The cyclicality of government foreign-aid expenditure: voter awareness in “good” times and in “bad”

Economists usually anticipate countercyclical government spending (Alesina et al. 2008), but evidence is mounting that governmental outlays are procyclical (Gavin et al. 1996; Kaminsky et al. 2004; Talvi and Végh 2005; Woo 2009). Public expenditure is procyclical if it increases as national income increases and declines as national income declines. While circumstances exist in which procyclical expenditure can maximize a community’s welfare (Lane 2003), such circumstances are very specific and extremely unlikely (Halland and Bleaney 2011).¹ The growing consensus is that systematic changes in political pressures over the economic cycle are determinants of procyclical public expenditure. Lane and Tornell (1996) and Tornell and Lane (1999) refer to “voracity effects”. When “voracity effects” operate, political pressures to increase public spending are more intense in economic upturns than in economic downturns.

In this paper, the first objective is to explore the question why are political pressures for increased public spending more intense in economic upturns? If income effects are positive, spending pressures are likely to strengthen as income increases, but when “voracity effects” operate additional factors must be considered. The proposition to be considered in Sect. 2 is that systematic changes in voter awareness (of fiscal policy) are relevant when explaining changes in the intensities of political pressures for more government spending.

Empirical studies have indicated that donor governments’ provision of foreign aid is procyclical. Dabla-Norris et al. (2015) report that foreign aid flows are influenced positively by the output gap between the donor country and recipient countries. Gravier-Rymaszewska (2012) explored the effects of macroeconomic shocks on the supply of foreign aid from donor countries, reporting evidence that aid budgets fall when own revenues fall. By comparison, in this paper, the second objective is to test the proposition that systematic changes in voter awareness over the economic cycle influence changes in the intensity of political pressure (for more foreign aid) and the cyclicality of that budget item.

In Sect. 2 the first question is whether it is possible to call on the public choice literature to predict systematic differences in voters’ awareness over time. Can we conclude that marked differences exist in voter awareness in “good times” and in “bad”?

Given the foregoing dichotomy, the intention herein is to focus on the impact of changes in voter awareness of public spending (1) in upturns and (2) in downturns. The cyclicality of expenditure over the full economic cycle will reflect different experiences when the economy is expanding or contracting. In that way, the paper is the first (to our knowledge) to employ a methodology that focuses separately on different cyclicalities (in upturns and in downturns) to explain the overall cyclicality of expenditure. The paper is also the first to consider the likelihood that vote-maximizing governments respond strategically to “fiscal illusion” in economic upturns. To date, the literature on “fiscal illusion” in economic upturns has focused on its impact on the level of aggregate public expenditure. Here, the proposition is that, as vote-maximizing governments benefit from a systematic reduction in voter awareness, they are likely strategically to time increases in potentially “unpopular” expenditures during economic upturns. Does evidence exist of that strategy (if governments are under pressure to accommodate lobby groups to increase public expenditure beyond the preference of the median voter)?²

Testable predictions are presented in Sect. 2 of the paper. Later sections present the data and empirical model employed to test the predictions. The intention is to test each of the predictions separately, but a further (and more demanding) question is whether the overall pattern of cyclicality (across budgets) is consistent with the pattern of cyclicality predicted in Sect. 2.

While changes in income are likely to be relevant when explaining changes in government expenditure, political pressures also are likely to influence changes in such spending.³ How is the intensity of political pressure for more public spending likely to change in economic upturns and in downturns?

