Who benefits more? The heterogeneous impact of highways on employment growth

ABSTRACT

This paper explores the relationship between the stock of interstate highways on employment growth of counties in Texas between 1983 and 2012. Using the Chernozhukov-Hansen instrumental variables quantile regression (IVQR) method, we examine the heterogeneous impact that highways have on employment growth at different quantiles of the conditional distribution while at the same time controlling for potential endogeneity. The results show that the employment growth effect monotonically increases as one shifts from the lower tail to the upper tail of the distribution. The range is 0.15 to 0.44, with the highest at the 95^th quantile, compared to 0.189 for OLS and 0.213 for 2SLS. A 10 percent increase in interstate highway kilometers in 1983 led to about a 1.5 to 4.4 percentage point increase in county employment over a 29-year period. Our results also indicate that the counties with low initial levels of employment grew faster than those with a high initial level of employment and that this convergence monotonically decreases from the lower tail to the upper tail of the growth distribution.

KEYWORDS:

• Transportation
• highways
• employment growth
• convergence
• instrumental variable quantile regressions
• Texas

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Ideally, one needs to examine the employment growth impact of both the initial level and accumulation rate of highway investments. But, since the kilometers of interstate highways within each MSA do not vary much over time, we do not include their growth rates in our model to examine the accumulation impact separately.

2 Our specification is similar to Duranton and Turner (2012), but our data for employment growth, initial employment, income, and education variables are at the county level. Duranton and Turner (2012) utilized the MSA level data for all variables.

3 We alternatively use the lane kilometers of interstate highways within each MSA, but the results are nearly identical.

4 We added more variables that represent the geographic characteristics (such as an index of the elevation range and heating degree days) and sociodemographic characteristics (such as the employment share of the manufacturing sector, percentage of poor, and a segregation index) of MSAs, but they were highly associated with the variables included in the final model. Thus, we only report the results from a model that is free from the multicolinearity problem.

5 The county-level employment and income data were obtained from the Bureau of Economic Analysis. The percent of adults completing four years of college or higher is from the Census Bureau. The MSA-level variables are from Duranton and Turner (2011, 2012). We thank Gilles Duranton and Matthew Turner for making their data publicly available.

6 We only include pairs of the variables that have a tolerance value greater than 0.10. Some variables (such as incomes for various decades, initial employment, heating degree days, cooling heating degree days) from Duranton and Turner (2012) are almost perfectly collinear.

7 This is common in the growth literature (see, for instance, Barro 1991).

8 See the seminal papers by Koenker and Basset (1978) and Buchinsky (1998) for details.

9 Note that the quantile regression approach has been utilized by various studies (see, for instance, Mello and Perelli (2003); Yasar, Nelson, and Rejesus (2006); Coad (2009); Andini and Andini (2014)). In this, study we use a recently advocated IVQR method that allows us to control for the likely endogeneity of our main variable of interest.

10 Note that for a standard linear quantile regression model of Koenker and Bassett (1978), we have: (${\alpha }_{\mathrm{l}\mathrm{n}HWY}\left(U\right)\equiv 0$ and Z=X.

11 See Wooldridge (2009) for details on the instrumental variable method and these tests.

12 Note that these tests were used by various other studies that utilized quantile regression (see, for instance, Yasar et al. 2006).