توزیع مجدد و کارایی بازار: مطالعه تجربی

We study the interaction between competitive markets and income redistribution that reallocates unequal pre-tax market incomes away from the rich to the poor majority. In one setup, participants earn their income by trading in a double auction (DA) with exogenous zero or full redistribution. In another setup, after trading, they vote on redistributive tax policies in a majoritarian election with two competing candidates. This results in virtually full redistribution, even when participants have the opportunity to influence taxes by transferring money to the candidates. We find that the high redistribution reduces trading efficiency, but not as much as predicted if market participants trade randomly. This is because, rather than capitulating to the much lower trading incentives, many participants respond to redistribution by asking and bidding more conservatively in the DA, and in this way help to prevent further efficiency losses.

A major concern of modern democracies is to implement an optimal degree of redistribution that reallocates income away from the well-to-do to the relatively poor. This is a delicate balancing act between the people's taste for equality and the potentially negative consequences of redistribution for economic efficiency. Citizens generate wealth by participating in a variety of different markets. Usually, this wealth is distributed unequally (e.g., due to differences in individual productivities). However, a majority of relatively poor citizens have, in principle, the opportunity to counteract the often dramatic inequality by determining the degree of income redistribution in elections. In the present paper, we study one important aspect of the balancing act between the fundamental conflict of the principles of markets (“one dollar, one vote”) and elections (“one person, one vote”), namely, the effects of redistribution on market trading behavior and efficiency. This is very different to the previous literature in which redistribution influences wealth through various other channels (e.g., Alesina and Giuliano, 2009), such as the people's labor-leisure choice (e.g., Meltzer and Richard, 1981 and Romer, 1975). So far, experimental markets and (redistribution in) elections have been studied in isolation. Many laboratory markets – in particular Smith's (1962) double auction (henceforth DA) – reliably clear at approximate equilibrium prices and quantities, and in this way generate close to the maximum possible wealth, or trading efficiency (for surveys, see Davis and Holt, 1993 and Kagel and Roth, 1995). Similarly, median voter preferences are reflected reliably by outcomes of laboratory two-candidate elections with compulsory majority voting (for a survey, see McKelvey and Ordeshook, 1990). While these literatures have substantially contributed to our understanding of how markets and elections function independently, they shed little light on how they interact and perform when they coexist. Given that trading efficiency can be negatively affected if there are interdependencies between the earnings of traders beyond those implied by market transactions (e.g., Dufwenberg et al., 2011, Goeree and Zhang, 2012 and Rostek and Weretka, 2010), it is important to ask: do lower trading incentives due to redistributive taxation result in lower trading efficiency? And if so, does the poor majority respond by reducing their tax demands in order to prevent a lower tax revenue base? In our experiment, citizens first earn their – unequally distributed – income in a DA. To mimic the prevalent income inequality observed in democracies around the world, we chose the market parameters such that, in equilibrium, the pre-tax income of a majority of citizens is below the average. As this is the first experimental study of the effects of redistribution on trading efficiency, we believe employing the DA is appropriate because it is the predominant market used in laboratory studies (and is often used in actual financial and commodity markets) and poses a challenge to our research question as we know that, without income redistribution, it reliably generates the maximum possible wealth. On the surface, introducing redistribution in a DA in the form of lump sum transfers financed with linear income taxes à la Meltzer and Richard (1981) and Romer (1975) does not change market prices and quantities. Without any redistribution, each transaction gives the two involved traders a purely private return. By contrast, their private return is smaller with redistribution (i.e., the untaxed share of their pre-tax income), but they and everybody else also receive a public return (i.e., a lump sum transfer from the taxed share). 1 Specifically, the marginal trading incentive is non-negative and decreasing in the tax. 2 Therefore, in equilibrium, the maximum trading efficiency is obtained even with full redistribution. Nevertheless, redistribution might affect trading behavior in different ways, and it is not obvious that the effort necessary to achieve high trading efficiency can be maintained. In particular, the lower marginal incentive to trade and bargain over prices might result in unpredicted price volatility, transactions, and efficiency losses. For example, even in games where variations in marginal incentives do not change equilibrium predictions, a bulk of laboratory studies show that people do respond to such variations (e.g., Goeree and Holt, 2001 and McKelvey and Palfrey, 1995). In the DA, this can manifest in careless pricing, which gives rise to substantial inefficient trading. 3 Even more dramatic, if the most efficient traders perceive the degree of taxation as unfair, one could imagine they abstain from market activity altogether. Indeed, as many bargaining experiments suggest, depriving people of their fair share of a surplus can trigger negative emotions that provoke them to destroy their pre-tax income ( Bosman and van Winden, 2002; see also Camerer, 2003). Therefore, it is reasonable to conjecture that too much redistribution decreases trading efficiency. We study whether this is the case by comparing experimental DAs with exogenous zero and full redistribution. We also examine two setups with endogenous redistribution in which, after citizens have earned their unequal incomes in the DA, two candidates compete via redistributive taxes in a majoritarian election. In one of the setups, prior to policymaking, citizens also have the opportunity to influence tax policies by transferring money to the candidates.4 Compared to exogenous redistribution, these setups allow us to draw more general conclusions about the interaction of markets and elections via income