2. LITERATURE REVIEW
4. EMPIRICAL RESULTS
A DEVELOPMENTAL COMPARATIVE ANALYSIS OF THE NATURE OF SUB-OPTIMAL PUBLIC CURRENT EXPENDITURE SHARES OF TOTAL PUBLIC EXPENDITURE
ABEL B.S. GAIYA *
An investigation into the patterns of general government current expenditure shares of total expenditure, as well as patterns of optimality and sub-optimality, between developed and developed countries, has been undertaken. Discussion over this metric has largely been limited to publicized comments, the informal blogosphere and news commentary; this paper presents a contribution towards the subject. Using econometric methods and statistical analysis, a wide range of results are uncovered regarding the cross-developmental patterns of current shares, their optimum values and sub-optimality. It is suggested that future research be directed at expanding and deepening the scope and technical rigour of the analysis.
Keywords: Public economics, National Government Expenditures and Related Policies,
JEL: H0, H5
On the 9th of December, 2011, a prominent Nigerian newspaper, Vanguard, reported the then Nigerian minister of finance, Ngozi Okonjo-Iweala, insisting that the federal government’s recurrent expenditure was too high and needed reduction: “When I left the administration last time (2006), we left a recurrent expenditure level of about 66 per cent [of total public expenditure], which was even considered to be too high back then. However, upon my second coming this year, I was surprised to find that recurrent expenditure in the budget has risen to more than 74 per cent. This is unacceptable.” (Onwuemenyi, 2011). Her sentiments were shared by the former governor of the Central Bank of Nigeria (Vanguard, 2012).
Moving to the east of Africa, similar concerns manifest in Kenya. For instance, on the 18th of August, 2009, Mr Mutava Musyimi, a Kenyan Member of Parliament, noted (during the legislative assessment of the executive budget proposal) that the Keynan public recurrent expenditure was “much too high” for a country that wants to develop (Kenya National Assembly, 2009). This is similar to Chenga’s (2005) media report on the concerns of the Zimbabwean populace with regards to the level of public recurrent expenditure being deemed too high at around 90 per cent.
Similar concerns over the size of public recurrent expenditure as a share of total public expenditure manifest in many other developing countries. However, it is surprising that they have largely been limited to news media, political statements, and the informal blogosphere without much comprehensive and technical assessment.
Notwithstanding, the primary aim of this research is not to provide estimates of recurrent shares for countries per se, but to do so for the purpose of assessing how they compare across developed and developing countries, and the characteristic behaviour of recurrent shares with respect to the optimum between these two developmental groups.
In other words, it is a comparative analysis of the pattern of current expenditure share of government expenditure.
In the first section, a review of the literature is presented. Both theoretical formulations and empirical analyses relevant to the topic are expounded upon. Existing criticisms in the literature are mentioned, and new criticisms are presented in this section. The review encompasses literature on the determinants of government expenditure (disaggregated), theories on the structural evolution of composition of government expenditure, and theories on the optimal distribution of government expenditure in terms of growth-maximization.
The next section outlines and explains the methodology adopted for empirical analysis. A comparison of the current expenditure shares of total general government expenditure between developed countries and developing countries. Performing this comparison would prove helpful in understanding the developmental significance, if any exists, of recurrent shares, optimal shares, and patterns of sub-optimality.
2. LITERATURE REVIEW
This section examines the theoretical and empirical literature related to determinants of the compositional shares of government expenditure, the literature on the determinants of compositions of government expenditure, the literature on economic growth effects of compositions of government expenditure, and the literature on the optimality of compositional shares of government expenditure. These categories are however intertwined and intuitively interrelated. These relationships are noted in the review. It must be noted that expenditure-disaggregated analyses are less prevalent in the empirical literature than aggregated analyses. Hence, there is a fundamental limitation to the breadth of academic works relevant to the topic of this research.
