Fiscal Reforms, Long-term Growth and Income Inequality

We estimate the effects on growth of nine fiscal reform episodes in seven high-income countries using the Synthetic Control Method. These episodes are selected using an indicator-based approach applied to the evaluation of growth-friendly fiscal reforms during 1975-2010. We find that in reform countries the annual growth rate of real GDP was on average about 1 percentage point above their synthetic units 10 years after each respective reform. Moreover, countries which were initially less developed seemed to experience a larger growth impact after their reforms. Results are broadly robust to controlling for structural reforms on business regulation, financial market, labor market, and legal and product markets, which may also affect growth. Our findings also suggest that inequality is not affected by the growth-friendly fiscal reforms analyzed in this paper.

Abstract

We estimate the effects on growth of nine fiscal reform episodes in seven high-income countries using the Synthetic Control Method. These episodes are selected using an indicator-based approach applied to the evaluation of growth-friendly fiscal reforms during 1975-2010. We find that in reform countries the annual growth rate of real GDP was on average about 1 percentage point above their synthetic units 10 years after each respective reform. Moreover, countries which were initially less developed seemed to experience a larger growth impact after their reforms. Results are broadly robust to controlling for structural reforms on business regulation, financial market, labor market, and legal and product markets, which may also affect growth. Our findings also suggest that inequality is not affected by the growth-friendly fiscal reforms analyzed in this paper.

I. Introduction

Raising long-term growth is a priority for the global economy to improve living standards while reducing poverty. Moreover, Berg and Ostry (2011) document that sustained growth is robustly associated with more equal income distribution. In this context, fine-tuning which specific government revenue and spending measures can achieve higher long-term growth, without compromising the sustainability of public finances, is particularly important at the current juncture. In fact, IMF (2016a) lists lifting long-term growth and making it more inclusive as one of the key priority areas within a broad-based policy effort to reinvigorate growth and contain the risks of reform reversals in the short and longer run.

In this regard, IMF (2015a) studies the effects of fiscal policy on growth using a battery of tools, and concludes that a well-designed package of tax and expenditure policies can be effective in raising long-term growth. IMF (2014, 2016b) also examines the role of fiscal policy, but focusing instead on its role in affecting inputs in the production function, which could have an additional growth impact through factor accumulation and increases in total factor productivity.

Without inclusion, however, growth is fragile and may not be long-lasting. Kumhof, Ranciere, and Winant (2015), for instance, have shown that excessive household leverage can be driven by changes in income distribution, and that this high indebtedness level could lead to financial crises. In fact, several years of widening disparities in income distribution were at the forerun to the most recent global financial crisis, which was in turn triggered by disruptions in a highly-leveraged mortgage market. Moreover, owing to the lackluster growth that followed the global crisis and the large disparities in income distribution still present in many economies, the world is witnessing sudden changes in established social contracts as reflected in the ongoing movements towards more nationalistic political regimes around the globe.

However, establishing the causal relationship among fiscal policy, growth, and inequality, is a complex exercise. First, the causation is likely to be bi-directional, as shown in Muinelo-Gallo and Roca-Sagalés (2013), and Ostry, Berg, and Tsangarides (2014). While changes in fiscal policies affect growth and income distribution, fiscal policies can also be affected by changes in the state of the economy and in how income is distributed. Second, there are factors other than fiscal policy that likely influence growth. For instance, growth-promoting structural reforms are likely to happen at the same time fiscal reforms take place, which creates a difficult identification problem. In this context, any observed change in economic activity cannot be easily attributed to fiscal policy decisions. Third, the growth effects of fiscal policy could be country-specific. For example, the impact of tax policy changes is likely to be different in countries with high levels of taxation relative to those where the tax burden is relatively low, as shown in Trabandt and Uhlig (2011).

In this paper, we aim to shed light on the causal relationship from fiscal policy to growth from a country-case study perspective. To do this, we use the Synthetic Control Method (SCM), developed in Abadie and Gardeazabal (2003), and Abadie, Diamond, and Hainmueller (2010). The SCM provides a data-driven procedure to construct a suitable non-treated counterfactual (i.e., the synthetic unit). A key advantage of the SCM is that it helps cope with the likely heterogeneous effects of fiscal reforms across countries by analyzing an individual economy separately, an advantage over standard panel regression methods where an average estimate is instead obtained for a whole sample of countries. It can also address endogeneity issues arising from omitted-variable biases due to time-variant fixed effects.

