Fiscal Illusion and Cyclical Government Expenditure: State Government Expenditure in the United States

A well‐established literature argues that fiscal illusion increases the level of government expenditure. This article focuses on the proposition that fiscal illusion also influences the cyclicality of government expenditure. Predictions are formed with reference to government reliance on high income elasticities of indirect tax revenues and on intergovernmental transfers. Predictions are tested with reference to the expenditures of 36 states in the United States from 1980 to 2000. Government expenditures are more likely to be procyclical when citizens systematically underestimate the cost of taxation.


I INTRODUCTION
Procyclical public expenditure increases as national income increases and falls as national income decreases. Economists usually anticipate counter-cyclical expenditure (Alesina et al., 2008), but empirical studies increasingly report procyclical expenditure (Gavin et al., 1996;Talvi and V egh, 2005;Alesina et al., 2008;Woo, 2009;Abbott and Jones, 2013). Studies that explain procyclical expenditure highlight the relevance of political pressures for public expenditure. The likelihood of procyclical expenditure increases if there are 'voracity effects', i.e. if political pressures to raise public expenditure increase as national income increases and fall as national income falls (Lane and Tornell, 1996;Tornell and Lane, 1999). Empirical studies indicate that political pressure to increase public expenditure is a significant determinant of procyclical expenditure (Lane, 2003;Abbott and Jones, 2013).
This article focuses on the way that the intensity of pressures to increase government expenditure changes over the economic cycle. The intensity of political pressure is gauged with reference to politicians' willingness to increase expenditure to win votes (Downs, 1957). Pressure is intense when politicians believe they are able to increase expenditure without alarming voters that taxation will increase.
A well-established literature focuses on the circumstances in which voters are likely to underestimate taxation systematically. 1 This literature has focused on the level of government expenditure. In this article, the objective is to introduce and explore the proposition that fiscal illusion is also likely to increase the cyclicality of government expenditure.
When focusing on the cyclicality of government expenditure, some of the sources of fiscal illusion are likely to be more relevant than others. In the next section of the article, the intention is to review the many sources of fiscal illusion. Studies that have tested the reasons why voters underestimate taxation have focused on variables that capture the extent to which individuals are likely to underestimate taxation. For example, one of these variables is the complexity of the tax system. Studies test the significance of these sources of fiscal illusion separately. In the case of tax complexity, they test the proposition that adding a single variable (to measure tax complexity) to the socioeconomic variables that are likely to influence the level of government spending, will reveal that tax complexity is a significant determinant of the level of taxation. In this article, a first attempt (to our knowledge) will be made to assess the relative importance of all of these different sources of fiscal illusion on the level of government expenditure. The intention is to add this full complement of variables (variables that are said to be relevant when considering different sources of fiscal illusion) to the socio-economic variables that are likely to influence the level of government spending, to compare the impact that they exert on the level of government spending. 2 In section III of the article, attention focuses on two sources of fiscal illusion that are likely to be particularly relevant when considering the level and the cyclicality of government expenditure. These sources of fiscal illusion are more important when national income is increasing. If politicians are myopic in economic upturns (Downs, 1957), they are likely to indulge pressures to increase government expenditure with little regard for the difficulties they will face when the economy moves into recession. If fiscal illusion increases as national income increases, fiscal illusion increases the likelihood that there will be 'voracity effects'. How relevant are these sources of fiscal illusion when national income is increasing? How relevant are these sources of fiscal illusion when anticipating that governments will spend procyclically?
If politicians indulge lobby-group pressures to increase public expenditure in an economic upturn, the pressures they face to sustain expenditure in a recession will depend on the extent to which governments are able to borrow.
In this article, predictions (presented in sections II and III of the article) are tested with reference to the level and the cyclicality of government expenditures in 36 states of the United States between 1980 and 2000 (because subnational governments find it more difficult to borrow in a recession). 3 Section IV of the article describes the data set and the models used to test predictions. The results are presented in section V. Conclusions and policy implications are discussed in the final section of the article.