Wagner’s (1883) law of increasing state activity is a popular theory of government expenditure. The theory predicts that as an economy grows, the government expands its role, and hence government size relative to the size of the economy grows as well. As explained by Ighodaro and Oriakhi (2010:186), three main factors for increased government spending were identified by Wagner. First, administrative and protective role of government increases as the development level of the economy increases. Second, government expenditures on “cultural and welfare” would rise, particularly on education and health, with the economy’s expansion. Implicitly, he assumed that the income elasticity of demand for public goods is more than unity. Lastly, progress in technology requires of developed nations governments to undertake certain economic services for which private sector may shy away from.
To empirically investigate this link between expenditure and national income, an examination of the income elasticity of expenditure is conducted. The condition for concluding Wagner’s hypothesis to be validated is that this estimated elasticity is greater than one, with a positive coefficient sign, and is of statistical significance (Hadjimatheou, 1976; Jackson, 1980; Diba, 1982). Using both cross-sectional and time series datasets, Wagner’s law has empirically been assessed for a broad range of developed and developing countries (Magazzino, 2010:3). A classification by Sideris (2007:6) categorizes the empirical research on the hypothesis into two groups: early studies conducted up to the mid-1990s (stationarity of data series was assumed and simple OLS regressions applied to test alternative versions of the law (Ram, 1987; Courakis et al., 1993); and studies based on Cointegration (performed from the mid-1990s till present). In addition, Granger causality tests to indicate causality direction between public spending and national income are also used in most of the recent studies (Wahab, 2004; Henrekson, 1993; Biswal et al., 1999; Islam, 2001; Burney, 2002;).
Disaggregated analyses are much less prevalent in the literature. Those which do so use function-based disaggregation (that is spending on public services, defence, health, education, environmental protection, and so on), and empirical studies which test the hypothesis using economic type spending (current and capital) disaggregation (in absolute terms or as proportion of total public spending) are relatively scarce in the literature. Hence, there is a limitation limited in what relevant papers examining Wagner’s law at the disaggregated economic-type level can be found in the literature relevant to this context. There are, nonetheless, some empirical papers on the subject that prove useful for this review.
Magazzino (2010a: 9) studies the linkages between public expenditure at a disaggregated level and GDP for Italy. Empirical evidence suggests that only for gross public investment expenditure the hypothesis is satisfied. Instead, Granger-causality exhibits unclear results: the direction of causality from public spending to aggregate income is observed for these categories of public expenditure: final consumption, public wages, gross public investment, and contribution to production. Albatel (2002:151) finds, using Cointegration and error correction analysis that Wagner’s law holds in Saudi Arabia, especially for government expenditure on services, investment, and consumption.
There are mixed results in the empirical literature on Wagner’s law. Bohl (1996) documents some of the conflicting studies. It is possible that the wide cross-country disparity in the results from tests of Wagner’s law (which show mixed and sometimes contradictory results), as noted by Magazzino (2010b: 4) is attributable to factors beyond variation in econometric approaches used, such as the diverse features exhibited by different countries during alternative periods of time. For example, it may be possible that countries whose aggregate output growth is driven industrially rather than by extractive sectors exhibit greater governmental expansion due to greater urbanization and hence greater demand for government services, or as suggested by a recent study by Lamartina and Zaghini (2011:161) on 23 OECD countries, which finds that the correlation between economic growth and government activity is higher in countries with low per-capita, according to Wagner’s hypothesis, the direct linkage between increasing state activity and economic growth might have a higher validity during early stages of development than at a later stage.
Interestingly, as Haque (2004:16) documents, an observation in the international public finance data is that high income countries on average have higher shares of public current expenditures in public spending, and as a share of GDP than developing countries. Furthermore, International Monetary Fund (2012) data (average over 1983 to 1990) clearly shows that the share of both central and government current expenditure as shares of GDP and total expenditure are, on average, higher for high income countries, and for middle income countries they are higher than low income countries but lower than high income countries. This demonstrates a steady upward transition of recurrent shares across developmental categories.
