Block Grants and Education Expenditure

Abstract

This article investigates the effects of block grants on education expenditures using panel data from Canadian provinces over the period 1982 to 2008. Our main empirical identification strategy relies on the use of the allocation formula for equalization grant—a component of the Canadian federal block grant. The results indicate that block grants have stimulative effects on provincial education expenditure. Our results suggest that a one dollar increase in per capita federal grants is associated with an increase in per capita education expenditure of about Can$0.21, which is roughly proportional to the share of education in total provincial spending. The results are robust to various sensitivity checks.

Keywords

block grants flypaper effect education expenditures federalism

Expenditure on education has long been considered as an essential public investment that helps foster productivity and economic growth. Education is also often considered as a key to both societal and individual prosperity and progress. For this reason, although the funding mechanism may vary, education accounts for a significant part of government outlays in many countries. In most federations, including Canada, the provisions of public services such as education are the responsibilities of subnational governments. However, it is well known that education expenditures have spillover effects to other jurisdictions in the federation. The presence of such spillover effects has long been considered as one of the main rationales for the presence of intergovernmental grants in a federation. An important policy question then is how do subnational governments respond to block grants? Do block grants have stimulative effects on education expenditure?

In the literature, the effects of block grants on recipient subnational governments’ spending are often discussed in the framework of the so-called flypaper effect—an empirical phenomenon where grants have a larger stimulatory effect on recipient subnational governments’ expenditure than what is predicted in economic theory (see Oates 1999; Gamkhar and Shah 2007; and Payne 2009, for in-depth discussions on flypaper effect). The justifications for such common empirical findings include potential endogeneity of grants (Knight 2002), institutional factors (Wyckoff 1988; Lutz 2010), politics (Inman 2008; Singhal 2008), and others (Dollery and Worthington 1996). Regardless of the justifications, previous studies find that federal block grants have stimulative effects on recipient subnational governments’ expenditures. Ladd (1993) also finds empirical evidence that the increase in state tax bases associated with the US Tax Reform Act of 1986 generates a flypaper effect as the increase in the tax revenue was not fully used to cut tax rates.

A number of previous studies also examine the effects of block grants on education expenditures. Most of these studies use data from US states.¹ Tsang and Levin (1983) provide an earlier survey of the literature. Cohn (1987) investigates the effects of educational grants on education spending. He finds that for each dollar of federal grants, recipient local governments raise their spending by seventy cents and the remaining part of the grant is used to lower taxes. For US states, Fisher and Papke (2000) also find that unconditional education grants stimulate education spending although the magnitude of the response is less than dollar-for-dollar implying that recipient governments may use some of the grants to lower their taxes. Recent studies such as Gordon (2004), Cascio, Gordon, and Reber (2013), and others, on the other hand, focus on the effects of education-specific grants in the United States. Gordon (2004) examines the effects of title I—US federal government grant to support elementary and secondary education—on schools’ revenues and spending. She finds that while such grants raise school spending initially, the effects of the grants on school spending become very minimal overtime. Cascio, Gordon, and Reber (2013), on the other hand, find that title I results in an increase in school spending for the average southern district.

As in many other federations, education expenditures in Canada are the responsibility of provincial governments. However, the federal government also helps finance education and other public services by providing grants to the provincial governments. Approximately 20 percent of the federal expenditure is directly paid to provincial governments in the form of grants.² Although currently there are no education-specific federal grants, provincial governments can use the various general block grants to partly finance education expenditures, as these grants are largely unconditional and recipient governments can spend those funds according to their own chosen priorities. Thus, Canada provides opportunity for a good natural experiment to examine the impact of block grants on subnational governments’ expenditure on education. However, there is a paucity of empirical studies that assess the effects of federal grants on Canadian provincial education spending.

The main objective of this article is to examine the effects of federal block grants on provincial education expenditures using panel data from Canadian provinces over the period 1982 to 2008. In our analysis, we address the potential problem of endogeneity of grants that has often plagued some of the previous empirical studies using the instrumental variable (IV) estimation method. We use the allocation formula for equalization grant—a component of the Canadian federal block grant—as our main exclusion restriction in the IV estimation method. Such an identification strategy is broadly similar in spirit to the growing literature of Regression Discontinuity Design. We employ an empirical methodology that is very similar to Dahlby and Ferede (2015).

Our empirical results show that federal block grants have stimulative effects on provincial education expenditures. We find that every one dollar increase in per capita federal block grants is associated with about a Can$0.21 increase in provincial education expenditure per capita. During the period under consideration, the average share of education spending in total provincial expenditure is about 22 percent. Thus, our empirical estimated results suggest that, for every dollar of federal grants, a fraction of the funds goes to education expenditure roughly in proportion to its share in the provincial total expenditures. The estimated results are robust to various sensitivity checks and well within the range of results obtained in previous studies for other countries.

The remaining part of the article is organized as follows. In the second section, we provide a brief description of the structure of federal grants in Canada. The empirical specification and the data are presented in the third section. We present and discuss the empirical results in the fourth section and conclude in the fifth section.

Institutional Background

The Canadian federal transfer payment system has evolved over time. At present, there are three main types of specific and general purpose federal transfers to the provinces: Canada Health Transfer (CHT), Canada Social Transfer (CST), and Equalization Program (EP). We provide a brief description of these major grants and how they evolve over time below.