II FISCAL ILLUSION AND THE LEVEL OF PUBLIC EXPENDITURE
An established literature has identified the circumstances in which voters are likely to underestimate the costs of taxation. When testing for fiscal illusion, variables are chosen to proxy the likelihood that there is fiscal illusion. As Oates (1988) explains, the hypothesis is that E (an estimate of budgetary size) depends on: X (a vector of variables that includes income, tax price, and socio-demographic variables that proxy for tastes and preferences); F (a variable that makes it difficult for voters to estimate tax costs) and u (an error term). If the coefficient b in equation (1) is significant, E is higher than would be predicted with reference to X: In this literature, the determinants of fiscal illusion are: (1) A complex tax system: Wagner (1976) argued that complexity is the important consideration. In Wagner's study, the vector F in equation (1) comprises a single variable, S. With R i equal to the share of revenues generated from each of n tax categories (e.g. property taxes; general sales taxes; charges and fees), S ¼ P n i¼1 R 2 i . The Herfindahl index achieves its maximum value (of unity) when a jurisdiction generates all of its own revenues from a single source. The minimum possible value is when revenues are divided equally across n different tax categories. Tax systems are not as complex the higher the value of the index. In Wagner's regression, S was significantly negative (suggesting that expenditure is lower the simpler the tax structure). Wagner could reject the null hypothesis of non-association at a 0.01 level of significance.
(2) Renter illusion: Renter illusion occurs when local taxes are levied on the owners of property and not on their tenants. In these circumstances, tenants believe that the costs of government expenditure are low (even if taxes are shifted forward in rent charges). It follows that the larger the number of renters in a jurisdiction, the higher the level of government expenditure (other things equal). There is evidence that local jurisdictions tend to spend more per capita on local public services when there is a relatively large fraction of citizens who rent accommodation (see, e.g. Bergstrom and Goodman, 1973;Peterson, 1975). Tenants fail to understand the link between levels of local spending and the level of rent that they pay. Once again, the F-vector in equation (1) is usually a single variable (the proportion of residents in a jurisdiction who rent accommodation).
(3) High Income elasticities of Tax Revenue: Oates (1975) introduced the argument that high income elasticities of tax revenue are likely to increase fiscal illusion. When national income increases, taxes increase the share of income spent in the public sector. Tax awareness is lower when there are forces in the economy that automatically increase government revenue (without any formal need to introduce a new tax or to change the tax rate). With a progressive tax rate structure, an increase in national income increases tax revenue and politicians have no reason to bring this to the attention of taxpayers. The empirical literature, focusing only on the impact of high income elasticities of tax revenue, is mixed. Oates (1975) drew attention to a positive relationship between high income elasticities of tax revenues and increases in the growth of government expenditure. Craig and Heins (1980:267) reported a positive relationship between high income elasticities of tax revenue and high levels of government expenditure. They supported ". . .the idea that elasticity drives spending". Craig and Heins focused on the impact of estimates of elasticities provided by the Advisory Commission of Intergovernmental Relations in 1976-77. They reported a statistically significant positive relationship for 1970. Di Lorenzo (1982 and Feenberg and Rosen (1987) did not find a significant relationship between higher income elasticities of tax revenue and higher levels of public sector spending. (4) Flypaper Effects. There is a 'flypaper effect' if sub-central government expenditure is higher than anticipated when sub-central governments receive lump sum intergovernmental transfers. When money moves in the public sector, money is spent in the public sector ('money sticks where it hits'). The 'flypaper effect' was first identified by Gramlich and Galper (1973), but many other studies have subsequently reported its existence (Hines and Thaler (1995) and Bailey and Connolly (1998) survey this literature). While there are many different explanations for the existence of the 'flypaper effect', local voters appear to underestimate the full costs of public expenditure when a local jurisdiction is in receipt of intergovernmental transfers (Oates, 1979). 4 In this instance, the F-vector would consist of a single variable that measures local jurisdictions' reliance on intergovernmental transfers. 4 Public choice theory explains the 'flypaper effect' with reference to pressure for increased public expenditure. A local jurisdiction's receipt of intergovernmental grants (paid, in part, by others in the country) provides local politicians with the opportunity to respond to pressures from: local bureaucrats (Niskanen, 1971;Romer and Rosenthal, 1980); local interest groups (Dougan and Kenyon, 1988); and local voters (Oates, 1979). However, at the heart of this analysis is the argument that voters systematically underestimate the tax cost of expenditure programmes. Voters focus on the average cost of public expenditure, rather than the marginal cost of public expenditure.