The SCM approach was recently used in a related policy paper which studies the fiscal policy and growth nexus (IMF, 2015a).2 As in that paper, we develop an indicator-based methodology to select the countries that undertook growth-friendly fiscal reforms during 1975-2010. However, we introduce several changes to enhance the quality of the estimations, while examining new dimensions. First, we revisit the criteria by which countries experiencing structural fiscal reforms are selected. While IMF (2015a) considers both quantitative and qualitative elements of fiscal policy and discretionary judgments to select the control group, we focus instead only on quantitative criteria to identify reform episodes and to select countries. Because the use of qualitative elements is subjective and to some extent debatable, the sole focus on quantitative indicators enhances transparency and is more consistent with the nature of the SCM, which advocates for reducing the discretion in the design of the study to avoid selection biases. Second, we restrict our sample of candidates to evaluate their fiscal reforms to high- and upper-middle-income countries, instead of a set of countries with a wide range of income levels. This enhances the quality of the synthetic unit given that we are choosing it from a pool of countries with a relatively homogeneous economic structure. This also allows us to use our quantitative criteria with more confidence, given that these countries tend to have more consistent data and over longer horizons. To identify more accurately the impact of structural fiscal policies on the economy, we also address explicitly the role of non-fiscal structural reforms (namely changes in business regulation, financial market, labor market, and legal and product markets), which may happen simultaneously with those in the fiscal front and could also be affecting growth. Finally, we also include income inequality as another dimension to be evaluated besides growth, and document how developments of different inequality indicators are affected by the identified fiscal reforms.

Turning to our main results, we show that our data-driven approach led us to select nine fiscal reform episodes in seven countries, which generally had positive effects on long-term growth. In particular, the average annual GDP growth rate 10 years after the fiscal reform started was on average higher in the fiscal reform country relative to its synthetic unit by about 1 percentage point, ranging from 0.1 percentage points for the German reform that started in 2003 to 4.3 percentage points for the Chilean reform that started in 1983.

We also observe that reform episodes in countries with a relatively lower initial level of development were associated with higher growth after the fiscal reform episode started. Placebo experiments confirm the robustness of baseline results for four out of the nine cases. They tend to be the ones whose growth effect in the baseline is larger. Results are borderline robust in three cases, but not robust in the two remaining cases. Although these placebo experiments suggest that caution should be taken when evaluating results, still two-thirds of the events are either robust or borderline robust to this assessment, a fact that provides support to our baseline findings.

Importantly, our findings hold after controlling for non-fiscal structural reforms that took place prior to the fiscal reforms. Thus, the growth effects of the fiscal reforms remain even when we account for pre-fiscal reforms differences in non-fiscal areas. This does not imply that higher growth can be unambiguously attributed to any of the identified reforms, as many other structural reforms could take place after the fiscal reforms started. However, we did not obtain definitive evidence that non-fiscal structural reforms were stronger in the reformer countries relative to their synthetic units after the fiscal reforms started.

Finally, although there is a potential tradeoff between growth and inequality, because some reforms that increase efficiency may have adverse consequences in terms of income distribution (IMF, 2015a), we did not observe such a tradeoff in the fiscal reform episodes analyzed in this paper. More precisely, we did not find clear-cut evidence that fiscal reform countries in our sample had different inequality outcome relative to a global trend or their synthetic units.

The rest of the paper proceeds as follows. Section II describes the data, including a description of the country selection process and the SCM approach. Section III shows the main results, and the extent to which these are robust to the use of placebo experiments. Section IV provides additional robustness checks, including through the control of various structural reform indicators. In that section, we also assess the implications on inequality. Finally, Section V concludes.

II. Country Selection and the Synthetic Control Method

A. Selection of Reform Countries

The selection of country cases is in line with IMF (2015a), which uses a set of indicators identified as growth-enhancing in the academic literature and the IMF’s extensive technical assistance experience, to identify significant and long-lasting fiscal reform episodes. However, we depart from the approach of that paper in several ways. For instance, we consider only quantitative indicators rather than a combination of quantitative and qualitative criteria, to provide a more objective basis for the country selection process and to avoid selection biases. We also select reform episodes from high- and upper-middle-income countries. This ensures a more homogeneous economic structure among the countries identified as being subject to structural fiscal reforms (treated unit), and those that qualify as not being exposed to such reforms (control group), improving the accuracy of our estimates. Moreover, looking only at countries with relatively high income levels allows us to rely solely on quantitative indicators, as sufficiently-long historical and consistent data exist only for these economies.