To investigate the cyclical properties of foreign aid expenditure we estimate⁸:where fa_it is the natural log of foreign economic aid expenditure for donor country i in period t, observations on which are taken from the Government Expenditure by Function section of the General Government Accounts (OECD 2018a), part of the OECD National Accounts (OECD 2018b). Foreign economic aid is a component of the broader General Public Services function and is defined as the government’s total aid budget, which is allocated to all purposes and given to all recipients.⁹ It is a relatively small part of total expenditure, averaging only 0.6% of all government spending across the sample and 4.7% of the larger General Public Services function. The change in income, Δy_it, reflects the economic cycle and while the literature sometimes uses the first difference in real GDP, we use the more common output gap, figures for which come from the OECD’s Economic Outlook: Statistics and Projections (OECD 2018c). All series are collected for a sample of 23 donor countries¹⁰ over the period from 1999 to 2015 and expressed in constant prices using data on the GDP deflator (OECD 2018b). We enter lagged values of Δfa_it to account for potential persistence in the growth of foreign economic aid budgets; Δy_it−1 and Δy_it−2 also are entered since the cyclical relationship between GDP and government spending on foreign economic aid might not be contemporaneous, e.g., owing to information lags. Our focus nevertheless remains on the contemporaneous relationship between Δy_it and Δfa_it, along with how the β₃ coefficient estimates the cyclicality of foreign aid spending. A finding that β₃ < 0 indicates that aid spending is countercyclical; β₃ > 0 suggests procyclical expenditure; while if Δy_it is found not to be statistically significant then aid expenditures are acyclical. Unobservable heterogeneity is accounted for by including time fixed effects λ_t and country-specific fixed effects μ_i. Given the presence of lagged dependent variables and the potential for finite sample bias (Nickell 1981) our preferred estimator is the System Generalized Method of Moments (SYS-GMM) (Blundell and Bond 1998). We also calculate robust standard errors for hypothesis testing, after adjusting for clustering in the donors. SYS-GMM estimates are said to be consistent if serial correlation is absent at second order or higher (though first-order serial correlation is anticipated) and if the instruments employed are valid. We test each of those assumptions for the models estimated in the following section. While our chosen estimator can control partially for potential endogeneity of the independent variables, the SYS-GMM instruments may not control fully for potential endogeneity of the output gap. We therefore enter the following instruments: contemporaneous oil price inflation in national currency terms; the lagged output gap for the whole of the OECD; and the contemporaneous change in exports as a percentage of GDP (e.g., Darby and Melitz 2008), which account for the influence of external exogenous shocks on the domestic output gap. The data for those additional instruments come from the U.S. Energy Information Administration (2019) and (OECD 2018c).

While foreign aid might be procyclical over the business cycle, asymmetries in the cyclical response could exist between economic upturns and downturns. We can test for differences in the cyclical responses of spending between upturns and downturns by estimating the following equation:

The specification of (2) differs from (1) by including \(\Delta {\text{y}}_{\text{it}}^{ + }\) and \(\Delta {\text{y}}_{\text{it}}^{ - }\), as well as the associated lagged terms, which enables us to test for differences in the cyclical responses of aid spending between upturns and downturns. Upturns are defined as periods when the actual level of output is above the trend level, i.e., a positive output gap (\(\Delta {\text{y}}_{\text{it}}^{ + }\)), whereas downturns are periods when actual output is below trend, i.e., a negative output gap (\(\Delta {\text{y}}_{\text{it}}^{ - }\)). So \(\Delta {\text{y}}_{\text{it}}^{ + } \ne 0\) and equals \(\Delta {\text{y}}_{\text{it}}^{ + }\), when \(\Delta {\text{y}}_{\text{it}}^{ + }\) takes a positive value and zero otherwise, while \(\Delta {\text{y}}_{\text{it}}^{ - } \ne 0\) (and equals \(\Delta {\text{y}}_{\text{it}}^{ - }\)) when \(\Delta {\text{y}}_{\text{it}}^{ - }\) takes a negative value and zero otherwise.

The above equations can be extended to test the extent to which governments rely on economic upturns to influence aid spending patterns. Our first hypothesis, that “left-wing governments are more likely to increase aid expenditures procyclically in economic upturns”, can be tested by estimating:

Specification (3) extends (2) by entering the dummy variable lft_it, as well as the interaction terms \(\left( {\Delta {\text{y}}^{ + } \times {\text{lft}}} \right)\) and \(\left( {\Delta {\text{y}}^{ - } \times {\text{lft}}} \right)\), where lft is a dummy variable that equals one whenever the government is classified as “left-wing”. Data on the ideological preferences of governments come from the Database of Political Institutions 2017 (Scartascini et al. 2018), which classifies countries based on party orientations with respect to economic policy. Countries are classified as “left-wing” when they are categorized as “communist, socialist, social democratic, or left-wing” (Ibid., p. 8); the alternatives are “right” or “center”. Support for hypothesis (1) arises when \(\left( {\Delta {\text{y}}^{ + } \times {\text{lft}}} \right)\) is statistically significant and δ₅₀ > 0.

