The Treasury

Global Navigation

Personal tools


Government and economic growth: Does size matter?

3.2  Evidence of causation: regression analysis

To move beyond correlation to identify whether there is a casual link between levels of government expenditure and economic growth requires cross-country economic growth regressions. These regressions attempt to take account of the range of variables or factors that may affect economic growth. The variables can include initial levels of ‘state' variables such as the stock of physical and human capital (in the forms of educational attainment and health). They also typically include ‘control or environmental' variables, such as the ratio of government consumption to GDP, share of domestic investment in GDP, the extent of international openness, movements in the terms of trade, the fertility rate, indicators of macroeconomic stability and measures of the maintenance of the rule of law (see Barro Sala-I-Martin, 2003).

There are challenges in identifying the appropriate variables to include, and in finding good measures of these variables, as economic growth theories are not conclusive about what variables are most important for economic growth (Sala-I-Martin, 1997a and b). There are also a number of other methodological challenges. The inter-relationship between expenditure and economic growth creates particular methodological problems (such as endogeneity). There are also data issues due to countries using different conventions for defining the public sector and testing problems created by a lack of variation in size of governments and differences in the efficiency of expenditure across OECD countries (see Barro and Sala-I-Martin, 2003 for more detail on the methodological issues). The 2025 Taskforce highlights that these methodological problems mean the results of cross-country regressions should be read as illustrative rather than determinative.

3.2.1  The mix matters

Partly reflecting these challenges, early econometric studies initially struggled to find a robust relationship between the level of government expenditure or revenue and economic growth. A number of reviews of the empirical evidence found that the link between government expenditure or revenue and economic growth is ambiguous, with similar fiscal variables producing strong positive, negative or insignificant results (see Nijkamp and Poot, 2004, Myles, 2000; Agell et al, 1997, and Temple, 1999). To try to overcome the problem that the results were sensitive to different model specifications, Sala-I-Martin (1997a and b) ran a large number of regressions with different specifications but still found that no measure of government spending appears to affect growth in a significant way.[5] More recent work by Barro and Sala-I-Martin (2003) found that government size did have a negative impact on economic growth in simple models with a relatively small number of variables. However, it became insignificant when additional variables were added to the model, suggesting that it was capturing the impact of these other variables in the simple models.[6]

However, there is now growing recognition that the lack of robustness of earlier studies was because they failed to take account of the composition of expenditure and how it was financed. The widely differing results across studies may reflect the net effect of alternative methods of financing expenditure or differences in expenditure mix across countries. By distinguishing between different types of expenditure and taxes, Kneller et al (1999a) and Bleaney et al (2001) found a positive impact of growth-enhancing expenditure and a negative impact of distortionary taxation on economic performance. Also in line with endogenous growth theory, they found that non-productive government expenditure and less distortionary taxes have no persistent, statistically significant effect on growth.

These studies have been reinforced by other studies by the same authors (such as Gemmell et al, 2001, 2009a and 2011), as well as by other researchers (see Box 2 for some examples and Gemmell et al, 2009a for a summary of relevant studies). As well as these “top-down” studies, there are a range of “bottom-up” studies that show a negative impact of taxes on factors critical to economic growth, such as investment and the size of the labour force (see Myles, 2007 for a review).

Box 2: Empirical studies of composition

Studies that have taken account of the composition of both expenditure and taxation have tended to find a positive impact of growth-enhancing expenditure and a negative impact of distortionary taxation on economic performance. Some examples are:

  • In a panel study of 23 OECD countries between 1970 and 2000, Angelopoulos et al (2007) found that OECD countries could improve their growth performance by reallocating public spending towards productive activities and that the average tax rate is negatively correlated with growth.
  • Bassanini et al (2001) found a large and significant negative effect of total government revenues on growth, with an additional negative effect coming from a tax structure focusing on direct (more distortionary) taxes, but positive impacts from consumption and investment expenditure.
  • Kocherlakota and Yi (1997) also found no impact of fiscal policy on long-run growth when including only taxation or expenditure in their model. However, distinguishing between taxation and public capital spending showed that they both had a statistically significant impact on long-run economic growth.
  • Barro and Redlick (2010) use a newly constructed series of average marginal income tax rates, which is a better measure of the disincentive effects of taxation than the traditionally used average tax measures. To overcome problems of the inter-relationship between government expenditure and GDP (endogeneity) they focus on increases in US defence expenditures associated with wars, which are unlikely to be driven by increases in GDP. They found that the simulative impact of increases in defence spending was offset by the economic costs of raising taxes to finance it, with increases in defence expenditure crowding out investment and net exports.

The studies that take account of compositional effects find that the greatest boost to economic growth comes from financing a reduction in the most distortionary types of tax by reducing unproductive expenditure. Some studies also suggest that the gains are still positive if financed by reductions in growth-enhancing or productive expenditure (for example, Kneller et al, 1999 and Gemmell et al, 2011). This suggests that, on average, the benefits of growth-enhancing expenditure in OECD countries may be somewhat outweighed by the costs of taxation. However, given the damaging impact of deficits or debt on economic growth, there is likely to be little or negative gain in financing tax cuts through increases in borrowing (Gemmell et al, 2011 and Romer and Romer, 2010).

There is significant diversity in estimates of the size of the economic growth boost even from reducing distortionary taxes and unproductive expenditure. For example, while Kneller et al (1999a) estimate that a one percentage point cut in distortionary taxation and unproductive expenditure could enable a 0.4 percentage point increase in annual economic growth, Gemmell et al (2011) suggest only a 0.1 percentage point increase in economic growth rates. While even the latter is significant, there is considerable uncertainty around these estimates and their applicability to New Zealand. In particular, it is risky to assume we can get at least a 1 percentage point increase in growth from a 10 percentage point reduction in expenditure as a percent of GDP. This would depend on the share of unproductive expenditure in our expenditure mix and the extent to which we can identify the right expenditures to reduce. It is also expected that there are diminishing returns from reductions in distortionary taxes and that, the more the state shrinks, the greater the chance that we begin to cut into expenditure that has some growth benefits.

Despite these caveats, overall it seems that the growth impacts of reducing distortionary taxes, particularly via reductions in ‘unproductive' expenditure, is likely to be moderate but not insignificant. Moreover, recent work by Gemmell et al (2011) suggests that the ‘long-run' growth effects of such changes are likely to be achieved within a few years of the policy change and are likely to persist, at least beyond a five-year horizon, as long as the changes are sustained. Unfortunately, growth-enhancing fiscal initiatives are often reversed. OECD governments have tended to finance increases in growth-enhancing expenditure with increases in growth-inhibiting taxes, or have increased expenditure again thereafter (Gemmell et al, 2011). This type of volatility in fiscal policy has its own dangers for economic growth. Theory and evidence suggests that uncertainty stemming from fiscal volatility can have a negative impact on capital formation and investment.[7] Unsustainable reductions in expenditure that are later reversed may be more harmful for economic growth than maintaining expenditure at a higher but consistent level.

3.2.2  What is the right mix?

Given that the mix matters, what types of taxes and expenditures should policymakers emphasize? Though there are some strong messages from theory and empirical evidence on the growth ranking of different types of taxes, distinguishing between ‘productive' and ‘unproductive' expenditures is more contentious.

A number of studies emphasise that it is the marginal rate of specific taxes, rather than overall tax to GDP ratios, that matter for growth (Treasury, 2008). In particular, a review of the evidence suggests that recurrent taxes on immovable property are the least distortive tax instrument in terms of reducing long-run GDP per capita, followed by consumption taxes (and other property taxes), with personal income taxes and corporate income taxes the most damaging for growth (Johansson, 2008). The Henry Tax Review (2010) in Australia also found the deadweight costs of corporate income tax (at 40 cents for every tax dollar collected) and labour income tax (at 24 cents) much higher than for consumption taxes (8 cents) and municipal rates (2 cents). Claus et al (2010) found high welfare costs from New Zealand's top marginal income tax rate of 39%,[8] while Gemmell et al's (2009b) open economy model suggests that bucking the OECD trend towards lower corporate tax rates could undermine a country's growth prospects.