There are in total nine indicators considered here. These indicators are defined by considering the direction in which these should be changing to ensure a positive growth impact, as explained below. Indicators cover the following three fiscal reforms areas: the tax mix, the composition of public spending, and the overall fiscal balance (Table 1).

  • The tax mix captures whether a country has re-balanced its taxes toward a more growth-friendly tax structure through a widening of the tax base, an increase (decrease) of standard rates, or a combination of tax base and rate changes. However, only shifts from direct to indirect taxes are considered as having a growth-promoting impact, in line with recent empirical studies (e.g., Arnold and others, 2011).

  • The composition of public spending evaluates whether a country has increased those outlays that are more likely to foster growth, such as physical investment, education, health and social protection (see Acosta-Ormaechea and Morozumi, 2017; Gosh and Gregoriou, 2008; Glomm and Kaganovich, 2008; and Gupta and others, 2005 for discussion).

  • Furthermore, changes in the overall fiscal balance are also included, to provide an indication of the impact of fiscal policy on macroeconomic stability, since fiscal reforms can be growth promoting only in the context of sustainable public finances.

Table 1.

Country-Case Selection Indicators

article image
Source: Author’s definitions

Direct tax denote taxes on income, profits and capital gains, social security contributions, recurrent taxes on immovable property, recurrent taxes on net wealth and other taxes on property; Indirect taxes denote taxes on payroll and workforce, taxes on goods and services and others.

Although increases in VAT standard rates might not be per se growth enhancing, most countries that undertook such reform reduced the corporate and/or the personal income tax rates around the same time. Such combination of reforms is likely to have growth promoting effects. If VAT rate increases were removed from the selection criteria, three additional European countries and two additional Western Hemisphere countries should be chosen in addition to our selected countries.

However, to reduce the risk of selecting outliers, countries reporting less than 20 years of data for a relevant fiscal indicator were excluded.3 In addition, we consider five-year non-overlapping averages to correct for the effects of business cycles fluctuations, starting in 1975 to 2010 or the latest available observation. Since reforms that promote long-term growth need to be sustained over time, the durability of the reform effort is controlled for by considering the change in the relevant fiscal variable over at least two consecutive five-year periods.

Reform countries are first selected by the number of reform indicators they satisfy. Thus, a country can satisfy at most nine reform indicators. Then we consider the regional breakdown of the sample. Specifically, following the IMF regional country classification, we select only those countries with the largest number of reforms in each of the following three country regions: European countries (EUR); Asian-Pacific countries (APD) and Western-Hemisphere countries (WHD).4 This selection process gives the seven countries considered here: Australia, New Zealand, Belgium, Ireland, Netherlands, Germany, and Chile (Table 2).5 Once countries are identified, we then define the exact year of the reform. This is done by reviewing the literature on the reform history of each country to define the year when a reform action started. Five of the seven countries overlap with (IMF, 2015a), and the reform year follows the identification of that paper. For New Zealand, we set the reform year as 1986, following OECD (1999) and Dalziel (2002). For Belgium, we set the reform year to 1992, following Carey (2003), IMF (1999), and IMF (2011).

Table 2.

Countries that Satisfy the Quantitative Criteria

article image
Note: (*) Selected countries.Source: Author’s calculations.

Includes only OECD countries with at least five reforms episodes.

World Bank’s country classification by income is included, as this classification will be required to construct the comparator countries as explained below.

Although seven countries are chosen, some of them could be exposed to more than one reform event, as observed in the cases of Australia and Chile.6

B. Selection of Comparator Countries

As part of the SCM (explained below), we also undertook the selection of countries in the control group to construct the so-called “synthetic unit” against which the effects of the fiscal reforms are assessed. Although the group of selected reform episodes includes only high-income OECD countries, the sample of those countries in the control group should necessarily go beyond the OECD to have a “sufficiently large” comparator base. Otherwise, the construction of a proper synthetic unit becomes difficult (see below). At the same time, we would like to retain the homogeneity of the control group to the extent possible. Thus, we consider high and upper middle income countries according to the World Bank’s country group classification in this broader sample, excluding oil exporting countries and small states.7,8 Table AI in Appendix 1 presents the full list of countries belonging to EUR, APD or WHD evaluated here, as well as their region and the number of reforms undertaken by each of them. It is important that the control group includes only those countries that did not undertake substantial fiscal reforms. In this regard, only countries which qualify as having five reforms or less according to the previous criteria are included in the control group.9