Similarly, we can test the influence of the government’s electoral majority on the cyclicality of foreign aid spending during economic upturns (hypothesis (2)) by entering the interaction terms \(\left( {\Delta {\text{y}}^{ + } \times {\text{maj}}} \right)\) and \(\left( {\Delta {\text{y}}^{ - } \times {\text{maj}}} \right)\), where maj is the “fraction of seats held by the government” (Scartascini et al. 2018):

Governments with smaller majorities are more likely to increase aid spending procyclically during economic upturns, so \(\left( {\Delta {\text{y}}^{ + } \times {\text{maj}}} \right)\) should be statistically significant and the estimated coefficient λ₅₀ negatively signed. Equation (4) also can be used to test the influence of elections from hypothesis (3) from the \(\left( {\Delta {\text{y}}^{ + } \times {\text{elec}}} \right)\) and \(\left( {\Delta {\text{y}}^{ - } \times {\text{elec}}} \right)\) terms, where “elec” equals one during election years and zero otherwise. Data on the election years comes from Scartascini et al. (2018), who reports information on the years remaining in the current term (elections are those periods with zero years left). Support for the hypothesis “governments are not as likely to spend procyclically in an economic upturn in an election year” arises when \(\left( {\Delta {\text{y}}^{ + } \times {\text{elec}}} \right)\) is statistically insignificant.

To compare the cyclicality of foreign aid with the cyclicality of domestic and social security spending, we estimate:where the variable de denotes domestic expenditures and sp refers to expenditure on social protection. Domestic expenditure is the sum of spending on health, education and social protection, with observations taken from the OECD’s Government Expenditure by Function database (OECD 2018a). From hypotheses (4) and (5), foreign aid expenditure is expected to be more procyclical than both domestic expenditure and social protection, implying that Δy_it should be statistically significant and the cyclicality coefficient on foreign economic aid, α₆ in (2), positive and larger in magnitude than φ₆ from (5) and τ₆ in (6).

Our sixth hypothesis tests whether the cyclical pattern of foreign economic aid exerts a permanent effect on the level of such spending. The α coefficients in (2) capture spending patterns in expansions and contractions. When the cyclical pattern is asymmetric, spending will react differently to \(\Delta {\text{y}}^{ + }\) and \(\Delta {\text{y}}^{ - }\). For example, if governments respond to growing tax revenues during expansions by increasing spending but choose not to contract spending when tax revenues fall during contractions, such behavior will increase spending levels over time, through cyclical upward ratcheting (Hercowitz and Strawczynski 2004). To measure the importance of that spending mechanism, we can calculate the ratcheting coefficient from (2), which is given by \({{\Phi }} = \left( {{{\upalpha }}_{3} + {{\upalpha }}_{4} + {{\upalpha }}_{5} } \right) - \left( {{{\upalpha }}_{6} + {{\upalpha }}_{7} + {{\upalpha }}_{8} } \right)\) (see Hercowitz and Strawczynski (2004). The value of Φ and its statistical significance will determine whether spending ratchets upwards or downwards.

Table 1 presents the descriptive statistics for the variables used in the empirical analysis, as well as the sources for the economic time series used. Before turning to the estimation results for our hypotheses it is useful to comment on the overall cyclical properties of foreign aid spending across the economic cycle. The evidence from our sample suggests that \(\Delta {\text{y}}_{\text{it}}\) is statistically significant from estimating (1) and that foreign aid is procyclical, since we find that \({\upbeta }_{3}\) > 0, with an estimated coefficient of 0.063.¹¹ The full estimation results are available upon request.

To test our first hypothesis, “Left-wing governments are more likely to increase aid expenditures procyclically in economic upturns”, the results of estimating (3) are reported in Table 2. We find that the interaction term \(\left( {\Delta {\text{y}}^{ + } \times {\text{lft}}} \right)_{\text{it}}\) is statistically significant and the coefficient estimate is positive, implying a larger procyclical marginal effect from left-wing governments during economic upturns, compared to all other political ideologies. Utilizing the estimates for \(\Delta {\text{y}}_{\text{it}}\) and \(\left( {\Delta {\text{y}}^{ + } \times {\text{lft}}} \right)_{\text{it}}\), we can conclude that during periods of left-wing governments, \(\Delta {\text{y}}_{\text{it}}\) is statistically significant and the estimated cyclicality coefficient for foreign aid spending is 0.218, while for the other ideologies \(\Delta {\text{y}}_{\text{it}}\) is statistically insignificant, implying that foreign aid is acyclical; the estimated coefficient is 0.099.