Historically, the federal government was providing transfers to provincial governments through the Established Program Financing (EPF) and Canada Assistance Plan (CAP). The EPF was meant to financially support provincial governments’ provision of health and education services. On the other hand, CAP was a matching grant designed as a cost-sharing arrangement with provincial governments for social assistance programs. In fiscal year 1996 to 1997, the federal government combined EPF and CAP and replaced those with one block grant known as the Canada Health and Social Transfer (CHST). This block grant provided transfers to provincial governments in support of the provision of health, social, and education services. Again in 2004, the CHST was restructured and divided into two separate forms of grants: CHT and CST.

As their names suggest, in principle, CHT and CST are specific grants meant to provide financial support to provincial governments in order to provide health and social services, respectively. However, these grants are largely unconditional block grants and only a few restrictions are imposed on the block grants, and the funds are included in the general revenues of the provinces. One of the minor conditions related to CHT (and also previously to EPF) is that provinces respect the five criteria of the Canada Health Act (universality, accessibility, portability, comprehensiveness, and public administration). To this effect, provisions for withholding funding were introduced. However, the federal government does not require provinces to have comprehensive reporting and auditing of the use of grants. Thus, provincial governments include these grants as part of their general revenues and use the funds according to their own chosen priorities.

Another important part of the Canadian intergovernmental transfer system is the EP, which has been around since 1957. The main principle behind the EP is to enable provincial governments to provide comparable levels of services using comparable levels of taxes. Thus, the main objective of the equalization grant program is to address differences in fiscal capacity of provinces. The Canadian federal government provides equalization grants using a formula that is based on the per capita fiscal capacity—which is a measure of provinces’ ability to raise revenue—of the province and that of the “standard” provinces.³ During the period 1982 to 2006, the “standard” provinces include British Columbia, Saskatchewan, Manitoba, Ontario, and Quebec. However, since 2007, the definition of “standard” provinces includes all the ten provinces.

For the different tax revenue categories, the per capita equalization grant entitlements are calculated as the difference between the per capita tax base of the standard provinces and that of the province both evaluated at the national average standard tax rate. Then, the total equalization grant to be provided for a province is determined by adding all the equalization grant entitlements from the various revenue categories. That is, the total per capita equalization grant entitlement for a province is calculated using the following formula:

Eq = Max {0, \sum_{j = 1}^{n} {\bar{τ}}_{j} ({\bar{B}}_{j} - B_{j})}

where Eq is total per capita equalization grant, ${\bar{τ}}_{j}$ denotes the national standard tax rate for revenue category j, and $B_{j}$ and ${\bar{B}}_{j}$ are the per capita tax base of the province and of standard provinces for revenue category j, respectively. If the total entitlement for all revenue categories is positive, then the province is categorized as a receiving province and receives per capita equalization grants Eq. However, if Eq is not positive, it means that the province has a better than national average standard fiscal capacity and is not entitled to receive equalization payments. During the period under consideration, Alberta and Ontario did not receive any equalization grants. British Columbia, Saskatchewan, and Newfoundland also did not receive equalization grants in some years.

In our empirical analysis, the federal grant variable (Grants) is the sum of CHT, CST, and equalization grants all expressed in terms of per capita 2002 Canadian dollars. During the years in which the matching grant CAP was in place, we exclude this grant and focus only on the other nonmatching block grants.⁴ The grant variable shows a lot of variations both over time and across provinces. During the period under consideration, the mean per capita grants range from Can$481 for Alberta to Can$2,436 for Prince Edward Island. For all ten provinces, the per capita grants range from Can$292 to Can$2,986 with an average value of Can$1,397 for the period.

Empirical Specification, Methodology, and Data

Specification

The literature on intergovernmental grants shows that lump-sum grants can stimulate receiving jurisdictions’ expenditure (see Gamkhar and Shah 2007 and the references contained therein). In order to analyze the effect of grants on the provincial education spending, we use the following basic specification:

E_{i t} = β_{0} + β_{1} G_{i t} + β_{2} X + μ_{t} + γ_{i} + u_{i t},

where E_it is the real per capita education expenditure in province i in year t, G_it is real per capita federal grants, and X contains all other relevant control variables. µ _t captures a full set of year effects. The time effects control for those factors that may have a common effect on provinces such as business cycle conditions and federal grant policy changes. Time-invariant provincial fixed effects are denoted by γ _i , and these allow us to control any secular differences in government spending across the provinces. Due to the variation in local government expenditure responsibilities and the corresponding differences in provincial governments’ own expenditures, as in Schmidt and McCarty (2008), we use consolidated provincial and local education spending as the dependent variable in our regression models. We convert the variable into real per capita by dividing by the total population and deflating with provincial gross domestic product deflator (2002 = 100).

In our analysis, the key variable of interest is federal grants. As indicated in the previous section, the federal grants variable is the sum of CHT, CST, and equalization grants, expressed in per capita terms in 2002 dollars. These are block grants that are to a large extent unconditional. When provincial governments receive higher federal grants, they may raise their spending on public services such as education or lower their tax rates or both. The literature on flypaper effects suggests that lump-sum intergovernmental grants have a stimulative effect on government spending. Thus, other things remaining the same, we expect that higher federal grants raise education expenditure. Thus, in equation (2), our main coefficient of interest is β₁ and we expect β₁ > 0.

As indicated before, the block grants are unconditional, and once the grants arrive they become part of general revenues and can be used to finance any public services. For this reason, in order to explicitly investigate the effect of the grants on education expenditure, it is important to control for changes in other provincial expenditure categories. In our analysis, we include changes in provincial spending on health, social, and others (those excluding provincial spending on education, health, and social services) as control variables. Obviously, in addition to federal grants, the spending capacity of provinces greatly depends on their own tax revenues. Thus, following Gordon (2004), we also include per capita provincial tax revenue as a control variable. We expect this variable to have a positive effect on education expenditure.