Once the comparator countries are identified, we group them in two types, following Billmeier and Nannicini (2013). Type A includes all countries that satisfy the above criterion within the whole pool of comparator countries. Type B includes only the subset that is in the same region as the country under consideration. After running the SCM for both types, we chose the type that provides the better pre-event match as the baseline case, where the better match is measured by the lower root mean squared error (RMSE) of the outcome variable (real GDP at constant 2005 prices (in million 2005 US$) in our case). Specifically, type A was chosen for Australia, Germany, and New Zealand, whereas type B was chosen for Belgium, Chile, Ireland, and the Netherlands. Appendix 2 provides the full list of countries that are available as part of the comparator group for each of the seven countries.10

C. Synthetic Control Method

Brief description

The synthetic control method provides a data-driven methodology to quantify the effects of a particular event in comparative case studies. It creates an artificial counterfactual (or synthetic unit) that closely matches the economic characteristics of the unit of interest prior to the event, and compares the difference in outcomes between that unit of interest and the counterfactual after the event. The synthetic unit is interpreted as the potential outcome of the treated unit if it did not experience such event, as the treated unit and its synthetic unit are matched in both observable and unobservable predictors. Thus, the divergence of the outcome variable after the event is interpreted as the quantitative estimate of the effects of the event under consideration during a particular number of periods. The method was first introduced in Abadie and Gardeazabal (2003), and has gained popularity in recent years. Notable applications include Abadie and others (2010, 2014), Billmeier and Nannicini (2013), and Cavallo and others (2013). The method can capture the heterogeneity in the effects of the event as it obtains case-specific estimates, an advantage over standard panel regressions whereby an average estimate is obtained for the whole sample. It can also address endogeneity issues due to omitted-variable biases arising from time-variant fixed effect. However, it would still suffer from reverse causation if the decision to embark on fiscal reforms was affected by expectations on future growth prospects. Moreover, its estimate can still be biased by events that take place after the event, which may affect the unit of interest and its counterfactual differently. For example, Cavallo and others (2013) report that their estimate of the effect of catastrophic natural disasters on economic growth is biased by the radical political revolutions that followed the natural disaster only in the country of interest.

More precisely, the method first uses data prior to the event to create a counterfactual unit as a weighted average of the comparator units, using a nested optimization algorithm that minimizes the distance between the unit of interest and its counterfactual, in terms of both the outcome variable of interest and its predictors. The comparator units are chosen so that they are similar to the unit of interest but are unaffected by the event under consideration.12 Once the counterfactual is created, the post-event outcome of the unit of interest is compared to the developments of such counterfactual. The estimated impact of the event is then represented by the difference between the two series of outcome variables over a specific period of time.

Implementation steps

Step 1: Choose potential comparator countries and explanatory variables
  • Comparator countries are those that are as similar as possible to the country of interest, but did not experience the same event within the sample period.

  • Predictor variables are those that are good predictors of the outcome variable of interest (real GDP at constant 2005 prices (in million 2005 US$) in our case).13

As elaborated above, we use two types of comparator countries. As predictor variables in the initial selection of countries we use standard variables chosen in empirical growth regressions, namely: the level of GDP per capita (at the beginning, half-way, and the end of the pre-reform period), trade openness, inflation rate, terms of trade and an index of human capital (see, Barro and Sala-i-Martin, 2004).

Step 2: Given the group of comparator countries and the outcome and predictor variables, we construct the relevant synthetic series as follows.14 The procedure calculates the weights of the comparator countries and predictor variables to create the counterfactual that is as close as possible to the unit of interest in the pre-event period. The method is based on a nested optimization algorithm as describe below.

Starting from an initial value of variable weights for the K×K diagonal matrix V, we choose the J×l vector W* of country weights that minimizes the distance ‖X1X0W‖, where X1 is a K×1 vector of pre-event averages of the K predictor variables for the unit of interest and X0 is a K×J matrix of the pre-event averages of the K predictor variables for the J comparator units, respectively, subject to the constraints that the weights must be between zero and one.15 In particular, W* minimizes the following distance in the pre-event period:

W*=argminWw=X1X0Wv=(X1X0W)V(X1X0W)(1)

Once the optimal country weights W* are chosen, the variable weight matrix V* is also chosen among all positive definite and diagonal matrices, such that the mean square prediction error (MSPE) of the outcome variable is minimized over the pre-event period. Specifically, this process considers:

V*=argminVv(Z1Z0W*(V))(Z1Z0W*(V))(2)

Where Z1 is a Tp×l vector of the time series of the outcome variable for the unit of interest, where Tp is the number of pre-event periods, and Z0 is a Tp×J matrix where each column is the time series of the outcome variable for country;. The resulting matrix V* is used as input in (1) for the next round of optimization to update W* (see Abadie and others, 2011, for details). Using such weights, the synthetic unit to create a counterfactual path of the outcome variable post-event can be constructed.