To consider the influence of government majorities on the cyclicality of foreign aid expenditures we estimate (4), the results from which also are shown in Table 2. Consistent with our prediction that “governments with smaller electoral majorities are more likely to increase aid expenditures procyclically in economic upturns”, we find that the interaction term \(\left( {\Delta {\text{y}}^{ + } \times {\text{maj}}} \right)_{\text{it}}\) is statistically significant and that the estimated coefficient is negatively signed. Smaller values of maj are consistent with narrower majorities for the incumbent government; the negatively signed estimated coefficient implies that as the majority gets smaller the extent of procyclicality rises for foreign aid. If the proportion of the seats held by the government falls by one standard deviation, then the marginal gain to procyclicality is 0.0661, which is relatively modest compared to the cyclicality estimate of 0.426 for \(\Delta {\text{y}}_{\text{it}}^{ + }\). The results from (4) also are supportive of the prediction that “Governments are not as likely to spend procyclically in an economic upturn in an election year”, since we find that \(\left( {\Delta {\text{y}}^{ + } \times {\text{elec}}} \right)_{\text{it}}\) is not statistically significant.

Hypotheses (4) and (5) indicate that during economic recessions foreign economic aid spending should be more procyclical than (1) domestic expenditures and (2) social security. The results of estimating Eqs. (2), (5) and (6) are presented in Table 3. Evidence in support of our hypotheses comes from finding that \(\Delta {\text{y}}_{\text{it}}^{ - }\) is statistically significant only for foreign aid spending; the estimated coefficient is 0.071. Both domestic current expenditures and social security spending are found to be acyclical during economic downturns. It is interesting to note that the cyclicality coefficients for foreign economic aid are of broadly similar magnitudes during upturns and downturns. We tested whether the cyclicality coefficients are statistically significantly different from one another i.e. \({\text{H}}_{0} :{{\upalpha }}_{3} - {{\upalpha }}_{6} = 0\). The difference in the estimated coefficients is only 0.009 and the associated Z-statistic is 0.16, which implies that we cannot reject the null hypothesis, and so conclude that the cyclical response of foreign economic aid is symmetric across economic upturns and downturns.

Finally, we consider whether the cyclical pattern of spending has a permanent influence on the level of spending for foreign economic aid (hypothesis (6)). Using the estimates presented in Table 3, the cyclical ratcheting coefficient is calculated to be 0.273; the estimated standard error is 0.121. The standard t test therefore is 2.26, implying significance at the 5% level. From that result, we can infer that foreign aid spending will be 0.27% higher at the end of the cycle compared to its beginning. While the estimated coefficient is smaller than earlier estimates of the ratcheting coefficient given for total spending in the OECD [see e.g., Hercowitz and Strawczynski (2004), who present a ratchet coefficient of 1.07], the result remains important in demonstrating the long-term consequences arising from the cyclicality of foreign aid spending. We derived that value since, during upturns, the combined coefficients of \(\left( {{{\upalpha }}_{3} + {{\upalpha }}_{4} + {{\upalpha }}_{5} } \right)\) yield an estimate of 0.273, implying that for a one percentage rise in the positive output gap, foreign economic aid rises by 0.27%. By contrast for the economic downturns, the combined cyclicality coefficients are only \(\left( {{{\upalpha }}_{6} + {{\upalpha }}_{7} + {{\upalpha }}_{8} } \right) = 0.003\), suggesting that the contraction in foreign economic aid is far smaller during the economic downturns than its increase in the upturns.

The initial motivation for this paper was the observation that procyclical public spending might be explained with reference to “voracity effects” (Lane and Tornell 1996; Tornell and Lane 1999). If political pressures to increase government spending are more intense in economic upturns than downturns, why are political pressures more intense in the former than in the latter?

The methodology employed herein differs from methodologies applied elsewhere that focus on the cyclicality of government expenditures. The usual approach is to analyze the cyclicality of government expenditure over a full economic cycle. However, in this paper it also is important to distinguish between cyclicality in upturns versus downturns. Our new approach offers additional insight because, even if two government spending programs are described as “procyclical”, important differences are likely if: (1) one set of expenditure is procyclical because it is procyclical in both upturns and downturns and (2) the other is procyclical because spending is highly procyclical in upturns and simply acyclical in downturns. In this paper, the evidence is that, between 1999 and 2015, OECD governments’ foreign aid expenditures were procyclical over a complete economic cyclical and procyclical in both economic upturns and recessions.