Previous studies of government expenditure, such as Gamkhar and Oates (1996), Di Matteo and Di Matteo (1998), Baker, Payne, and Smart (1999), Dahlby and Ferede (2015), among others, include the jurisdiction’s average income as one of the control variables. Thus, we control for per capita personal income in our analysis. This variable captures the effect of the overall economic condition of the province. Generally, public services such as education are considered as normal goods and the demand for these services will increase with income. Thus, we expect the coefficient of per capita income to be positive.

Arguably, one of the most important driving forces behind public education expenditures is the population share of the school-aged population. Other things remaining the same, jurisdictions with a higher share of school-aged population will be forced to spend more on education. To capture this effect, as in Poterba (1998), Harris, Evans, and Schwab (2001), and Schmidt and McCarty (2008), we include Schoolage, the share of the province’s population that is school-aged (those between five and nineteen years of age inclusive) as a covariate. As the share of school-aged population increases, provincial governments will be expected to spend more on education in order to satisfy the education services needs of these cohorts. Thus, we expect this variable to have a positive effect on public education expenditures.

The demographic makeup of a province can influence both the total amount and the type of public services to be provided. The elderly and the very young in particular necessitate various social and health care–related spending by the government. Thus, an increase in aging population and very young population may induce policy makers to favor directing more funds toward health services ultimately affecting education expenditure adversely (see Poterba 1997; Gradstein and Kaganovich 2004; and Arvate and Zoghbi 2010, for more in-depth analysis). To account for such demographic effects, we include the shares of the population who are sixty-five years and above (old) and those below five years of age (young) as control variables. We expect these coefficients to have negative effects on education expenditure as more and more funds are directed toward health services and away from education outlays.

Often, a government’s spending decision and the composition of government expenditure can be influenced by the political ideology of the governing party and the presence of an election. As education spending is sometimes viewed as redistribution in kind, the ideology of the governing party may matter (see Brender and Drazen 2013, for an exposition on the larger changes in expenditure composition during election years in mature democracies). As discussed in Baker, Payne, and Smart (1999), Kneebone and McKenzie (2001), and others, left-leaning governments generally have a tendency to be pro-spending. Thus, to capture this ideological effect on education expenditures, we include dummy variables (New Democratic Party [NDP]) that are equal to one if the premier of the province belongs to the NDP and Conservative that is equal to one if the governing party belongs to the Conservative party. The excluded category is the Liberal party.⁵

The literature on political cycles indicates that incumbent politicians have opportunistic incentives, and they attempt to maximize their chances of reelection by introducing popular public policies (see Alesina, Silvia, and Trebbi 2006). If this is indeed the case and voters have a strong preference for higher public spending on education, we expect provincial governments to raise their education expenditures at election times. In order to account for such potential effects of political cycles, we include an election dummy variable (Election) that is equal to one in the years in which there is an election. We expect this variable to have a positive effect on education expenditures if provincial governments tend to bias their outlays toward education services at election times.

Identification Strategy

As discussed previously, the allocation of some of the federal grants is based on a formula that compares the per capita fiscal capacity of the province and that of the standard provinces. While provinces with fiscal capacities below the standard fiscal capacity receive equalization grants, those provinces with fiscal capacities above the standard do not receive grants. Thus, the equalization allocation formula exhibits discontinuity in the relationship between equalization grants and fiscal capacity at the point where the per capita tax base of the province is equal to the average per capita tax base of standard provinces. In our empirical analysis, we rely on this discontinuity to identify the exogenous effects of grants on education spending. Thus, following Dahlby and Ferede (2015), we use the equalization grant formula, denoted as Formula, as the main excluded instrument for grants.⁶ Dahlberg et al. (2008) also use a somewhat similar identification strategy. In a similar situation, Angrist and Lavy (1999) use the class size allocation formula as an instrument for class size conditional on a smooth function of enrollment. Since fiscal capacity of the province can have a direct effect on government spending, we need to control for this element of the equalization grant formula. As is common in similar studies, we use a smooth polynomial function of fiscal capacity.

In order to further address the potential endogeneity problem associated with the nonequalization component of the block grants, we rely on one of the main grant allocation conditions. Both CHT and CST (and their predecessors EPF and CHST) have been allocated through a mix of cash and equalized tax points transfers. The total transfers are allocated on equal per capita basis. Thus, the growth rate of the population share of the province plays an important role in the allocation of this block grant component. Consequently, we use the growth rate of the population share of the province in the country (PopShare) as an additional excluded instrument for grants. Thus, in our empirical analysis, we use the equalization grant allocation formula (Formula) and the growth rate of population share of provinces (PopShare) as instruments for block grants (table 1).

Table 1.

Summary Statistics, 1982 to 2008.