Step 3: Comparing the actual and post-event outcome variables. As indicated previously, the difference between the two series can then be interpreted as the estimated impact of the event under consideration (assuming that all other factors potentially affecting the variable of interest have been properly controlled for).

III. Baseline Results

Fiscal reforms have led to higher growth relative to the synthetic unit, in all reform episodes but with heterogeneity in the growth impact (Figure 1). In particular, average annual GDP growth for the 10 years after the fiscal reform started was higher in the reform country relative to its synthetic units by about 1 percentage point—ranging from 0.1 percentage point in 2003 for Germany to 4.3 percentage points in 1983 for Chile.16 The speed of the materialization of the growth effect also varied significantly. While growth accelerated quickly in Chile and Ireland, the growth effect remained more modest during the first few years in other cases (year-by-year results are summarized in Appendix 3).17

Figure 1.
Figure 1.

Growth Effects of Fiscal Structural Reforms: Baseline

Citation: IMF Working Papers 2017, 145; 10.5089/9781484303689.001.A001

Source: Author’s calculations.Note: Growth effects are defined as differences in average annual real GDP growth rates of the treated unit and its synthetic unit 10 years after the fiscal reforms started.

In terms of the size of the effects, we found that the episodes which took place in countries with a relatively lower level of development at the time of the fiscal reforms were associated with a higher growth effect afterwards (for instance, Chile 1983 and Ireland 1987). To formalize this, Figure 2 associates the GDP per capita relative to that of the US in the year the reform started with the subsequent growth effect. A lower GDP per capita ratio is broadly related to a higher subsequent growth effect, along the lines of what exogenous growth models would predict in terms of catching-up effects (Barro and Sala-i-Martin, 2004). These results broadly confirm that fiscal reforms could have different effects depending on the stage of development. Still, our sample is relatively small and it is true that many other factors could be at play, as highlighted by the very different growth effects of 1974 Chile and 1983 Chile, although the stage of development was similar in both cases.

Figure 2.
Figure 2.

Pre-Reform GDP Per Capita Relative to the US and After-Reform Growth Effect

Citation: IMF Working Papers 2017, 145; 10.5089/9781484303689.001.A001

Source: Author’s calculations.

While the SCM does not allow for the use of standard inference techniques due to the lack of large sample properties, we use placebo experiments to assess the robustness of the baseline results, following Abadie and others (2010) (Appendix 5). The idea is to evaluate the likelihood that the growth estimate of the non-reform countries exceeds that of the reform country by running the SCM for each country in the control group and generating their “growth estimates” and comparing the resulting “growth estimates distribution” with the baseline growth estimate. When this likelihood (p-value) is high, the robustness of the baseline estimates can be put under question. About half of the baseline results are robust to placebo experiments, in the sense that the treated countries indeed show that the growth estimates after the reform episodes are not obtained by chance. This finding holds typically in those cases with the larger growth estimates under the baseline. In particular, very few placebo permutations are above the growth effects of the treated country ten years after the treatment, namely Belgium 1992 (1/4), Chile 1983 (0/8), Ireland 1987 (0/3), and Netherlands 1983 (0/6).18 In contrast, baseline estimates are not robust for Chile 1974 (4/7) and Germany 2003 (20/21), as most placebo permutations show growth effects above that of the treated country. Australia 1985 (4/20), Australia 1998 (4/18), and New Zealand 1986 (5/19) are intermediate cases where baseline results are borderline robust. Although these placebo permutations suggest that caution should be taken when evaluating results, still two-thirds of the events are either robust or borderline robust to such assessment.

In addition, there are certainly other factors that could affect long-term growth besides fiscal policy. Structural reforms in non-fiscal areas are among the most relevant candidates, which we try to control for in the next section.

