How a race to the bottom can make you fat

ABSTRACT

This article measures the effect of fiscal competition on obesity rates in the United States through education and health spending. We hypothesize that fiscal competition to attract firms results in lower business tax revenues and higher public infrastructure spending which crowds out education and health spending leading to an increase in obesity rates. We empirically test this hypothesis. We find that there is significant fiscal competition to attract firms. Next, we show that when business tax revenues are lowered and public infrastructure spending favouring businesses increased, public health and education spending declines and obesity rates significantly increase. Thus, fiscal competition significantly contributes to obesity rates through the education and health spending channel.

KEYWORDS:

• Fiscal competition
• tax competition
• education and health spending
• obesity rates

JEL CLASSIFICATION:

• H51
• H71
• I18

Acknowledgements

The authors thank two referees who helped clarify several points made in the article and Benjamin Cowan who provided important comments and suggestions in the development of the study.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ We recognize that there may be other mechanisms by which fiscal competition affects obesity. Our study only shows how these two variables are linked through one potential channel: an education and health spending channel. Other mechanisms may complement or even have a countervailing effect on obesity.

² There is strong empirical evidence showing that local governments within states are in tax competition when setting property taxes (Brueckner and Saavedra 2001) and other measures that control growth of a city (Brueckner 1998). At the state level, there is also evidence of a state’s level of per capita expenditure positively correlating with its neighbouring state’s spending level per capita (Case, Rosen, and Hines 1993). Strategic interaction across states in deciding environmental regulations is also significant (Fredriksson and Millimet 2002). Similar strategic interaction of environmental regulations exists across European countries when examining sulphur and nitrogen oxide emissions (Murdoch, Sandler, and Sargent 1997).

³ To do so, we define a state’s ‘business-friendly fiscal measure’, which indexes a state’s business or corporate taxes and, in some cases, its public infrastructure spending.

⁴ Schlenker and Walker (2016) adopted a similar strategy when investigating the impact of airport networks on health outcomes.

⁵ For reviews of the large literature on the relationship between education and health, see Cutler and Lleras-Muney (2006) and Grossman (2006).

⁶ Note that the calculated short-run and long-run effects of neighbour’s government business-related revenue and spending variable on obesity are only through education and health spending. There may be other channels by which fiscal competition can affect obesity.

⁷ A common example of $w_{ijy}$ is a border weight. It puts the same weight on states that border the given state i and zero weight on non-bordering states.

⁸ Such a panel data model has been commonly used in this literature. See Gadenne (2017) and Cantarero (2005) as examples.

⁹ Note that we also test to see if higher obesity rates affect education and health spending and we find that the effect is insignificant as shown in Table A1 in the Appendix.

¹⁰ The Environmental Kuznets Curve literature linking income to demand for environmental quality has shown that higher demand for environmental quality is correlated with more wealth (see Dinda 2004 for a survey of the literature). In a parallel series of studies, higher average income also increases voting support for environmental referendums (Allen 2003; Salka 2001).

¹¹ The difference GMM was developed by Arellano and Bond (1991) to solve the endogeneity issue in a dynamic panel model by transforming the original level equation to first differences and using twice-differenced lagged dependent variables (or more) as instruments for the lagged difference dependent variable. The system GMM constructs two equations where the first equation is the first-differenced equation in a difference GMM framework and the second equation is the original level equation which the lagged dependent variable is instrumented by once and/or more lagged first-differenced dependent variables (Arellano and Bover, 1995; Blundell and Bond 1998).

¹² Such an assumption can be tested by the Difference-in-Hansen test where the null hypothesis is that all instruments, including the fixed effects, are uncorrelated with the error term. Failure to reject it indicates that instruments are valid and system GMM can be used. We show in our results that we satisfy this additional assumption in the bottom of Table 4 along with other robustness checks in the Appendix.

¹³ We are able to run the entire set of equations simultaneously with the last equation integrated using difference GMM instead. We find very similar results in terms of sign for all our variables. However, the main difference is the significance level of our variables of interest in the third equation where it is not as significant when we opt to run it simultaneously but with difference GMM. Contact authors for results from this regression.

¹⁴ All revenues and expenditures are deflated by the consumer price index using 1982–1984 as the base year.

¹⁵ Note that by using aggregate revenues instead of the tax rates, we account for all the possible business-related fiscal choices a government can make which not only includes the corporate tax rate choice but also the licence fee level, operation fee, occupation fee and other related business fees.

¹⁶ We also construct an alternative measure of business-friendliness that focuses only on the corporate tax rate set by the state – corporate tax revenues per unit of public infrastructure spending and corporate tax revenue per capita. The average corporate tax revenues per unit of public infrastructure spending in our sample is 0.145 which means that for every dollar spent on infrastructure, $0.145 came from corporate taxes. The average corporate tax revenue per capita is $63. The lowest levels are in Nevada, South Dakota, Texas, Washington and Wyoming where corporate taxes are zero while the highest corporate tax revenue per unit of spending and corporate tax revenue per capita are New Hampshire and New York, respectively, with values equal to 0.52 and $203.

¹⁷ To examine the robustness of the weighting matrix, we randomly assign spatial weights for each state and find that the new weighted neighbour’s fiscal policy measure is insignificant in our fiscal competition equation. Thus, our measure of proximity between states appears to have economic content. The results are available from the authors upon request.

¹⁸ Results using corporate revenues are available upon request.