(5) Public Debt Illusion. If voters ignore future tax liabilities, government borrowing appears to reduce the costs of taxation. Vickrey (1961:133) refers to ". . . "a public debt illusion" . . .(when). . . individuals pay no attention to their share in the liability represented by the public debt. . .". Buchanan and Wagner (1977:142) argue that ". . .events of fiscal history strongly support the hypothesis that unconstrained access to public borrowing will tend to generate excessive public spending". Once again, there is fiscal illusion if a single variable (as a proxy of government borrowing) is significant (e.g. Kormendi, 1985).
For each hypothesis, the empirical evidence is mixed. 5 Surveys of the literature on fiscal illusion are critical of tests that have been employed. A first concern is that the proxies employed to measure fiscal illusion are far from perfect. When focusing on complexity, and when focusing on income elasticities of tax revenue, some taxes are simply more transparent than others (irrespective of the Herfindahl Index, or the size of the income elasticity of tax revenue). It has long been argued that taxpayers are likely to know less about indirect taxes than they know about direct taxes. John Stuart Mill ([1848] 1994) argued: "If all taxes were direct, taxation would be much more perceived than at present, and there would be a security, which now there is not, for economy in the public expenditure" (cited in Dell'Anno and Mour ao, 2012: 272). 6 In this study, the intention is to distinguish between the impact exerted by high income elasticities of direct tax revenue and the impact exerted by high income elasticities of indirect tax revenue.
A second criticism endemic in this literature is that OLS regressions are not able to deal with the problem of endogeneity. Oates (1988) is critical that it might not be an increase in fiscal illusion (proxied by variables chosen to capture fiscal illusion) that increases government expenditure. Instead, plans to increase government expenditure might explain reliance on variables used to proxy fiscal illusion. For example, governments that anticipate an increased demand for public expenditure might chose to rely on a wider diversity of taxes (a more complex system) to lower the excess burden associated with increased taxation. Also, local jurisdictions (that anticipate that voters will demand greater expenditure as income increases) might choose to rely on tax systems that have high income elasticities of tax revenue, to reduce the 5 Not all studies report evidence of fiscal illusion (e.g. Gemmell et al., 2003), and in some studies, there is also evidence that voters underestimate public expenditure (e.g. Cullis and Jones, 1987). 6 Buchanan (1967) argued that a sales tax creates fiscal illusion because it is absorbed in the payments citizens make for private goods and services. Sausgruber and Tyran (2005) argued that indirect taxation is less visible than direct taxation. Pommerehne and Schneider (1978), Dollery and Worthington (1996) and Dell'Anno and Mour ao (2012) report evidence that public spending is higher (other things equal) if governments rely on indirect taxation. When Oates (1999:2) considered the optimal tax for sub-central governments, he argued that: ". . .sales taxes do not get very good marks on a fairness or ability-to-pay criterion. In addition, they do not stack up at all well on the public-choice criterion of providing the electorate with accurate and visible signals of the costs of public programs". (Oates was aware of the argument that property taxes might create fiscal illusion but he was more concerned with the obfuscation created by sales tax).
transactions costs of continually increasing tax rates in the future. In this article, the intention is to deal with the problem of endogeneity by instrumenting changes in income and federal aid.
To date, studies have focused on each single source of fiscal illusion. 7 In this article, it is important to test the relative influence of different causes of fiscal illusion, because (as will be argued in the next section) some determinants are likely to be more important than others when focusing on the cyclicality of public expenditure. Here, the objective was to offer a first test of the relative importance of these determinants of fiscal illusion, controlling for the problem of endogeneity.

III FISCAL ILLUSION AND THE CYCLICALITY OF GOVERNMENT EXPENDITURE
Pressures to increase public expenditure are more intense when politicians believe that they are able to win votes (by increasing public expenditure) without alarming voters of any increase in taxation. Downs (1957) argued that politicians focus myopically on electoral payoffs over a 4-or 5-year electoral interval. If fiscal illusion increases as national income increases, politicians 'sow the seeds' of problems they will face in recessions. The more that fiscal illusion increases as national income increases, the greater the likelihood that there will be 'voracity effects'.