This may suggest that Wagner’s state increasing mechanism is current expenditure compositionally-intensive. The implication of this may be that as countries transition between developmental states, public current expenditures may be gaining increasing shares in total public expenditure and GDP alike. Hence, the long-run growth of public expenditure composition shares may be dependent upon the structural evolution of society.
The ratchet hypothesis, first formulated by Peacock and Wiseman (1961), is another theory which may prove relevant for this topic. According to Halcombe, (2005:100) it is, “the theory is that government responds to crises like wars and depressions by ratcheting up expenditures, and then, after the crises pass, expenditures fall somewhat but remain above their pre-crisis level”.
The hypothesis has seen empirical application in the area of economic cycles. Backus, Kehoe and Kydland (1995), Gavin and Perotti (1997) and Talvi and Vegh (2000), among others, have investigated the cyclical pattern of government expenditure in industrial economies. In particular, Gavin and Perotti (1997:21) present evidence of asymmetrical ﬁscal behaviour over the cycle: government consumption is moderately procyclical in expansions, while in contractions government consumption and transfers are strongly countercyclical. The contribution of this paper to this literature is to test the link between this asymmetry and the spending/output drift.
Hercowitz and Strawczynski (2004:21) using a panel of 22 OECD countries, found that cyclical ratcheting is exhibited in the three components of government spending – government consumption, transfers and subsidies, and capital expenditure (all as ratios of GDP). The estimates imply that following an artiﬁcial 2-year economic cycle of 1% amplitude, the government consumption to output ratio is 1.30% higher than prior to the cycle, although for capital expenditure, the ratcheting coefficient is found to be insigniﬁcant.
The detection of ratcheting in these public expenditure components (relative to GDP) may provide justification for investigating the ratchet effect in compositional shares of government expenditure, rather than as shares of GDP alone. Unfortunately, it is difficult to find any studies (if any exist at all) which have embarked on such analysis. But if this effect is indeed detected, it may mean that some of the variation in compositional shares may be due to past fiscal responses to economic shocks that have not been corrected. However, technical econometric inquiry into the ratchet hypothesis with regards to compositional shares is beyond the scope of this paper.
Endogenous Growth Models
As Marica (2014:7) concurs, although the problem of optimal aggregate public expenditure has been largely debated in literature especially from a neoclassical perspective, only a few studies have addressed the optimal composition of public expenditure. Even the empirical literature on effects of aggregate public expenditure on economic growth has seen mixed and inconclusive results. Some find effect to be negative or non-significant (Folster and Henrekson, 1999; Grier and Tullock, 1999; Barro, 1990, 1991; Landau, 1983; Aschauer, 1989; Romer, 1990; Easterly and Rebelo, 1993). Others find that the impact of government is positive and significant (Sàez and García, 2005; Aschauer, 1989). Folster and Henrekson (1999:338), however, contend that the econometric problems arising when examining this relationship are not considered by many empirical studies. Providing evidence, they demonstrate that, after dealing with several econometric concerns, the relationship between government spending and growth may be more robust than it first appears. Likewise, larger emphasis on the econometric issues in most studies is placed by Slemrod (1995:401) when he states that largely due to fundamental identification problems, persuasive evidence for or against the relationship is lacking.
As Marica (2014:36) explains, it is only recently that researchers began to assess the effect of the composition of government expenditure on economic growth. Scholars have therefore tried to estimate the elasticity of economic growth with respect to components of government spending. These empirical attempts have largely been guided by endogenous growth models. It was indeed Barro’s (1990) sparking publication, as Chen (2005:123) notes, which precipitated growth effects of public spending to be one of the popular topics in economic research
Among others, pertinent papers in the area are Cassou and Lansing (1998), Turnovsky and Fischer (1995), and Futagami et al (1993). The theoretical connection between economic growth and government spending composition and growth is explored by Devarajan et al. (1996), Ghosh and Roy (2004), Chen (2006) and Ghosh and Gregoriou (2008). Devarajan et al. (1996) is a more recent and notable model focusing on the effect of compositions of public expenditure on growth. They developed a model considering two productive services in a CES production function. Of these two components of public expenditure, one is more productive than the other is. A shift in favour of an objectively more productive type of expenditure may not raise the growth rate if its initial share is too high. The difference between these two public spending types are expressed in Devarajan et al.’s model by highlighting how the long-run growth rate of the economy is altered by a shift in the mix between the two. In other words, not only are the conditions dependent upon the physical productivity of different components of public spending (in this context, current and capital spending) but also on the shares of government expenditure allocated to them. It is this model linking current and capital spending shares that forms the theoretical support for the importance of compositional shares of government spending (and likewise research into it), as well as justification for employing growth-maximization as the condition for optimality of shares of public current and capital spending in total public spending.