Variable	Definitions	Observation	Mean	SD	Min	Max
Education	Provincial education expenditure per capita	270	1,917	319	1,269	2,614
Grants	Federal block grants (sum of CST, CHT, and equalization grants) per capita	270	1,397	754	292	2,986
Income	Personal income per capita	270	24,044	3,911	14,748	33,879
School age	Share of the population who are between five and nineteen years of age (in %)	270	21.4	2.4	16.9	31.0
Young	Share of the population who are under five years of age (in %)	270	6.5	1.0	4.5	8.7
Old	Share of the population who are sixty-five and above years of age (in %)	270	12.2	1.8	7.2	15.4
Conservative	Dummy variable for the Conservative government	270	0.5	0.5	0.0	1.0
NDP	Dummy variable for the NDP government	270	0.2	0.4	0.0	1.0
Election	Dummy variable for the election year	270	0.3	0.4	0.0	1.0
Tax revenue	Provincial total tax revenue per capita	270	3,692	836	1,816	6,231
Social	Changes in provincial social expenditure per capita	260	18	111	−976	580
Health	Change in provincial health expenditure per capita	260	58	118	−295	489
Others	Changes in provincial total expenditure per capita (excluding those on education, health and social services)	260	20	256	−1,038	824
Fiscal capacity	The sum of provincial per capita tax bases evaluated at their respective national average standard tax rates	270	4,662	1,897	1,501	12,839
PopShare	Growth rate of the population share of the province in the country	260	−0.004	0.008	−0.028	0.019
Formula	Equalization grant allocation formula	270	0.141	0.257	−0.568	0.583

Note: All monetary values are per capita in 2002 Canadian dollars. Government expenditure data are consolidated provincial and local (in fiscal year). CST = Canada Social Transfer; CHT = Canada Health Transfer; NDP = New Democratic Party.

Data

The data set for our empirical analysis comes from various sources. Annual provincial and local government expenditures, personal income, prices, total population, and its different categories come from Statistics Canada database (CANSIM). We obtain the data on governing political parties and elections from the Canadian Parliamentary Guide. The federal grants and provincial fiscal capacity used in the allocation of equalization payments are administrative data obtained from Finance Canada.

Empirical Results and Discussion

Results

We begin the discussion of our empirical results by presenting the first-stage regression results in table 2. Columns (1) to (4) show the regression results associated with various nonlinear specifications of fiscal capacity. One can check the relevance of the excluded instruments, Formula and PopShare, by looking at the t values of the instruments. The results suggest that the instruments for grant are statistically significant in all cases. Various statistical tests also confirm the relevance of our excluded instruments. More specifically, our empirical model is not under identified as the null hypothesis of under identification is rejected at the 5 percent significance level for all regressions. Furthermore, the Kleibergen–Paap F statistics for weak instruments bias indicates that the results are not affected by the presence of weak instruments. The validity of the excluded instruments is further confirmed by the standard Hansen overidentification test. Thus, we conclude that our instruments for grants are statistically valid.

Table 2.

First-stage Regressions, 1982 to 2008.

Dependent variable: Block grants per capita (Grants)	(1)	(2)	(3)	(4)
Income	0.0694**	0.113***	0.0977***	0.115***
Income	(0.0351)	(0.0337)	(0.0358)	(0.0343)
Young	85.21	−17.25	1.528	−28.50
Young	(71.75)	(56.37)	(65.61)	(60.38)
School age	−130.1***	−74.18***	−88.04***	−80.27***
School age	(32.27)	(25.02)	(29.59)	(26.37)
Old	−217.8***	−180.5***	−228.8***	−201.1***
Old	(70.37)	(53.90)	(68.73)	(55.47)
Conservative	−36.02	−19.58	−62.78*	−28.61
Conservative	(33.05)	(25.44)	(32.76)	(26.88)
NDP	−10.46	−7.195	−99.90*	−31.50
NDP	(43.82)	(32.73)	(51.56)	(36.23)
Election	−4.994	−1.980	−4.266	−3.077
Election	(16.19)	(14.83)	(15.87)	(15.36)
Social	−0.0335	−0.104	−0.128	−0.122
Social	(0.0726)	(0.0786)	(0.0867)	(0.0833)
Health	0.0789	−0.0796	0.0188	−0.0571
Health	(0.0981)	(0.0847)	(0.0891)	(0.0857)
Others	0.0138	0.0205	0.0111	0.0168
Others	(0.0365)	(0.0292)	(0.0336)	(0.0304)
Tax revenue	0.0818	−0.0319	0.0204	−0.0197
Tax revenue	(0.0651)	(0.0529)	(0.0600)	(0.0551)
	Instruments	Instruments	Instruments	Instruments
Formula	3,931.9***	5,549.2***	3,742.0***	5,198.4***
Formula	(752.2)	(694.8)	(601.1)	(694.7)
PopShare	−9,048.2***	−8,823.8***	−12,346.3***	−9,662.7***
PopShare	(3,349.9)	(2,738.7)	(3,550.9)	(2,812.5)
Fiscal capacity	Quadratic	Cubic	Linear spline	Quadratic spline
Under identification (p value)	.001	.001	.000	.001
Kleibergen–Paap F statistic	14.05	32.01	19.72	28.20
Overidentification (p value)	.30	.27	.27	.29
Number of observations	260	260	260	260
Adjusted R ²	.966	.975	.969	.974

Note: All regressions include provincial fixed effects and year effects. Heteroscedasticity and autocorrelation robust standard errors are in parentheses. NDP = New Democratic Party.

Significance levels are indicated by *** for 1 percent, ** for 5 percent, and * for 10 percent.

We now turn to the discussion of the effects of grants on provincial education expenditure. The empirical results are reported in table 3. As indicated previously, the dependent variable is real per capita provincial and local education expenditure. All regressions include provincial fixed effects and year effects. We include additional covariates to control for the effects of various economic, demographic, and political factors on provincial education spending. We present Newey–West standard errors that are robust to arbitrary heteroscedasticity and arbitrary autocorrelation up to two lags.⁷

Table 3.

Education Expenditure Regression Results, 1982 to 2008.