Two sources of fiscal illusion are particularly sensitive to changes in national income: (i) high income elasticities of sales taxes and (ii) receipts of intergovernmental transfers. When governments rely on high income elasticities of indirect tax, it is not simply the case that they will not need to introduce a new tax (or a new tax rate), it is also the case they can remain more at ease that the electorate will also not be aware of the increase in tax that they are paying automatically as national income increases. Abbott and Jones (2013) have shown that intergovernmental transfers are often procyclical. They argue that politicians at federal government are more likely to increase intergovernmental transfers when national income is increasing. When national income is falling, federal governments are more constrained by falling tax revenues. When national income is falling, politicians at federal government are also more mindful that there are pressures to increase 'direct' federal government expenditure on health, education, defence, social security, etc. These expenditures are more transparent than intergovernmental transfers. At federal government, politicians are mindful that they will win votes if they persuade voters they are able to manage the economy (e.g. Jones and Hudson, 1996). As they are under pressure to increase 'direct' expenditures in an economic downturn, they are less receptive to pressures for intergovernmental transfers in an economic downturn. With this mix of incentives and constraints, political pressures for intergovernmental transfers are more successful in economic upturns than in economic downturns.
In this article, the argument is that fiscal illusion is likely to increase as income increases, because pressures for governmental transfers are more likely to be successful in economic upturns than in downturns. Mueller (2003) has argued, local politicians are aware that ". . .the more the government spends holding taxes constant the happier voters are. . ." and ". . .the higher the probability of incumbent politicians being re-elected" (p. 223). With high income elasticities of sales tax revenues and with heavy reliance on intergovernmental transfers, politicians in sub-central governments are more likely to act indulgently when national income is increasing.
Other things equal: • high income elasticities of indirect tax revenue and reliance on intergovernmental transfers increase the level of sub-central government expenditures; • high income elasticities of indirect tax revenue increase the likelihood of procyclical government expenditure; • reliance on intergovernmental transfers increases the likelihood of procyclical government expenditure.

IV FISCAL ILLUSION AND SUB-CENTRAL GOVERNMENT EXPENDITURES IN THE UNITED STATES
To test the extent to which fiscal illusion impacts on the level of government spending, we can adapt the model of Craig and Heins (1980) as follows: where G is state current government spending per capita; Y is state personal income per capita; F is federal aid per capita; A is the percentage of population aged over 65 years; C is the percentage of the population that lives in an urban area; S is the proportion of revenues originating at the state level; B is state borrowing as a proportion of GDP; D is total state debt outstanding as a proportion of GDP; GI is the Gini index of income inequality; ID is a measure of state government ideology; NW is the proportion of the population that is non-White; SE is the income elasticity of state sales tax revenue; PIE is the income elasticity of state personal income tax revenue; RHF is a Herfindahl index of state revenue proportions; OWN is the rate of owner occupation. This model differs from the original Craig and Heins' (1980) specification in a number of respects. Firstly, A covers the population aged 65 years and over, rather than the population aged 18 and over, because this is now a more significant driver of state spending. Secondly, population density is no longer included because it has a strong association with the proportion of the state population that lives in an urban area. Thirdly, we add a number of control variables not previously considered, which could be important in influencing the level of government spending, such as state ideology (measured between 0 for 'conservative' and 100 for 'liberal') and the proportion of non-White population (e.g. Facchini, 2014). Fourthly, fiscal illusion is measured through six alternative approaches: the rate of owner occupation (OWN), which inversely picks up the proportion of renters; the Herfindahl index of state revenue proportions (RHF); the percentage of revenue originating locally (S); the amount of state borrowing (B) and the total stock of debt (D); the income elasticities of sales tax (SE) and personal income taxation (PIE). We rely on more recent estimates of tax elasticities in Bruce et al. (2006), rather than the estimates available in the Advisory Commission of Intergovernmental Relations used by Craig and Heins (1980). Through including six proxies for fiscal illusion we hope to establish whether a consistent pattern of fiscal illusion emerges, or whether one proxy emerges as being particularly significant for determining government spending. Except for the coefficients for RHF and for OWN, the expectation is that all of the estimated coefficients will be positive.