The model has been extended by Ghosh and Gregoriou (2008) within an optimal fiscal policy perspective. Additionally, Agénor (2010:945) demonstrates that a reallocation of public expenditure from “unproductive” public spending to infrastructure component would lead to a higher steady-state growth.
Ghosh and Gregoriou’s (2006) extension of the Devarajan et al. (1996) model is characterized by a decentralized economy set-up. They characterize the welfare-maximizing fiscal policy for a benevolent government. This government chooses the fiscal instruments at its disposal to maximize the representative agent's utility. Ghosh and Gregoriou’s (2006) model solves the problems relating to the optimal growth rate, the optimal tax rate, and the optimal expenditure shares of the two services. In Devarajan et al.’s (1996) model, it is in terms of the expenditure shares and tax rate (both exogenous) that the economy's growth rate is expressed; in Ghosh and Gregorious’s (2006) model, the optimal growth rate is expressed in terms of optimal values of the same two variables (Marica, 2014:39).
The Devarajan et al (1996:316) model starts from a CES production function of the form
[illustration not visible in this excerpt] (1)
Where 𝑦 is output, 𝑘 is private capital, 𝑔1, 𝑔2 are two types of government spending, and
illustration not visible in this excerpt
The budget constraint of the government is
[illustration not visible in this excerpt] (2)
Where 𝜏 is the (constant over time) income tax rate.
The shares of government expenditure that go towards 𝑔[illustration not visible in this excerpt] and 𝑔[illustration not visible in this excerpt] are given by
illustration not visible in this excerpt
Where [illustration not visible in this excerpt]
The representative agent's utility function is isoelastic, utility is derived from private consumption (𝑐) and is given by
[illustration not visible in this excerpt] (4)
Where [illustration not visible in this excerpt] is the rate of time preference
The agent's budget constraints is
[illustration not visible in this excerpt] (6)
An expression for the ratio is derived by Devarajan et al. (1996:317)
[illustration not visible in this excerpt] (7)
And 𝜆, the endogenous growth rate of the economy is given by
[illustration not visible in this excerpt] (8)
The objective of this model is to characterize the optimal fiscal policy. It is at this point that Ghosh and Gregoriou (2006) take over. Equations (1) - (6) are exactly as stated by Devarajan et al. (1996:316). The representative agent’s problem is to choose 𝑐 and 𝑘 to maximize utility, U in (4), subject to the budget constraint (5), taking 𝑔1 and 𝑔2, 𝑘0 and 𝜏, as given. The first order condition gives rise to the Euler equation (Ghosh and Gregoriou, 2006:6):
[illustration not visible in this excerpt] (9)
The government’s problem is choosing 𝜏, 𝑔1 and 𝑔2 for the maximization of the representative agent’s utility subject to (2), (5) and (8), and taking 𝑘0 as given. The first order conditions with respect to 𝜏, 𝑔1 and 𝑔2 respectively yield
[illustration not visible in this excerpt] (10)
[illustration not visible in this excerpt] (11)
[illustration not visible in this excerpt] (12)
Where µ and are the co-state variables associated with the private and government budget constraints – (6) and (2) – respectively.
