"But For" Percentages for Economic Development Incentives: What Percentage Estimates are Plausible Based on the Research Literature?

Abstract

This paper reviews the research literature in the United States on effects of state and local "economic development incentives." Such incentives are tax breaks or grants, provided by state or local governments to individual firms, that are intended to affect firms' decisions about business location, expansion, or job retention. Incentives' benefits versus costs depend greatly on what percentage of incented firms would not have made a particular location/expansion/retention decision "but for" the incentive. Based on a review of 34 estimates of "but for" percentages, from 30 different studies, this paper concludes that typical incentives probably tip somewhere between 2 percent and 25 percent of incented firms toward making a decision favoring the location providing the incentive. In other words, for at least 75 percent of incented firms, the firm would have made a similar decision location/expansion/retention decision without the incentive. Many of the current incentive studies are positively biased toward overestimating the "but for" percentage. Better estimates of "but for" percentages depend on developing data that quantitatively measure diverse changes in incentive policies across comparable areas. JEL Classification Codes: R30, R12, H71

Economic development incentive program deadweight: The role of program design features, firm characteristics, and location

“…Based on a review of studies that adopt these various methods, Bartik (2018aBartik ( , 2018b argues that the plausible range for "but for" effects is about 2%-25% with an average effect of state economic incentives at about 10%-15%, implying average deadweights of 85%-90%. However, empirical findings have varied widely, partly because of limitations of individual research designs, especially small sample sizes, and failure to account for selection bias.…”

“…However, empirical findings have varied widely, partly because of limitations of individual research designs, especially small sample sizes, and failure to account for selection bias. Survey-based analyses tend to find lower deadweight effect due to suspected response biases (Bartik, 2018a).…”

“…It is important to acknowledge that firm surveys are subject to various types of bias. Moreover studies that are based on other empirical methodologies generally find substantially larger deadweight effects (Bartik, 2018a). First, firms may not be honest about the impact of an incentive on their | 67 REPHANN choices/performance due to self-interest in preserving the financial benefit or for having previously inaccurately "certified" with state agency officials that the incentive was needed.…”

“…While some may construe the "but for" effect of 40.1% reported here as high, it is within the range reported by other survey-based studies. Based on an inventory of firm surveys compiled by Bartik (2018a), estimates of the effect range from 6% to 57% with an unweighted average of 34.8% and a weighted average by survey size (when the number of survey observations is reported) of 41.7%. 7 While there may be bias in the overall results (Indeed, Bartik's analysis of "but for" effects using non-survey based ex post empirical studies likely to be less biased finds effects in the range of 2%-25%), the operating assumption of this analysis is that some variation in the responses is nonrandom and can be accounted for by program, firm, or location characteristics.…”

Catching the whale: A comparison of place promotion strategies through the lens of Amazon HQ2

“…Though incentives are highly prevalent, the literature shows substantial doubt regarding whether they alter firm decisions. Bartik (2018) reviews studies investigating decisions and concludes that incentives infrequently alter firm decisions (2% to 25%). A poll of CEOs in North Carolina revealed that talent was the biggest driver of location decisions, with 30% not aware that their firm had received cost-based assistance (Jolley, Lancaster, & Gao, 2015).…”