Government spending per capita, the proportion of revenue originating at state level, borrowing for each state and the total stock of outstanding debt are available in the Annual Survey of State Government Finances. We also use the same source to calculate the Herfindahl index of revenue sources. The population data for C, A and NW are taken from the US Census Bureau along with the Gini coefficient data for income inequality and the rate of owner occupation. The personal income per capita series is derived from the Bureau of Economic Analysis. Our measure of political ideology is the NOMINATE measure of state government ideology (Berry et al., 2010). Federal aid per capita data are available from the US Office of Management Budget. The estimates of income elasticities are provided by Bruce et al. (2006) for Personal Income Taxes and for Sales Taxes (based on tax base data between 1967 and 2000). They produced estimates for 48 states but this article focuses on the 36 states for which both elasticities are reported (to analyse a consistent sample of states). 8 The model is a cross-section regression, estimated separately for 3 years: 1980, 1990 and 2000, to test the robustness of the model over time. Given the potential for problems of endogeneity, we use a three-stage least squares (3SLS) estimator that controls for potential endogeneity in personal income per capita and federal aid per capita. We instrument state personal income per capita using personal income of the USA regional division that the USA state belongs too. There are nine regional divisions used by the US Census Bureau: New England; Mid-Atlantic; East North Central; West North Central; South Atlantic; East South Central; West South Central; Mountain; Pacific. By using the regional personal income per capita, we control for the impact of economic shocks arising from neighbouring states. The instruments for federal aid per capita are urban population; population density; and political party coincidence between the state governor and the USA president. To test the second set of hypotheses, the cyclicality of government spending can be estimated as follows (see for example Arena andRevilla (2009) andLane (2003)) for similar models: where g it is the log of government spending per capita for the ith state at period t, y is the log of state gross domestic product per capita, while br denotes state government borrowing per capita. We add time dummies D t to account for idiosyncratic time shocks, while e it is a white noise error. The dynamic specification of our model allows for potential persistence in g it , through the inclusion of the lagged dependent variables, and for the possibility that government spending might react with a lag in response to shocks in GDP or government borrowing. These lagged responses could be due to administrative delays or lags caused by delays in the budgetary process. The cyclicality of government spending is picked up through the b 1 and b 2 coefficients, with b 1 < 0 for example indicating government spending is counter-cyclical and b 1 > 0 suggesting procyclical behaviour. The growth in state borrowing could also be expected to positively influence the growth in government spending, so p 1 and p 2 are expected to be positive. This dynamic panel data model can be estimated through the SYS-GMM estimator of Blundell and Bond (1998). The use of SYS-GMM estimator may be limited in this application, as it is expected that the number of time periods are relatively few and the number of individual units relatively large. We use data from 1980 to 2000 for these estimations.
We are also interested in the extent to which the procyclicality varies with the income elasticity of tax revenues and the reliance of states on federal government transfers. Investigating the effect of the tax elasticity, equation (3) is amended as follows: where E i denotes the income elasticity of sales tax revenue or personal income tax revenue. Statistically significant interaction term Þand positively signed coefficients are consistent with the hypothesis that procyclical spending reacts positively to a higher income elasticity of tax revenues.
To account for the influence of transfers on the cyclicality of spending, we interact the growth in federal transfers (Dtr it ) with the growth in state GDP (Dy it ), so that (3) becomes: The predictions are that φ 1 and φ 2 will be positively signed and the interaction terms will be statistically significant.

V ESTIMATION RESULTS
The estimation results from equation (1) 1980, 1990 and 2000). We find the following: (1) Levels of government expenditure and fiscal illusion. Among the six chosen measures of fiscal illusion, income elasticity of sales tax revenue has statistical significance. The SE variable is statistically significant for the 2000 period and the estimated coefficient is positively signed as expected. The fact that the sales tax income elasticity is statistically significant but the personal income tax elasticity is not, is important when reflecting on the unresolved debate in the literature as to whether or not high income elasticities of tax revenue increase levels of government spending. The results indicate a growing trend in fiscal illusion and suggest that the debate is sensitive to differences in the mix of income and sales tax revenues in different tax bases. The impact of fiscal illusion is likely to be more obvious the more sub-central governments rely on indirect tax revenues.