IV. Non-Fiscal Structural Reforms and Links with Inequality

A. Non-Fiscal Structural Reforms

The role that structural reforms in other areas, such as product market, labor market, business regulations, legal environment, and financial reforms play on long-term growth, have been well researched and documented in the literature.19 This implies that if the treated unit had more non-fiscal structural reforms than its synthetic unit in addition to the assessed fiscal reforms, the previous growth effects could be biased and overestimated.

However, the difference in non-fiscal structural reforms may have started prior to or concurrent with the analyzed fiscal reforms. To address the former case, we include non-fiscal structural reforms as additional growth predictors and re-run the SCM accordingly. This ensures that the non-fiscal structural reforms were similar between the treated unit and its synthetic counterpart prior to the fiscal reform events, thus controlling for the effects of non-fiscal structural reforms from the growth estimates. The list of the structural reforms considered are business regulations, financial reforms, legal reforms, labor market reforms, and product market reforms, following IMF (2008, 2015b). We add them one at a time for each episode.

We cannot do the same when the non-fiscal reforms take place simultaneously or immediately after the fiscal reform episode, however. We instead document the developments of non-fiscal structural reforms after the fiscal reforms started, and test whether the difference in the strength of structural reforms between the treated unit and its synthetic unit is statistically significant.

Non-fiscal structural reforms prior to fiscal reforms

As explained, to control for the effects of differences in structural reforms prior to the fiscal reforms, we add a series of structural reform events as input to the SCM one-by-one each time, as an additional growth predictor.

Our baseline estimates of the growth effects are generally confirmed. Dots in Figure 3 represent the point estimates of the SCM results with the inclusion of different structural reform indicators, and pink bars take their average, weighted by the inverse of the RMSE, as similarly considered in Acemoglu and others (2016). Measured by the weighted average, these findings show positive growth effects which are broadly similar in magnitude to the baseline results. An exception is New Zealand 1986, for which the growth effect turns negative.20 On average, output growth was higher by 0.9 percentage points relative to the control units, ranging from −0.7 percentage point in New Zealand 1983 to 4.3 percentage points in Chile 1983. Some of the point estimates in Australia 1985, Australia 1998, and New Zealand 1986 do show negative growth effects, however.

Figure 3.
Figure 3.

Growth Effects: Baseline and Controlling for Non-Fiscal Structural Reforms

Citation: IMF Working Papers 2017, 145; 10.5089/9781484303689.001.A001

Source: Author’s calculations.Note: Blue bars indicate the baseline results (B), red bars non-fiscal structural reforms results (SR).

Non-fiscal structural reforms concurrent with fiscal reforms

Part of the estimated growth effects may well reflect the impact of structural reforms that took place simultaneously or immediately after the analyzed fiscal reforms, which may not be properly captured through the previous approach. In particular, if the treated country engaged in a higher degree of structural reforms relative to its control group around the time of the fiscal reforms, then the SCM results would not correct for such effects and the overall results might be biased. To test for this possibility, we analyze the evolution of indices of non-fiscal structural reforms between the fiscal reform country and its synthetic units after the analyzed fiscal reforms started. Two types of exercises are undertaken.

First, we document ten-year changes in non-fiscal structural reform indicators and compare whether the treated or its synthetic unit had more of these reforms during the fiscal reform period. If the treated and synthetic units show similar developments, it indicates that the estimated growth effect is likely to be due to the fiscal reforms. In contrast, if the treated unit exhibits larger improvements in these non-fiscal structural reform indicators, we cannot rule out the possibility that the non-fiscal structural reforms contributed to the estimated growth effect. We use the weights derived in the SCM that included the pertinent reform as an additional growth predictor to construct the synthetic unit and the corresponding non-fiscal structural reform indicators. In other words, the weights differ depending on the reform. Second, we examine whether the treated unit also exhibits a stronger impact from non-fiscal structural reforms than its placebos during the fiscal reform periods. For this purpose, we construct “non-fiscal structural reform gaps” for each of the SCM permutation in the placebo experiments. The “non-fiscal structural reform gaps” are measured by differences in ten-year changes between the treated and synthetic unit in non-fiscal structural reform indicators. If the “non-fiscal structural reform gap” has high p-values, it suggests that the treated unit did not have non-fiscal structural reforms that were particularly stronger than its synthetic unit during the fiscal reform periods. We consider only the six cases for which the baseline growth effect was either robust or borderline robust in the placebo permutations. Moreover, we focus on business regulation and legal environment reforms due to data limitations.

