An experimental test of the crowding out hypothesisAn experimental test of the crowding out hypothesisBackground and motivationExperimental design and procedureExperimental designProcedureResultsNo-tax frameTax frameConclusionAcknowledgementsInstructions (US$18-US$2 no-tax treatment, with changes for the 15/5 treatments underlined and the changes for the tax treatment in bold italics in parentheses)ReferencesAn experimental test of the crowding out hypothesisCatherine C. Eckela,*, Philip J. Grossmanb, Rachel M. JohnstonbaDepartment of Economics, (0316) Virginia Tech Blacksburg, VA 24061, USAbDepartment of Economics, St. Cloud State University St. Cloud, MN 56301, USAReceived 10 April 2003; received in revised form 5 December 2003; accepted 25 May 2004Available online 6 November 2004AbstractWe report the results of laboratory experiments that examine whether third-party contributionscrowd out private giving to charity. Subjects play a single dictator game with a charity as therecipient. The subject chooses his preferred charity from a list. There are four treatmentcombinations: two initial allocations and two frames. Initial allocations are either US$18 for thesubject and US$2 for the charity, or US$15 and US$5, respectively, and the subject is then giventhe opportunity to allocate additional funds if desired. The decision frame is also varied to affectsubjects’ perceptions of the task. In one frame, subjects are simply informed of the initialallocations between themselves and their chosen charity. In the other, subjects are told that theirUS$20 allocation has been taxed, and the amount allocated to their chosen charity. The structureof payoffs is identical in both frames. In the first frame, we see a level of crowding out that isclose to zero, far less than other experimental studies; in the second frame, we observe nearly100% crowding out.D 2004 Elsevier B.V. All rights reserved.Keywords: Crowding out; Third-party contribution; Tax1. An experimental test of the crowding out hypothesisHow is private giving affected by government or third-party provision of a publicgood? Theorists, empiricists, and experimentalists have addressed this question, with no0047-2727/$ - see front matter D 2004 Elsevier B.V. All rights reserved.doi:10.1016/j.jpubeco.2004.05.012* Corresponding author. Tel.: +1 540 231 7707; fax: +1 540 231 5097.E-mail addresses: [email protected] (C.C. Eckel)8 [email protected] (P.J. Grossman).Journal of Public Economics 89 (2005) 1543– 1560www.elsevier.com/locate/econbaseclear resolution. We present evidence that, in laboratory experiments, the extent ofcrowding out may depend on the decision-maker’s perceptions about the source of thefunding. Our design varies the framing of a dictator game, and is based on the procedure inBolton and Katok (1998). Subjects are paired with a charity of their own choosing selectedfrom a list of 10. In one pair of treatments, we use a neutral frame, where third-partysupport for the charity is implemented in the form of a difference in initial allocationsbetween the subject and his chosen charity. This frame allows for fiscal illusion; there is noexplicit indication of the source of funding for the third-party transfer to the charities. Thesecond frame introduces fiscal transparency: The source of the funding for the third-partytransfer is an explicit tax on the subject’s own endowment. This frame approximates theassumptions of theoretical models where agents know the source and target of third-partyfunding. Because the transfer comes from the subject’s own initial endowment, our secondframe represents an extreme form of fiscal transparency. We find no evidence of crowdingout in the first frame, but rather an indication of a small amount of crowding in. In thesecond frame, we report almost complete crowd ing out.In the next section, we briefly review the theoretical and empirical studies. In Section 3,we review in greater detail the previous experimental studies. We then discuss theprocedures of our experiment, Section 4, and our findings, Section 5. We conclude inSection 6.2. Background and motivationTheoretical research suggests that government giving should be neutral in its effect onnet contributions to nonprofit institutions under the condition that no one is taxed morethan his pretax contribution to the charity (Warr, 1982, 1983; Roberts, 1984, 1987;Bergstrom et al., 1986). Under these conditions, the effect of any third-party transfer to thenonprofit can be neutralized by the donor’s response. This conclusion is premised on theassumptions of fiscal transparency (i.e., donors are fully aware that government giving isfinanced by the taxes they pay) and that donors’ benefit from the public good arises solelyfrom the nonprofit’s activity.Government giving need not be neutral in the absence of fiscal transparency, (i.e., whenthere is fiscal illusion). Fiscal illusion results when donors/taxpayers do not understand thesources and opportunity costs of funding for activities that they support. Oates (1988)defines fiscal illusion as b...the notion that the systematic misperception of key fiscalparameters may significantly distort fiscal choices by the electorateQ (p. 65). Suchmisperceptions can result in a public sector that is eith er too small (Downs, 1960) or toolarge (Brennan and Buchanan, 1980 ).1Giving by taxpayers suffering from fiscal illusionneed not be completely crowded out by government giving.Crowding out also may be incomplete if the donor cares not only about totalcontributions, but also about his own contribution. Andreoni (1989, 1990) shows that a1Empirical studies tend to support the conclusion that fiscal illusion contributes to an excessive public sector(see, for example, Dollery and Worthington, 1999; Gemmell et al., 2002; Grossman, 1990; McGillivray andMorrissey, 2001; Mitias and Turnbull, 2001; Oates, 1991; Turnbull, 1998).C.C. Eckel et al. / Journal of Public Economics 89 (2005) 1543–15601544warm glow from giving (imperfect altruism) can lead to incomplete crow ding out.2Wedraw on the models of Bergstrom et al. (1986) and Andreoni (1989, 1990), modifying theirapproach to include the possibility of fiscal illusion.Following these models, let xi=private consumption; G=C+T is total public goodprovision consisting of C=voluntary contributions and T=tax revenue; and gi=ci+atiis theconsumer’s perceived contribution, consisting of voluntary contribution ciand perceivedtax support ti. The parameter