(2) Levels of public expenditure and intergovernmental transfers. In Table 1, the federal aid per capita series is the most statistically significant control variable and it is signed positively throughout. Own revenues are statistically significant for the 1990 period but the estimated coefficient does not have the expected sign.
(3) Borrowing and the level of public spending. The role of government borrowing across the USA states is found to have a statistically significant effect on the level of government spending for the 1990 period only. In this case, the estimated coefficient has the expected positive sign. In contrast, the total stock of outstanding debt is not statistically significant.
(4) The cyclicality of public expenditure and high income elasticities of indirect tax revenue. Turning to the estimation of the cyclicality of state spending, the estimated cyclicality coefficients are reported in Table 2. We find that Dy it is statistically significant and the positively signed coefficient of 0.106 implies a procyclical pattern of spending, albeit with a lag. In contrast, the growth in borrowing does not affect the growth in spending.
Whether the income elasticity of sales tax revenue positively influences the procyclical nature of public spending can be captured by the interaction terms detailed in (4). The results presented in the second and third columns of Table (2) show that Dy itÀ1 Â E i ð Þis statistically significant and the estimated coefficient positively signed as expected. The lagged procyclical public spending grows when the illusory impact of taxation is stronger for high incomes.
The interaction terms are not statistically significant when using the personal income tax elasticity. (5) The cyclicality of public expenditure and intergovernmental transfers. The results in the final column are also consistent with the hypothesis that procyclicality increases with the growth in intergovernmental transfers. The estimated coefficient φ is positively signed and the interaction Dy itÀ1 Â Dtr itÀ1 ð Þ is statistically significant. The value of the coefficient is greater than one and this is consistent with the hypothesis that the 'flypaper effect' can be explained with reference to fiscal illusion. Once again, fiscal illusion increases the procyclicality of government expenditures over and above the inherent procyclicality in these sources of governmental revenue. Notes: Estimates produced using 3SLS estimation, controlling of for the endogeneity of personal income and federal aid. The instrumented equations are not reported. T-ratios are reported in parentheses calculated from robust standard errors and * indicates significance at the 5% level.

VI CONCLUSIONS
In the literature that explains the existence of procyclical expenditure, attention has focused on the importance of 'voracity effects' (Lane and Tornell, 1996;Tornell and Lane, 1999). In this article, attention has focused on the proposition that fiscal illusion increases the likelihood of 'voracity effects'. With 'voracity effects', political pressures to increase public expenditure are more intense as national income increases. In this article, it has been argued Notes: Estimates of the cyclicality equation derived using the SYS-GMM dynamic panel data estimator of Blundell and Bond (1998) (estimates of the time dummy coefficients are also generated but not reported to conserve space). The number of instruments displayed refers to the number of lagged values of the regressors (in level and first difference form) that are used as instruments for the explanatory variables. SC denotes the tests for serial correlation at 1st, 2nd and 4th order. The autocorrelation tests are for zero autocorrelation in first-differenced errors. The Hansen test of over-identifying restrictions is reported, which is robust to heteroskedasticity and autocorrelation. * indicates significance at the 5% level (t-ratios from robust standard errors are shown in parentheses).
that the 'intensity' of political pressures depends on the way they are received (as well as on the way that they are exerted). VoteÀmaximising politicians are more inclined to increase public expenditures when they believe they are able to increase expenditures, without alarming voters that there is an increase in taxation.
There are two sources of fiscal illusion that increase the likelihood of 'voracity effects'. With high income elasticities of sales tax revenue and dependence on intergovernmental transfers, voters are more likely to underestimate the costs of taxation as national income increases. The first testable prediction in this article is that high income elasticities of sales tax revenue and dependence on intergovernmental transfers increase the level of fiscal public expenditure (other things equal).
The analysis of government spending in 36 USA states between 1980 and 2000 produced results that are consistent with the proposition that high income elasticities of sales tax revenue and receipt of intergovernmental transfers are important determinants of fiscal illusion. This is the first test (to our knowledge) of the relative importance of the full complement of potential sources of fiscal illusion (as described in this article). It is also a test that has been designed to address the problem of endogeneity. The results are consistent with the proposition that high income elasticities of sales tax revenue and receipt of intergovernmental transfers are important determinants of fiscal illusion.