Dependent variable: Provincial education expenditure per capita (Education)	(1) OLS	(2) OLS	(3) OLS	(4) IV	(5) IV	(6) IV	(7) IV
Grants	0.118**	0.170***	0.113***	0.205**	0.194***	0.239***	0.204***
Grants	(0.0526)	(0.0463)	(0.0412)	(0.0804)	(0.0687)	(0.0824)	(0.0715)
Grants _it ₋₁	0.0723
Grants _it ₋₁	(0.0713)
Grants _it ₋₂	0.0175
Grants _it ₋₂	(0.0784)
Income	0.0171	0.0199*	0.0273**	0.0245**	0.0243*	0.0258*	0.0250*
Income	(0.0105)	(0.0104)	(0.0109)	(0.0122)	(0.0129)	(0.0133)	(0.0133)
Young	−74.60	−54.82	−115.4***	−125.9***	−123.4**	−138.8***	−127.0***
Young	(45.26)	(40.69)	(43.25)	(43.58)	(47.49)	(46.31)	(48.21)
School age	3.649	2.134	26.77*	29.09**	27.39	28.98*	30.05
School age	(14.21)	(12.35)	(14.57)	(14.16)	(22.01)	(16.87)	(20.70)
Old	−51.84*	−59.56**	−55.17**	−43.20*	−45.54	−46.27*	−42.85
Old	(27.58)	(26.38)	(23.86)	(24.85)	(28.40)	(26.02)	(27.11)
Conservative	36.50*	34.41	41.39**	50.54**	49.47**	49.89**	50.35**
Conservative	(21.89)	(20.95)	(20.49)	(20.05)	(19.71)	(20.19)	(19.74)
NDP	105.8***	108.3***	113.4***	118.2***	117.6***	104.6***	117.8***
NDP	(28.78)	(28.10)	(28.31)	(28.27)	(28.05)	(31.25)	(28.67)
Election	5.489	7.126	8.793	9.966	9.816	10.95	9.954
Election	(13.82)	(13.71)	(12.20)	(12.43)	(12.36)	(12.38)	(12.37)
Social	0.0703	0.0673	0.0480	0.0493	0.0501	0.0445	0.0483
Social	(0.0781)	(0.0776)	(0.0539)	(0.0544)	(0.0548)	(0.0564)	(0.0555)
Health	0.0839	0.0687	0.0290	0.0162	0.0197	0.0122	0.0149
Health	(0.0626)	(0.0618)	(0.0536)	(0.0550)	(0.0541)	(0.0548)	(0.0543)
Others	0.0255	0.0214	0.0317	0.0330	0.0329	0.0351	0.0329
Others	(0.0313)	(0.0295)	(0.0262)	(0.0271)	(0.0270)	(0.0274)	(0.0271)
Tax revenue	0.131***	0.142***	0.0721	0.0647	0.0673	0.0552	0.0635
Tax revenue	(0.0486)	(0.0464)	(0.0455)	(0.0448)	(0.0486)	(0.0468)	(0.0483)
Fiscal capacity			Quadratic	Quadratic	Cubic	Linear spline	Quadratic spline
Number of observations	250	260	260	260	260	260	260
Adjusted R ²	.888	.895	.910	.907	.907	.906	.907

Note: All regressions include provincial fixed effects and year effects. Heteroscedasticity and autocorrelation robust standard errors are in parentheses. OLS = ordinary least square; IV = instrumental variable; NDP = New Democratic Party.

Significance levels are indicated by *** for 1 percent, ** for 5 percent, and * for 10 percent.

Although our main empirical analysis is based on IV estimation methods, for the sake of comparison, we begin our discussion by presenting ordinary least square (OLS) estimates in column (1). Arguably, block grants may potentially have lagged effects on education expenditure. We first check for this in our basic regression by including up to two-period lag of grants.⁸ However, the results suggest that the lagged effects of grants are statistically insignificant and as a result we drop them from our main analysis. The coefficient of contemporaneous grants, on the other hand, is positive and statistically significant at the 1 percent level. The result indicates that a one dollar increase in per capita federal block grants is associated with about Can$0.12 increase in provincial education expenditure per capita.

In column (2), we drop the lagged grants and reestimate the model with OLS. The coefficient of grants is again positive and statistically significant at the 1 percent level. The numerical magnitude of the coefficient estimate is also much higher. Note also that the adjusted R ² of the model improved once we drop the statistically insignificant lagged grants variable, confirming the irrelevance of these variable in explaining education expenditure. The coefficient of real income per capita is positive and statistically significant.

For ease of comparison with IV estimates to be discussed later, we reestimate the model with OLS after controlling for a quadratic form of fiscal capacity. These results are reported in column (3). The results indicate again that block grants have stimulative effects on education expenditure.

So far, we have focused on OLS estimation results assuming that federal grants are exogenous. However, as we have argued before, the OLS estimation results may be biased and unreliable as the federal grant variable is endogenous. Consequently, we present the IV estimation results in columns (4) through (7) of table 3. The specifications of columns (4) to (7) are basically similar, and they differ only in the nonlinear functional form of fiscal capacity that we include as part of our identification strategy. We use the equalization grant formula, Formula, and the growth rate of provincial population share (PopShare) as instruments. Based on adjusted R ² and other conditions, the regressions with quadratic specification for fiscal capacity show the best fit for the data. So we focus our discussion on results presented in column (4), which control for a quadratic form of fiscal capacity.