Overall, results are not definitive regarding whether fiscal reform countries also had more non-fiscal structural reforms than their synthetic units concurrent with the fiscal reform episodes. Table 3 summarizes results for the first exercise. (More detailed developments of structural reform indicators are presented in Appendix 6.) In general, the non-fiscal structural reforms were more pronounced in the treated units but it was not predominant, as indicated in the last row of Table 3. Regarding the second exercise, p-values for the non-fiscal structural reforms are often large, suggesting the subdued role of non-fiscal structural reforms in the reform countries (Table 4). The p-values were also larger than those for growth effect for almost all cases, highlighting the likelihood that fiscal reforms were the key contributing factor of the assessed growth effects (Table 4).

Table 3.

Summary Comparison of Non-Fiscal Structural Reforms

article image
Note: Figures reflect the difference in the average annual change in percent of the pertinent structural reform index, ten years after the start of the fiscal reform period, between the treated unit and its synthetic unit.

The numerator is the number of episodes for which the structural reform preceded more in the treated unit than its synthetic, while the denominator is the total number of episodes for which data was available. A positive difference for business regulations, financial and legal means the reform preceded more in the treated unit, while the opposite is true for labor and product market.

Source: EPW, OECD, and author’s calculations.
Table 4.

Growth Effect and “Non-Fiscal Structural Reform Gap”

article image

Clearly, there is a more general issue which is the magnitude of the growth effect caused by the non-fiscal structural reforms, which may also vary across countries (i.e., because of differences in the implementation of a particular structural reform, country A could have a larger or lower growth effects than country B, even though in principle the same structural reform is applied).21 Thus it still remains an open question to assess quantitatively how growth is affected by the non-fiscal structural reforms that take place simultaneously with the fiscal reforms analyzed here.

B. Inequality

As part of the structural fiscal reforms analyzed here, there is the possibility of a growth-inequality trade-off, whereby growth acceleration may come at the expense of increases in income inequality. One of our criteria for choosing fiscal reforms is the change in the tax mix from direct to indirect taxation. However, it is well documented that indirect taxes tend to be more regressive than direct taxes (e.g., Martinez-Vazquez, Moreno-Dodson, and Vulovic (2012)). Moreover, if improvements in the overall balance, another criterion chosen by us, come at the expense of a reduction in the transfers targeted to the relatively poorer segment of the population, such policies could also have negative effects on income distribution. Focusing on fiscal consolidation episodes, Ball and others (2013) and Woo and others (2013) show that fiscal consolidations typically raise inequality, especially when they are expenditure-based. At the same time, increases in health and education expenditure, two of our other criteria, may have the potential to reduce inequality (e.g., Martinez-Vazquez, Moreno-Dodson, and Vulovic (2012)).

Our approach in this subsection is to evaluate how fiscal reform countries fared on inequality relative to a global trend, and relative to their synthetic units. For instance, Immervoll and Richardson (2011) and Caminada, Goudswaard, and Wang (2012) report that between mid-1980s and mid-2000s, market-income inequality increased, being only partially offset by an accompanying increase in fiscal redistribution. We investigate whether the fiscal reform countries deviated from this global trend, and whether they had a different pattern relative to the synthetic units. In this regard, we intend to simply document the developments of inequality indicators using the country weights obtained in the baseline. Thus, we are not claiming any causal effects from fiscal reforms to inequality, but rather merely showing associations. We are also not making a welfare or normative evaluation of whether an increase in inequality is desirable or undesirable.

To proceed, we use Solt (2014)’s Standardized World Income Inequality Database (SWIID) and track developments in market income and net income Gini coefficients, which measure inequality of market-based income and that after fiscal redistribution effects (due to budget transfers and progressive taxation), respectively. Thus, an increase in inequality is represented by an increase in these coefficients. We focus on the seven cases where data are available, leading to the need to drop Chile 1974 and Netherlands 1983.

Income inequality relative to a global trend

On a first look, there is no deviation from a broader trend specific to countries which undertook fiscal reforms. While income inequality increased in most cases, fiscal redistribution still played an important role. Figure 4 documents the changes in Gini coefficients over time for all fiscal reform cases. It shows that both market income Gini and net income Gini increased in five out of the seven cases considering ten-year averages after the fiscal reform started (except for Chile 1983 and Ireland 1987).22 At the same time, the net income Gini coefficient either increased less or decreased more than the market income Gini coefficient in most cases (except for Belgium 1992 and Ireland 1987), suggesting an active role of fiscal policy to redistribute resources in the economy. Thus, the general tendency is that an increase in market Gini is partially offset by fiscal redistribution during the reform episodes analyzed here.23

Figure 4.
Figure 4.