When reporting that some potential sources of fiscal illusion are not significant, the results in this article are consistent with results reported by other studies of fiscal illusion. 9 The results also shed insight into the existence of the unresolved debate about the relevance of high income elasticities of tax revenue. Differences in empirical results that refer to the significance of high income elasticities of tax revenue are likely to depend on the extent to which taxation is raised by direct taxation and by indirect taxation. The results in this article indicate that voters are more inclined to underestimate the costs of taxation when taxation is raised by indirect taxation.
The results that both high income elasticities of sales tax revenues and intergovernmental transfers are relevant sources of fiscal illusion is important because, in this article, the second testable proposition is that fiscal illusion also increases the cyclicality of government expenditure. Having addressed the issue of endogeneity, the results are consistent with the proposition that the level of public expenditure is higher than anticipated (other things equal) because voters underestimate the costs of taxation.
Turning to the second set of empirical tests reported in this article, the proposition is that fiscal illusion will increase the level and the cyclicality of government expenditure. The first important observation (in Table 2) is that the coefficient on the interaction term Dy itÀ1 Â E i ð Þis significant and substantially larger than the 9 The results in this article are consistent with results in other studies that question the relevance of other sources of fiscal illusion, e.g. Blom-Hansen's (2005) assessment of the relevance of renter illusion in the Netherlands. coefficient on Dy itÀ1 . The presence of high income elasticities of sales tax revenue is important (other things equal) when explaining the likelihood that sub-central governments will spend procyclically. The second important observation (in Table 2) is that the coefficient on Dy itÀ1 Â Dtr itÀ1 ð Þ is significant (and higher than the insignificant coefficient on the volatility of income). As in other studies (e.g. Abbott and Jones, 2013), sub-central governments are more likely to spend procyclically, the more they rely on intergovernmental transfers. For both income elasticities of sales tax revenues and intergovernmental transfers, the relative sizes of coefficients are consistent with the hypothesis that fiscal illusion increases the procyclicality of government expenditure over and above the inherent procyclicality in these sources of government revenue.
In this article, the first set of results (correcting for endogeneity) indicate that fiscal illusion increases the level of government spending. The second set of results indicate that the same determinants of fiscal illusion (high income elasticities of sales tax revenue and dependence on intergovernmental transfers) and their impact on fiscal illusion are also relevant when explaining the cyclicality of government expenditure.
The tests in this article are consistent with the proposition that myopic, vote-maximising politicians increase the likelihood of procyclical government expenditure by indulging pressures to increase governments expenditures in economic upturns (and by depleting the budgetary reserves that might otherwise have been employed to mitigate economic recession). While there are circumstances in which procyclical spending might maximise a community's welfare, commentators are dismissive because the circumstance are very restrictive (Halland and Bleaney, 2009). 10 As fiscal illusion is relevant when explaining the existence of government failure, the costs of fiscal illusion exceed the static misallocation costs that are associated with 'excessive' levels of public expenditure (e.g. Mueller, 2003;Cullis and Jones, 2009). Fiscal illusion also imposes costs because there is additional uncertainty if fiscal illusion creates 'excessive' cyclicality. Low-income individuals are vulnerable in economic recessions (e.g. Bordo and Meissner, 2011) and governments are not able to offer the levels of social protection that the vulnerable rely on in recessions. 11 With evidence that the sources of fiscal illusion that increase the likelihood of 'voracity effects' are important determinants of the level of public expenditure, fiscal illusion is also relevant when analysing the cyclicality of government expenditure. 10 Lane (2003) argues that there are circumstances in which procyclical government spending might maximise welfare if private and public consumption are complements. Alesina et al. (2008) argue that procyclical spending might increase welfare in developing countries because of the failures they experience in capital markets. 11 While this article has focused on sub-central governments, fiscal illusion is also likely to be relevant when explaining the difficulties that federal governments experience when they must take action to mitigate the costs of recessions (Abbott and Jones, 2012). R.F. Harrod (Keynes' biographer) stated that Keynes believed economic policy would be made by ". . .a small group of enlightened men. . .in accordance with the 'public interest'" (cited by Buchanan and Wagner, 1977: 84), but here the evidence is consistent with the argument that politicians' own political ambitions may also be relevant when explaining the cyclicality of government expenditure.