In column (4), we present the IV estimation results that control for a quadratic form of fiscal capacity. Comparing results of columns (3) and (4), we see that the numerical magnitude of the effects of grant on provincial education spending is much higher when we use the IV estimation suggesting that the effects of grants may be downward biased if the endogeneity of the variable is not taken into account. The results indicate that, as expected, federal block grant has a statistically significant positive effect on education expenditures. The results suggest that a one dollar increase in per capita federal grant is associated with about a Can$0.21 increase in education spending per capita. This result is broadly consistent with the findings of previous related studies. During the period under consideration, the average share of education spending in total provincial expenditure for all provinces is about 22 percent (ranges from 17.6 percent to 27 percent). Thus, our empirical estimated results suggest that for every dollar of federal grants a fraction of the funds goes to education expenditure roughly in proportion to its share in the provincial total expenditures. Thus, the stimulative effect of intergovernmental grants on education spending is quite large.

The other control variables have generally the expected signs. In order to better examine the flypaper effects of block grants, we control for changes in all other noneducation expenditure categories. The coefficients of these other expenditure categories are, however, statistically insignificant throughout perhaps due to the potential endogeneity problems of their own. The other fiscal variable that we control in our analysis is the provincial tax revenue per capita. As expected, this variable is positive; however, it is statistically insignificant. One may be surprised with the statistical insignificance of the variable. But the insignificance of this variable is due to the potential correlation of the variable with the fiscal capacity variable.

It is known that one of the most important deriving factors for education spending is the population share of those in school age. The coefficient of Schoolage is, as expected, positive and statistically significant at the 1 percent level. Our results suggest that a one percentage point increase in the share of school-aged population increases education expenditure per capita by about thirty-six dollars. Fernandez and Rogerson (2001) also find similar results for the United States using state-level data.

The coefficients of the share of the population who are below five (young) and above sixty-four (old) are as expected negative and statistically significant at the 1 percent level. The results suggest that a one percentage point increase in the share of the young reduces per capita provincial education expenditure by about 126 dollars. Further, the results indicate that every one percentage point increase in the share of the population who are sixty-five and above is associated with a decrease in per capita provincial education expenditures by about forty-three dollars. These results are consistent with the findings of previous studies for other countries such as Poterba (1997, 1998), Gradstein and Kaganovich (2004), and others. The implication of this finding is that, as many commentators and researchers highlighted in the past, the aging of the population and the associated increase in public outlays for the provision of health services puts a strain on education outlays.

Our results also indicate that both NDP (left-wing government) and Conservative (right-wing government) have positive effects on education spending even though the former seems to have a much larger effect on education expenditure. This suggests that all political parties see education expenditure as important and are favorable to this expenditure category. While left-leaning governments are generally known to be pro-spending on various public services including education, it is surprising that we find a positive and significant effect of Conservative governments on education expenditure. However, such a result is not uncommon in the literature. See, for example, Potrafke (2011) who finds a somewhat similar result for Germany.

In column (5), we control for a cubic form of fiscal capacity. The results are very close to what we obtain in column (4), suggesting the robustness of our results to the degree of polynomial function of fiscal capacity controlled for. Although we control for a polynomial form of fiscal capacity in columns (4) and (5), theoretically any nonlinear form of the variable can be used. Thus, as a robustness check, we control for a linear- and quadratic-spline form of fiscal capacity in columns (6) and (7), respectively.⁹ These results are generally similar to those obtained in columns (4) and (5) except that the magnitudes of the coefficients of grants are slightly larger when we use spline forms of fiscal capacity.

In sum, as expected, our results indicate that block grants have strong stimulative effects on provincial education expenditure. Although, to the best of our knowledge, there are no empirical studies that focus on the effects of Canadian federal grants on provincial education spending, Di Matteo and Di Matteo (1998) find that grants have stimulative effects on provincial health care expenditure. Dahlby and Ferede (2015) also find that intergovernmental grants have stimulative effects on provincial total government expenditure. Thus, our results are broadly consistent with results obtained in other related studies.

Sensitivity Analysis

In this section, we check the sensitivity of our main result to various robustness checks. The sensitivity analysis is conducted based on our preferred results of column (4) of table 3, which controls for a quadratic form of fiscal capacity. The regression results include all relevant control variables as discussed before, but we report the coefficient estimates of just grants and income for the sake of brevity.

The literature on intergovernmental grants suggests that matching and block grants can have different effects on government expenditure. As we are interested in the investigation of the effects of block grants, we exclude the matching grant CAP that was in place in Canada prior to 1995 from our analysis. This is in fact a common approach employed in previous Canadian studies; see, for example, Dahlby and Ferede (2015).¹⁰ However, one may wonder how the inclusion of this grant type affects our results. Results reported in column (1) of table 4 show that the coefficient estimate of grants is strikingly similar to what we report previously. Thus, our result is robust whether one includes CAP in the grant variable or not.

Table 4.

Robustness Checks (IV), 1982 to 2008.

Dependent variable: Provincial education expenditure per capita (Education)	(1)	(2)	(3)	(4)	(5)	(6)
	Including CAP	Spatial factors	Five-province standard	Including Trend	Equalization grants only^a	Equalization grants only^b	(7) Log-linear^c
Grants	0.205***	0.214***	0.215**	0.207*			0.133***
Grants	(0.077)	(0.0800)	(0.0993)	(0.106)			(0.0448)
Equalization					0.189**	0.187**
Equalization					(0.0862)	(0.0845)
Income	0.024**	0.0246**	0.0326***	0.0606***	0.0277**	0.0220*	0.427***
Income	(0.012)	(0.0119)	(0.0106)	(0.0149)	(0.0120)	(0.0123)	(0.164)
Year effects	Yes	Yes	Yes	No	Yes	Yes	Yes
Province effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Fiscal capacity	Quadratic	Quadratic	Quadratic	Quadratic	Quadratic	Quadratic	Quadratic
Number of observations	260	260	240	260	260	260	260
Adjusted R ²	.909	.908	.915	.834	.905	.906	.919

Note: All regressions include provincial fixed effects, year effects, and all the relevant control variables. The coefficients of these variables are not reported for the sake of brevity. The robustness checks are based on column (4) of table 3 and the same set of instruments are used unless otherwise stated. Heteroscedasticity and autocorrelation robust standard errors are in parentheses. CAP = Canada Assistance Plan.