On the expenditure side, empirical studies can be broken down by how expenditure is classified, typically using either an economic or a functional classification:

  • Using the economic classification, the results for public investment have been mixed. In a review of empirical studies, Gerson (1998) reports that only some studies found a positive link between total public investment and growth. More recent studies are also inconclusive. For example, Romero de Avila and Strauch (2003) estimate public investment to have a positive effect on growth in the EU, while Afonso and Furceri (2008) do not find public investment to be significant in explaining growth in the EU and OECD. By contrast, public transfers and consumption are typically estimated to negatively impact on growth (Barrios and Schaechter, 2008).
  • Using a functional classification, the types of public expenditure that have been found to raise growth vary strongly with the data sample. Some studies find only education, research and development and public infrastructure spending to be growth enhancing; others also include spending on health, public order and safety, and environment protection (Barrios and Schaechter, 2008). However, in a review of a large number of studies, Nijkamp and Poot (2004) highlighted that studies tend to find that infrastructure and education are positive, while general government consumption and defence are typically negative for economic growth.
  • Some studies combine economic and functional classifications. For example, in a review of empirical studies, Barrios and Schaechter (2008) found that investment in transportation and communication was more systematically matched with higher growth. However, in a review of studies using a variety of econometric approaches, Égert et al (2009) found that infrastructure investment in telecommunications and the electricity sectors has a robust positive impact on long-term growth but not in railways and road networks.

Therefore, empirical evidence is not conclusive about what expenditures are good for growth. Even expenditure areas that are generally modelled as non-growth enhancing are not uncontroversial. For example, the endogenous growth literature typically defines social welfare benefits as ‘unproductive' or providing no growth benefit to offset the cost of financing them. Other commentators would go further and argue that social benefits further undermine economic growth through undermining incentives to work and save. On the other hand, there may be offsetting or growth-enhancing benefits from having a safety net. For example, people may wait for jobs that better match their skill set, or be less risk averse and more entrepreneurial, or invest more in human capital (Kneller et al, 1999b).[9] In addition, though the aim of social benefits may be to reduce poverty or increase social mobility, they should also reduce inequality. There are mixed views on the relationship between inequality and growth (see Box 3). Whether it is desirable for government to redistribute wealth will depend upon what factors are driving any inter-relationship and the method by which government chooses to reduce it. Whatever the evidence on the impact of inequality on growth, if government wants to reduce inequality for other reasons, it is important that it does so in a way that minimises any possible disincentive effects.

Box 3: Is income inequality bad for growth?

Traditionally, economic models have suggested that income or wealth inequality has a positive effect on growth. Savings rates models hold that as the marginal propensity to save is higher for rich than for poor people, more unequal societies are likely to save and invest more (Kaldor, 1960; Kalecki, 1971). Work incentives models hold that output is increased where workers are incentivised - through wages relating work to output - to put in additional effort (Mirrlees, 1971).

However, in the early 1990s, numerous theoretical and empirical studies found evidence to suggest that inequality of income or wealth has a negative effect on growth. Political-economy/median voter models and empirical studies suggest that income or wealth inequality negatively affects growth as in unequal societies, a majority of people will vote for redistributive policies, undermining growth because of the economic costs of financing those policies (see Persson and Tabellini, 1994). In credit market imperfection models the combination of borrowing constraints and inequality limit the ability of the less wealthy to invest in human capital, which has important inter-generational impacts (for example, Galor and Zeira, 1993). Socio-political models suggest that income or wealth inequality discourages investment by encouraging non-market activities, such as crime, eroding trust and increasing socio-political instability (see Alesina and Perotti, 1996). These frameworks suggest the need to balance the negative costs of financing redistributive policies with the benefits of creating a socio-political climate more conducive to productive activities and capital accumulation. However, the type of policy response depends upon the cause of the income inequality - social benefits, for example, may not address low human capital in some parts of the population and may undermine economic growth.

