The Ends of Your Means

Abstract

This chapter lays out the first steps to bringing closure to an empirical study. Above all, this involves pursuing “coherence,” in which the study’s findings, economic theory, and vernacular knowledge about the phenomenon of interest coalesce into a logically consistent, unified whole. The pursuit of coherence is multifaceted, and extends to the larger literature to which the study belongs. These ideas inform studies of the demand for cigarettes, zero tolerance drunk driving laws, The Great Moderation, and more.

Food for Thought

• 1. Governments use estimates of the Value of Statistical Life (VSL) to allocate resources that increase the public’s safety. These estimates typically come from wage studies, by extrapolating from the compensating differential associated with more dangerous jobs. If people would sacrifice $1000 in earnings to avoid a 1:10,000 risk of dying on the job, then the VSL = $10 million.

  Unfortunately, this approach has a consistency problem: other job characteristics are not well controlled for, and the labor literature –to the extent it even tries any more–has a difficult time uncovering compensating differentials associated with job disamenities generally (see Smith 1979, the most recent focused review I could find). How should studies of the VSL attempt to deal with this nettlesome problem? What about reviews or assessments of the VSL literature? Take a look at recent studies, meta-analyses , or literature reviews and see what has actually been done.

• 2. Given the facts highlighted in the two real-life scenarios listed below, how should you summarize these literatures in the way that will be most useful for policymakers?

  • (a) The minimum wage in the United States has fluctuated, in real terms, considerably over its lifetime. While the federal minimum is currently on the low side, historically, state minima are gradually being raised to $15/hr. in California and New York, and similar increases are being considered elsewhere. This minimum wage would be far above the historical norm.

  • (b) The evidence on the effectiveness of charter schools is fairly mixed, as are the characteristics of charter schools and the state laws under which they operate. Some states, such as Texas , where charter schools are quite limited in number and scope, are considering legislation that would open up this market substantially.

• 3. The Journal of Economic Surveys is all about summarizing the literature ; you name it, they have an article summarizing it. Several such articles use a common literature-listing technique that I call a “tablelist”–a table that just lists articles, their characteristics and findings, one after the other. Recent examples include Abedifar et al. (2015) on Islamic finance, Becker (2015) on public policy ’s effect on private R&D investment , and van Ours and Williams (2015) on the labor market and health effects of cannabis use. Choose a tablelist from one of these articles or any other, and create a figure that summarizes the literature effectively.

• 4. Anderson et al. (2017) conduct a meta-analysis of studies analyzing the effect of government spending on income inequality. An abbreviated version of their main results (Table 3, column 3) is as follows:

  Y =  − 0.01 + 0.63SE + 0.01DEVELOPED + 0.01PUBLISHED − 0.10OLS+ 0.07TAX + 0.02TRADE + 0.05GOVERNANCE − 0.10INFLATION+ 0.01EDUCATION − 0.10SOCIAL + error

  where Y is the estimate of government spending’s effect on the Gini coefficient, and the independent variables are as follows: the standard error of the relevant coefficient estimate ; dummies for whether an economically developed country is analyzed, whether the study was published, and whether the estimator was OLS ; measures of tax revenue, trade policy , the type of governance, and inflation ; and dummies for whether the spending measure was education spending or general social spending. Significant estimates are in bold.

  Based on these results, is there evidence of publication bias ? Do concerns about the estimation method and potential omitted variables need to be taken seriously? What policy implications can be taken away from these findings?

• 5. My analysis of the “recycling problem” has arisen in Chaps. 3, 8, and 9 of this book, each of which contains an anecdote about one facet of the study. Assemble these facets to show the cohesive nature of that study’s findings.