^aIn column (5), we use equalization grants only and ignore nonequalization grants. We use Formula as an excluded instrument.

^bIn column (6), we focus on equalization grants, but nonequalization grants are controlled for as part of the revenues available for provincial governments. We use Formula as an excluded instrument.

^cIn column (7), all variables (with the exception of the election and ideology dummies) are in log-linear form.

Significance levels are indicated by *** for 1 percent, ** for 5 percent, and * for 10 percent.

In closely interdependent jurisdictions such as Canadian provinces, for various political and economic reasons, the education spending decision of a provincial government may well depend on education expenditures of other provinces in the federation. In fact, empirical results of earlier studies such as Case, Rosen, and Hines (1993) and Acosta (2010) find empirical support for the presence of such spillover effects. In order to check the robustness of our main results to the potential presence of spillover effects in education expenditures from other provinces in the federation, in column (2), we include the weighted-average (weighted by the inverse of the distance between the main population centers of the neighboring provinces) per capita education expenditure of other provinces as an additional explanatory variable to control for the spending-spillover effects. The results indicate that our empirical result is not sensitive to the inclusion of such education spending-spillover effects.

As indicated before, during the period under consideration, the number of “standard” provinces changed from five to ten. As a robustness check, in column (3), we restrict our estimation period only to the years in which the five-province standard is in place. Again our key variable of interest seems to be robust to this sensitivity check.

In column (4), as in Schmidt and McCarty (2008), we include a time trend instead of year effects to allow for a general increase or decrease in education expenditure. The coefficient of the trend variable is positive but statistically insignificant. But more importantly, our key variable of interest, federal block grants, is positive and statistically significant showing the robustness of our results.

Our article focuses on analyzing the effects of all block grants on education expenditure. For this reason, we include all relevant block grants in the country. However, not all block grants are the same. While CHT and CST are at least nominally meant to support specific expenditure categories, equalization grant is a general block grant that simply supplements recipient provinces general revenue.¹¹ For this reason, one may wonder on the effects of equalization grants alone on education expenditure. As part of the sensitivity analysis, we focus exclusively on equalization grant in columns (5) and (6) of table 4. We use only Formula as an excluded instrument in both cases. In column (5), we ignore CHT and CST grants altogether. In column (6), we include CHT and CST as a control variable by lumping them together with the provincial tax revenue. Again, results in columns (5) and (6) show that equalization grant has stimulative effects on education expenditure even though the numerical magnitude is slightly lower as one may expect.

Some of the previous empirical studies in the literature such as Poterba (1997, 1998) and Schmidt and McCarty (2008) employ log-linear specification as their interests were largely to obtain elasticity estimates. As a robustness check, in column (7), we employ a log-linear specification. These coefficient estimates are not directly comparable to those of table 3 due to differences in specification and resulting differences in interpretation. However, one thing is clear from the results: federal grants stimulate provincial education expenditure suggesting the robustness of our main empirical finding.

Conclusions

Intergovernmental grants have been the cornerstones of many federations around the world. In Canada, federal grants to the provincial governments have been used as powerful policy instruments to encourage provincial governments’ financing of essential public services. While previous studies focus on how these federal grants affect provincial tax policy or total expenditure, the effect of these grants on education spending has been largely ignored. In this article, we investigate the effects of federal block grants on education expenditure using panel data from Canadian provinces over the period 1982 to 2008. We address the common empirical problem of endogeneity of grants by using the allocation formula for equalization grant—a component of the Canadian federal block grant—as our main identification strategy.

We find that block grants have stimulative effects on provincial education expenditure. Our results suggest that a one dollar increase in per capita grants is associated with about Can$0.21 increase in per capita education expenditure. This is roughly equivalent to the average share of education in total provincial spending during the period under consideration. Thus, the stimulative effect of grants on education spending is quite large. The results are robust to various sensitivity checks.

Our results also shed light on other determinants of provincial education expenditure. Consistent with the general concern that many commentators have raised, we also find that aging population has a negative effect on provincial education expenditure. While we find that both left- and right-leaning provincial governments have positive effects on education spending, the magnitude of the effects seems to be larger for the former. The literature on political cycles suggests that governments tend to spend more during election times to maximize their chance of reelection. However, we do not find such effects for provincial education expenditure in our analysis.

Footnotes

Acknowledgments

We would like to thank three anonymous referees and James Alm for their helpful comments and suggestions.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Acosta

Pablo

. 2010. “The ‘Flypaper Effect’ in Presence of Spatial Interdependence: Evidence from Argentinean Municipalities.” Annals of Regional Science 44:453–66.

Alesina

Alberto

Silvia

Ardagna

Trebbi

Francesco

. 2006. “Who Adjusts and When? On the Political Economy of Reforms.” NBER Working Paper 12049, Cambridge, MA, USA.

Angrist

Joshua D.

Lavy

Victor

. 1999. “Using Maimonides’ Rule to Estimates the Effect of Class Size on Scholastic Achievement.” Quarterly Journal of Economics 114:533–75.

Arvate

Paulo R.

Zoghbi

Ana C.

. 2010. “Intergenerational Conflict and Public Education Expenditure When There Is Co-residence between the Elderly and Young.” Economic Education Review 29:1165–75.