Changes in Market and Net Gini Coefficients in Reform Episodes

Citation: IMF Working Papers 2017, 145; 10.5089/9781484303689.001.A001

Source: Author’s calculations.

Income inequality relative to the Synthetic Units

We also compare the developments of the Gini coefficients between the treated and the synthetic units to measure whether there are discernible differences. The idea is again to identify the marginal effects of fiscal reforms. The Gini coefficients for the synthetic units are constructed using a weighted average with weights obtained from the baseline SCM results. The assumption is that the synthetic unit in the baseline is similar enough to the treated unit, in the sense that it can be used also as a synthetic unit for the inequality indicators.24

Results in Figure 5 suggest that market income Gini increased more in four out of seven reform cases relative to the synthetic units (i.e., Australia 1985, Australia 1998, Germany 2003, and New Zealand 1986). Fiscal redistribution also increased more in all four cases. For the cases in which market income Gini increased less than the synthetic unit, two of them (Belgium 1992 and Ireland 1987) also had the fiscal redistribution increasing less than the synthetic unit. Finally, in Chile 1983, fiscal redistribution increased more than the synthetic unit, although market Gini increased less. Thus, we do not observe a consistent pattern whereby fiscal reforms led to either higher market-income inequality, or that the fiscal redistribution function weakened more in countries which undertook the structural fiscal reforms analyzed in the paper.

Figure 5.
Figure 5.

Changes in Market Gini Coefficients and Fiscal Redistribution, Treated vs Synthetic

Citation: IMF Working Papers 2017, 145; 10.5089/9781484303689.001.A001

Source: Author’s calculations.

V. Concluding Remarks

We have shown that those countries identified as being exposed to growth-friendly fiscal reforms had experienced higher growth relative to their synthetic units. Although the size of the effects varies among reform episodes, it is broadly observed that those countries which were initially at a lower level of development experienced a larger growth impact. Results tend to be broadly robust to placebo permutations in most fiscal reform events. Also, as a key robustness check, we found that our main results remain broadly unaffected after controlling for various non-fiscal structural reforms prior to the fiscal reform events that could contribute to growth, namely business regulation, financial, labor market, legal and product market reforms. Moreover, we could not find definitive evidence that non-fiscal structural reforms were stronger in the countries that undertook the fiscal reform relative to their synthetic unit even after the reform period started.

From an inequality perspective, we could not find clear-cut evidence on whether those countries that went through the analyzed fiscal reforms experienced noticeable differences in their inequality indicators. That is, the possible tradeoff between growth-friendly fiscal reforms and inequality often discussed among policymakers, appears to be rather absent at the aggregate level in the evidence presented here. In fact, although in some cases the fiscal reform periods have coincided with increases in income inequality, there is no evidence pointing to any form of causality from structural fiscal reforms to a more unequal income distribution.

Although our findings provide convincing preliminary evidence about the long-term positive impact of structural fiscal reforms on growth, further analyses seem to be warranted to shed further light on the causal relation between fiscal policy and growth. More detailed considerations to help foster policy recommendations can also be considered. For instance, one relevant avenue to explore from a tax policy perspective could be to identify whether a broadening of tax bases (rather than changes in standard rates) has more desirable benefits from a growth perspective. Similarly, assessing the impact of structural expenditure measures, when controlling for the quality of the different spending categories, is a fruitful issue to be assessed, which can provide clear policy insights. To the extent that data availability permits, exploring those questions using the synthetic control method is an interesting and exciting area for future research.

Fiscal Reforms, Long-term Growth and Income Inequality
Author: Mr. Santiago Acosta Ormaechea, Mr. Takuji Komatsuzaki, and Carolina Correa-Caro
  • View in gallery

    Growth Effects of Fiscal Structural Reforms: Baseline

  • View in gallery

    Pre-Reform GDP Per Capita Relative to the US and After-Reform Growth Effect

  • View in gallery

    Growth Effects: Baseline and Controlling for Non-Fiscal Structural Reforms

  • View in gallery

    Changes in Market and Net Gini Coefficients in Reform Episodes

  • View in gallery

    Changes in Market Gini Coefficients and Fiscal Redistribution, Treated vs Synthetic