To test food truck opponents’ theory, we examined 12 years of data on food trucks and restaurants from the Census Bureau and the Bureau of Labor Statistics. Our study uses a specialized regression to see if a greater number of food trucks in one time period significantly predicts a lower number of restaurants in a later period. If so, this would suggest food truck growth leads to restaurant closures.
This analytical method accomplishes two things. First, it tests the intuition that if more food trucks force restaurants to close, this effect will not occur immediately but instead after a passage of time, one year in our analysis. Our use of a one-year lag, rather than a longer lag (e.g., two years, three years, five years), was informed by media reporting 1
and academic literature 2
suggesting that food trucks’ potential effects on restaurants—if any—would be observed sooner rather than later. We discuss this in greater detail in the appendix. Second and related, our analytical approach clearly identifies food trucks as a cause temporally by having the cause (food truck growth) precede the effect (presumed restaurant closures).
Our analysis improves on work first completed by The Economist magazine. In 2017, the debate about food trucks and restaurants had become so prevalent that The Economist conducted an analysis comparing food truck and restaurant growth using Census Bureau data. The results indicated counties with higher growth in mobile food services also saw higher growth in restaurants and catering businesses. 3
While suggestive, The Economist’s analysis was purely correlational; put differently, its results do not indicate whether changes in the number of food trucks caused changes in the number of restaurants. Our analysis more directly tests food truck critics’ claims that food truck growth causes restaurant decline by exploring whether there is a causal relationship between the number of food trucks and the number of restaurants.
Note, however, that our analysis is not a true experiment in which the number of food trucks can be identified as the single cause of changes in the number of restaurants. 4
Factors we did not or could not measure, such as the financial health of individual businesses, weather patterns or county food truck laws, may help explain changes in the number of restaurants. Yet the type of analysis we used reduces the effects of other potential factors that could confound the results and therefore further highlights the relationship, if any, between food trucks and restaurants.
We also improved on The Economist’s analysis by controlling for factors we could measure that could confound the relationship between food trucks and restaurants. Outside of a true experiment—which again this is not—an examination of the influence of a factor on some outcome could be blurred by a third factor, casting doubt on the extent to which the primary factor of interest actually influences the outcome. In this study, for example, changes in the number of restaurants might be influenced more by changes in the economy over time than by the number of food trucks. One way of addressing this is by statistically controlling for—or removing the influence of—other potential explanatory factors. We did this by controlling for county population size using population estimates drawn from annual Census Bureau data. We also controlled for economic effects by including county-level unemployment rates from the BLS Local Area Unemployment Statistics. This economic control also enabled us to account for the Great Recession, which occurred during the time period covered in this report.
To execute the analysis, we extracted the annual number of mobile food service establishments and full-service restaurant establishments by county (n = 3,133) from the 2005 to 2016 annual Census Bureau County Business Patterns database. This is the same data The Economist used. We also ran the analysis a second time excluding rural counties, which are sparsely populated and may have no food trucks or restaurants—potentially distorting results. The sample size for the second analysis was 1,165 counties. See the appendix for detailed methods and full results.