Analysis of the determinants of the cyclicality of public spending and of its policy implications depends on insights that are possible when focusing (independently) on economic upturns and downturns. When focusing on the determinants of procyclical expenditures, the question posed in the introduction of the paper was: ‘Do marked differences exist in voter awareness in “good times” and in “bad?” Here, the evidence is that differences in awareness are salient, and that such differences strengthen the intensities of the political pressures that governments experience to increase aid expenditure in economic upturns. When focusing on foreign aid, an obvious reduction in pressure to increase government expenditure materializes when donors move into recession.

In economic upturns governments have discretion in responding to pressures (for example, from domestic lobby groups and international agencies) to spend more on programs that are not as popular with the median voter. As evidence of the conjecture that they act strategically, this paper focuses on both the motivation and the opportunity to rely on an informational advantage in an economic upturn. Strategically, governments increase expenditures when voters are least aware of programmatic tax costs. Left-wing governments also are particularly inclined to rely on economic upturns to increase foreign aid. Governments of all political persuasions are more likely to rely on upturns the more they are “vulnerable” because of small legislative majorities (Frey and Schneider 1978, 1981). As an informational advantage is not as marked when an election is forthcoming (because electoral debates increase tax awareness), governmental foreign aid expenditures are not as procyclical in economic upturns.

In economic downturns voters are more fiscally aware. They question the “tax price” of government spending and the adequacy of specific expenditures (the inspection effect). With a more isolationist approach, the differences in voters’ awareness of the benefits of foreign aid, relative to domestic expenditure (Downs 1960), become even more extreme. With falling tax revenues and difficulty relying on borrowing, governments divert expenditure from foreign programs to domestic programs.

While the evidence presented is consistent with each prediction independently, it also is the case that the pattern of evidence supports the proposition that governments act strategically. Governments time increases in less popular expenditures to occur in economic upturns. While the literature on “fiscal illusion” focuses on changes in the level of government spending, “fiscal illusion” also is relevant when analyzing the composition of government expenditure.

Turning to assessments of the policy implications of our analysis:

Given those policy criticisms, the question is whether it is possible to introduce reform. One possibility is that each nation-state “should” independently introduce “expenditure rules” that apply equally in both upturns and recessions. The problem is that politicians are unlikely to have incentives to constrain the discretion they might otherwise exercise over public expenditure. An alternative approach is to raise the pressure on politicians to adopt international spending targets. In 1970, the United Nations called on member states to spend 0.7% of gross national income on foreign aid in upturns and in recessions. Vaubel (1986) argues that politicians have an incentive to respond to pressure from international organizations because they are able to blame those organizations if they introduce changes that prove unpopular. Griffith-Jones (2003) and Dreher and Vaubel (2002) refer to the blame politicians placed on international financial institutions when they adopted unpopular monetary policies.

The existence of incentives to accept expenditure constraints is important, but the prospects for reform are not unqualified. While reliance on the same expenditure target in economic upturns and in recessions will reduce the cyclicality induced by “fiscal illusion”, reliance on a target defined as a proportion of gross national income will not eradicate cyclicality. Moreover, while politicians may have incentives to respond, the extent to which they are likely to respond is debateable. For example, the United Kingdom agreed in principle to the UN’s 1974 target, but it was 2013 before it was achieved. The United Kingdom enshrined a duty to comply in law in 2015, but that commitment is not enforceable through the law courts¹³ and discussion of whether political parties will continue to adhere to the “constraint” remains on-going (Full Fact 2018). Politicians might have an incentive to respond to pressure to adopt expenditure targets, but the expenditure targets are not self-enforcing “rules”.

The costs of failing to reduce politicians’ discretion to rely on “fiscal illusion” in economic upturns depends on the nature of the distortion. If voters underestimate the benefits (Heinrich et al. 2016) and overestimate the costs of overseas aid (Heinrich and Kobayashi 2018), the question is whether reliance on “fiscal illusion” (to increase overseas aid) is the most efficient way of providing a merit want. Many might argue that governments should rely on the alternative policy option of providing accurate information (to correct misperceptions). Even if governments have their own policy agendas (e.g., Cusack 1997; Keech 1995; Wittman 1983), assessments of good governance are likely to depend on governmental transparency and sensitivity to the preferences of well-informed voters.

The overarching conclusion in this paper is that systematic differences exist in voter awareness of public expenditure in “good” times and in “bad”. Differences in voter awareness over the economic cycle are relevant when explaining differences in the intensities of political pressures for increased government spending and they are relevant when assessing the effectiveness of such expenditures.