From (11) and (12) is obtained, from which it can be derived the optimal ratio of the two public goods with a benevolent government:
[illustration not visible in this excerpt] (13)
From the following is obtained:
[illustration not visible in this excerpt] (14)
From obtained is:
[illustration not visible in this excerpt] (15)
The optimal tax rate is then
[illustration not visible in this excerpt] (16)
It is then finally possible to derive the optimal share of the first public good by combining equations (3), (14) and (16):
[illustration not visible in this excerpt] (17)
That for the second public good is likewise derived by combining equations (3), (15) and (16):
[illustration not visible in this excerpt] (18)
Combining (13), (17) and (18), the following equation is derived:
[illustration not visible in this excerpt] (19)
A clear implication of this equation (19), given the research context, is that given two countries with equal elasticity of substitution, the country with the higher relative productivity of current expenditure to capital expenditure (β/γ) will tend to have a resulting higher optimal ratio of current expenditure share to capital share. It is then possible to link the endogenous model to Haque’s (2004:16) aforementioned empirical observation (that is, that developed countries have higher current expenditure shares of total public expenditure, on average) by positing that perhaps, as countries develop the relative productivity of the share of current expenditure in total expenditure rises; hence the optimal share of current expenditure in total government expenditure rises as a consequence. However, more tangible econometric analysis is required to come to a more reliable conclusion.
Using this theoretical framework, Devarajan et al (1996) empirically investigate the compositional effects of public expenditure on growth for two separate panels of 21 developed countries and 43 developin. They find a result contrary to conventional wisdom: for the developing countries, a statistically significant positive growth effect was found for the share of current expenditure on total spending, a negative one was found for capital spending; for the panel of developing countries, the opposite relationship was found (Devarajan et al, 1996:338). Ghosh and Gregoriou (2008:5) independently confirm the result using GMM technique on a panel dataset for 15 developing countries over 28 years. On the other hand, Easterly and Rebelo (1993:431) find contradictory results. However, Devarajan et al. (1996:327) note that one reason that their results conflict with those of Easterly and Rebelo (1993) is that theirs is on the compositional effect of public expenditure on growth, with the level effect controlled for separately in the regression model by the variable total government expenditure as a proportion of GDP. Easterly and Rebelo (1993) do not control for this, and hence their results are less robust than Devarajan et al.’s.
Gupta et al. (2005) evaluate the impacts of expenditure composition and fiscal consolidation on economic growth in a 39 low-income country sample during the 1990s, estimated by GMM and LSDV estimators. They find that, in both the short-run and long-run, strong budgetary positions are generally associated with higher growth (Gupta et al, 2005:26). Evidence was also found that that in nations where higher shares of expenditure go to capital and nonwage goods and services, higher growth is experienced, while lower growth is experienced by those whose spending is concentrated on wages (Gupta et al, 2005:26).
For a panel of 30 developing countries from 1970 to 1980, Bose et al. (2007:550) investigate government expenditure effects on growth. The finding is that the share of current spending in GDP is insignificant, that of capital spending is significantly and positively associated with growth. Additionally, they find that the key sector to which government should direct its resources in order to promote economic growth is education.
Romero-Avila and Strauch (2008:187) examine data for 15 EU countries from 1960 to 2001. A positive and significant impact of government investment on growth and a negative and significant association is found between government consumption and transfers, on growth.
Armey (1995) presented the homonym curve, analogous to the Laffer curve, which shows the relationship between government size and economic growth as nonlinear in nature. It was indeed Armey (1995:3) who formulated the concept of optimal size of government. As Cosimo and Forte (2010:4) explain, if public expenditure is at very low levels, there would be failure of the government to ensure property rights protection and contract compliance, and economic growth would be zero or negative. Additionally, given that productivity of public enterprises grows less compared to the market, the more probable a decrease of total productivity would occur as the public share of resources increases. In contrast, when public expenditure shares are high, there would be little incentive for citizens to invest and produce, since the fiscal burden levels would be high, and growth suffers. Thus, there is an optimal value of public expenditure share of GDP from the point of view of the maximization of GDP growth.
* This research project was undertaken to in fulfilment of the research requirement for the award of Bachelor of Commerce (Honours) degree in economics at Rhodes University.