini (2011), an increase in the productive public expenditure has a positive effect on economic growth However, the marginal effect of this one depends on the state of fiscal balance As productive expenditure rises in the context of large fiscal deficits, the government is forced to borrow Debt expansion has a negative impact reducing the positive effect of productive expenditure on economic growth Therefore, when the fiscal deficit is small, it makes less the negative effect on the growth effect of productive expenditure Similarly, the empirical results imply that the level of fiscal deficit exacerbates the negative effect of unproductive expenditure on economic growth Consequently, regardless of whether an increase in public expenditure is productive or unproductive expenditure, the smaller the fiscal deficit is, the less the negative impact on growth effect of public expenditure is created

The empirical results indicate significant negative effects of the interaction between public expenditure components and public debt on economic growth The marginal effects of public expenditure components on economic growth are affected by the size of public debt An increase in public spending leads to higher interest rates, while

government pay interest on loans (Teles & Mussolini, 2011) Large interest payments overwhelm government spending, so governments with large public debt tend to cut spending on public investment in order to interest payments (Lora & Olivera, 2007; Sturm, 2001)

Table 2

The role of public expenditure components and budget constraint in public expenditure and economic growth nexus in developing countries (1998 – 2016)

*, **, and *** show the statistical significance at 1, 5, 10 % levels respectively.

4.2 In non-linear context

We examine the growth effects of public expenditure components in non-linear context The productive expenditure and unproductive expenditure are respectively found at 2.838% GDP and 17.025% GDP The empirical results are presented in Table 2

Table 3

The effect of public expenditure on economic growth in developing countries in non-linear context (1998-2016)

g1 ≤ 2.838%

GDP

g1 > 2.838%

GDP

g2 ≤ 17.025%

GDP

g2 > 17.025% GDP

Private investment (lnkit) 0.134*** 0.097*** 0.024 0.177*** Human capital (lnkit) 0.605*** 1.066*** 1.106*** 0.969*** Trade openness (lntrade it ) -0.039*** -0.016 -0.019 0.004 Public debt (lnd it ) -0.057*** -0.104*** -0.091*** -0.095*** Productive expenditure (lng1it) -0.006** 0.042* 0.001 0.006 Unproductive expenditure (lng 2it ) -0.057 -0.073*** 0.040 -0.170***

*, **, and *** show the statistical significance at 1, 5, 10 % levels respectively.

In general, productive expenditure has a positive impact on economic growth However, if its scale is extremely small (g1 ≤ 2.838% GDP), productive expenditure is too weak to support economic growth Until it passes the threshold value (g1 > 2.838% GDP), productive expenditure is large enough to have the positive effect on economic growth Meanwhile, unproductive is typically detrimental to economic growth Nonetheless, its negative effect is only significant when it is larger than the threshold value (g2 > 17.025% GDP)

In addition, we explore the role of public expenditure components and budget constraint in each threshold regime by adding suitable interaction variables The results are displayed in Table 3