Using new longitudinal datasets, several studies since the mid-1990s suggest that the effect of income or wealth inequality on growth is neither wholly negative nor wholly positive. For example, Barro (2000) finds that the effect depends on a country's stage of development, with income inequality slowing growth in poor countries but encouraging it in wealthier ones. Voitchovsky (2005) finds that the effect is sensitive to both the methodology used (with time series data tending to indicate a positive relationship and cross-sectional data indicating a negative one) as well as the size and shape of the income distribution (with inequality having a positive effect in the upper tail of the distribution but a negative effect in the lower tail).

A debate also continues on the impact of inequality on wider social outcomes. For example, in The Spirit Level (2009), epidemiologists Wilkinson and Pickett argue that increases in income inequality lead to increases in a range of social problems that affect people at all points across the income distribution. While gaining much prominence in the media, Wilkinson and Pickett's work has been criticised for using correlations to claim a causal relationship between inequality and social problems.

Overall, as Dominicis et al (2006) note in their meta-analysis of 22 key studies undertaken over the past 16 years, there is still no theoretical or empirical consensus on the impact of inequality on growth. The Treasury is closely following the evolution of research on this topic and, when more conclusive evidence emerges, will consider the implications for policy settings. In the meantime, we will continue to outline the distributional effects of policy choices in our advice to Ministers.

The mixed empirical results on how different expenditures impact on economic growth may perhaps reflect the empirical challenges. Gemmell et al (2009a) argue that much of the literature is methodologically weak as many studies do not include all elements of the government's budget, leading to biased results (see section 3.2.1). The more recent literature is beginning to produce some degree of consistency at least in the finding that infrastructure and educational spending tends to be growth-enhancing (Gemmell et al, 2009a). Another issue is that modelling expenditure at the level of functional and economic classifications may mask the significant diversity in the type and quality of different expenditure programmes within these classifications. Contradictory findings may also reflect the reality that more expenditure is not always better, even when it is growth-enhancing. The impact of ‘productive' expenditure on economic growth is not likely to be linear, as the marginal benefit of expenditure will decline as expenditure increases (see section 2.1.3). For example, while Égert et al (2009) found a positive impact of infrastructure investment on growth, the impact varies across countries, sectors and time, with some evidence of over-investment or inefficient use of infrastructure. The mixed empirical results suggest that it is well-targeted and high-quality public expenditure, rather than just expenditure in particular areas, that is growth enhancing.


  • [5]Sala-I-Martin (1997a and b) ran over two and four million regressions respectively and assigned a level of confidence to each variable, depending on the number of regressions where it shows a statistically significant relationship with economic growth.
  • [6]The researchers do caveat that this analysis would not pick up a non-linear relationship between economic growth and government ‘size' as predicted by the inverted U of endogenous growth theory.
  • [7]For example, Fatás and Mihov (2003) found that, for a set of 91 countries, higher volatility of discretionary government spending significantly increased output variability which, in turn, lowered growth. In a recent study, Afonso and Furceri (2008) show that even volatility in the cyclical component of public expenditure can undermine growth performance.
  • [8]The top marginal tax rate was increased from 33% to 39% in 2001 but was reduced again to 33% in 2010.
  • [9]Note that Gemmell et al (2009a) find that, once the inter-relationship between social spending and economic growth is controlled for (i.e. that social spending will tend to increase in economic downturns), social spending appears to be neutral, rather than adverse, for growth. However, neutral expenditure is still likely to have a negative impact on economic growth as it does not provide any positive impacts for growth to offset the economic cost of financing it.
Page top