Baker

Michael

Payne

Abigail A.

Smart

Michael

. 1999. “An Empirical Study of Matching Grants: The Cap on CAP.” Journal of Public Economics 72:269–88.

Brender

Abdi

Drazen

Allan

. 2013. “Elections, Leaders, and the Composition of Government Spending.” Journal of Public Economics 97:18–31.

Cascio

Elizabeth U.

Gordon

Nora

Reber

Sarah

. 2013. “Local Responses to Federal Grants: Evidence from the Introduction of Title I in the South.” American Economic Journal: Economic Policy 5:126–59.

Case

Anne C.

Rosen

Harvey S.

Hines

James R.

. 1993. “Budget Spillovers and Fiscal Policy Interdependence: Evidence from the States.” Journal of Public Economics 52:285–307.

Cohn

Elchanan

. 1987. “Federal and State Grants to Education: Are They Stimulative or Substitutive?” Economics of Education Review 6:339–44.

10.

Dahlberg

Matz

Mork

Eva

Rattso

Jorn

Agren

Hanna

. 2008. “Using a Discontinuous Grant Rule to Identify the Effect of Grants on Local Taxes and Spending.” Journal of Public Economics 92:2320–35.

11.

Dahlby

Bev

Ferede

Ergete

. 2015. “The Stimulative Effects of Intergovernmental Grants and the Marginal Cost of Public Funds.” International Tax and Public Finance. doi:10.1007/s10797-015-9352-5.

12.

DiMatteo

Livio

DiMatteo

Rosanna

. 1998. “Evidence on the Determinants of Canadian Provincial Government Health Expenditures: 1965–1991.” Journal of Health Economics 17:211–28.

13.

Dollery

Brian E.

Worthington

Andrew C.

. 1996. “The Empirical Analysis of Fiscal Illusion.” Journal of Economic Surveys 10:261–97.

14.

Fernandez

Raquel

Rogerson

Richard

. 2001. “The Determinants of Public Education Expenditures: Long Run Evidence from the States.” Journal of Education Finance 27:567–84.

15.

Fisher

Ronald C.

Papke

Leslie E.

. 2000. “Local Government Responses to Education Grants.” National Tax Journal 53:153–68.

16.

Gamkhar

Shama

Oates

Wallace

. 1996. “Asymmetries in the Response to Increases and Decreases in Intergovernmental Grants: Some Empirical Findings.” National Tax Journal 49:501–12.

17.

Gamkhar

Shama

Shah

Anwar

. 2007. “The Impact of Intergovernmental Transfers: A Synthesis of the Conceptual and Empirical Literature.” In Intergovernmental Fiscal Transfers: Principles and Practice, edited by Boadway

Shah

, 225–258. Washington, DC: The World Bank.

18.

Gordon

Nora

. 2004. “Do Federal Grants Boost School Spending? Evidence from Title I.” Journal of Public Economics 88:1771–92.

19.

Gradstein

Mark

Kaganovich

Michael

. 2004. “Aging Population and Education Finance.” Journal of Public Economics 88:2469–85.

20.

Harris

Amy Rehder

Evans

William N.

Schwab

Robert M.

. 2001. “Education Spending in an Aging America.” Journal of Public Economics 81:449–72.

21.

Inman

Robert P.

2008. The Flypaper Effect. NBER Working Paper 14579, Cambridge, MA, USA.

22.

Kneebone

Ronald

McKenzie

Kenneth

. 2001. “Electoral and Partisan Cycles in Fiscal Policy: An Examination of Canadian Provinces.” International Tax and Public Finance 8:753–74.

23.

Knight

Brian

. 2002. “Endogenous Federal Grants and Crowd-out of State Government Spending: Theory and Evidence from the Federal Highway Aid Program.” American Economic Review 92:71–92.

24.

Ladd

Helen F.

1993. “State Responses to the TRA86 Revenue Windfalls: A New Test of the Flypaper Effect.” Journal of Policy Analysis and Management 12:82–103.

25.

Lutz

Byron F.

2010. “Taxation with Representation: Intergovernmental Grants in a Plebiscite Democracy.” Review of Economics and Statistics 92:316–32.

26.

Oates

Wallace

. 1999. “An Essay on Fiscal Federalism.” Journal of Economic Literature 37:1120–49.

27.

Payne

Abigail A.

2009. “Does Government Funding Change Behavior? Empirical Analysis of Crowd-out.” Tax Policy and the Economy 23:159–84.

28.

Poterba

James M.

1997. “Demographic Structure and the Political Economy of Public Education.” Journal of Policy Analysis and Management 16:48–66.

29.

Poterba

James M.

1998. “Demographic Change, Intergenerational Linkages, and Public Education.” American Economic Review 88:315–20.

30.

Potrafke

Niklas

. 2011. “Public Expenditures on Education and Cultural Affairs in the West German States: Does Government Ideology Influence the Budget Composition?” German Economic Review 12:124–45.

31.

Schmidt

Stephen J.

McCarty

Therese A.

. 2008. “Estimating Permanent and Transitory Income Elasticities of Education Spending from Panel Data.” Journal of Public Economics 92:2132–45.

32.

Singhal

Monica

. 2008. “Special Interest Groups and the Allocation of Public Funds.” Journal of Public Economics 92:548–64.

33.

Tsang

Mun

Levin

Henry H.

. 1983. “The Impact of Intergovernmental Grants on Education Expenditure.” Review of Educational Research 53:329–67.

34.

Wyckoff

Paul G.

1988. “A Bureaucratic Theory of Flypaper Effects.” Journal of Urban Economics 23:115–29.