redistribution, because now market efficiency and income equality are the outcome of a balancing act between the rich and the poor. First, candidates might attempt to woo the poor majority with drastic redistribution, but if this lowers the trading efficiency then the poor may in fact support more moderate taxes. Second, if transfers to candidates are possible, the rich might attempt to lower their tax burden by compensating them for taking the electoral risk of choosing low tax policies, which can in turn trigger counteractive transfers by the poor.5 If candidates respond to money transfers, the high concentration of pre-tax market income among the well-to-dos might work in their advantage in the rent-seeking race (see also Karabarbounis, 2011). However, as in everyday politics where quid pro quo is generally banned and thus not contract-enforceable, candidates are not bound to return any favors and the success of money transfers depends on their willingness to reciprocate. In addition to gaining better knowledge of the influence of favors on redistribution, the transfer levels give us valuable insights on the preferences for redistribution in our experiment. By keeping our laboratory democracy simple, we obviously abstract from a variety of other phenomena related to income redistribution.6 However, for policymaking to be effective it seems vital to understand how redistribution impacts the basic functioning of coexisting markets and elections. Therefore, our paper can be seen as a first step toward a more basic understanding of the interaction of the two institutions. To this end, laboratory studies are ideal since in the field we generally cannot observe how they work together without them being confounded by various other influences. Naturally, in experiments specific procedures and parameters must be chosen. We want to stress, however, that our choices are nonetheless representative of many of the incentives people face outside of the laboratory. In the following, we discuss some of the important studies that are relevant to ours, starting with the DA and continuing with empirical work on voting and preferences for redistribution. In laboratory DAs prices and quantities quickly and reliably converge toward predicted outcomes, even when participants have minimal information about overall supply and demand. To understand this important result, trading behavior has been modeled and experimentally tested in more detail. For example, Cason and Friedman (1996) perform a systematic analysis of inefficiencies in experimental DAs and compare trading behavior to various theoretical benchmarks (see also Friedman, 1991 and Gjerstad and Dickhaut, 1998; Gode and Sunder, 1993). As benchmarks for trading behavior in the DA with income redistribution, we formulate a simple model that builds upon Gjerstad and Dickhaut's rational traders who form subjective beliefs about the DA environment and we run simulations with Gode and Sunder's zero-intelligence traders. A growing literature investigates more thoroughly the origins of preferences for redistribution (Alesina and Giuliano, 2009). In the present paper, office-seeking candidates have an incentive to woo the poor majority by selecting full redistribution. Of course, such extreme policies are not observed in the field, and a variety of explanations have been offered why median voter preferences for redistribution are more moderate. For example, Alesina and Giuliano (2009) find that the taste for redistribution differs in personal traits (such as age, gender, race, and socioeconomic status), social traits (such as history, culture, and ideology), and fairness concerns (as in Fong, 2001); and see also footnote 6. Moreover, in reality the rich have other options to avoid high taxes such as migrating or indulging in leisure, which are not available in our study and would limit the poor's desire for maximal redistribution. In short, our interest lies in the effects of redistribution on trading behavior and efficiency per se, and how these effects influence the interplay between competitive markets and elections. The laboratory studies of Durante and Putterman (2009), Esarey et al. (2012), and Tyran and Sausgruber (2006) are related to ours in that they too analyze income redistribution derived from citizens' choices. Their results often agree with those in observational studies, supporting the notion that redistributive preferences revealed in experiments are externally valid. In the first two experiments mentioned, self-interest explains much of the redistribution observed, while the third study explores the conditions under which fairness concerns can change voting outcomes (using poor minorities). However, Durante and Putterman (2009) show that preferences for redistribution are sensitive to changes in the setup: for example, there is lower support for redistribution if taxation is associated with costs and deadweight losses and if participants earn their income in a real effort task. In the abovementioned studies, taxes are determined either using the tax proposal of the median voter or a randomly chosen participant. By contrast, in our experiment endogenous redistribution is emerging from candidate competition, voting, and sometimes monetary transfers. To the best of our knowledge, ours is the first study on the mutual impact of redistribution and trading efficiency.7

We experimentally study how income redistribution affects market behavior and, thus, the generation of economic wealth. We use a double auction (DA) where, in equilibrium, market outcomes create a rich minority and a poor majority. When we impose zero redistribution, the DA yields nearly maximum wealth (as in many previous DAs). By contrast, in the novel setup with exogenous full redistribution (i.e., the after-tax incomes will be the same for everyone), wealth drops noticeably, though not as much as predicted for random trading. We attribute the efficiency loss to lower trading incentives caused by high redistribution, which increases the variance of asks and bids and thereby increases the number of transactions involving inefficient units. However, interestingly, participants avoid the large efficiency loss seen with random trading by selling and buying more conservatively with than without redistribution, as predicted by our simple model in which rational players take into account that trading their own units can prevent the trade of more efficient units. To avoid a lower tax revenue base, sellers adjust their WTA above unit costs and buyers adjust their WTP below unit reservation values, which causes the more conservative pricing. In fact, this phenomenon should occur more generally in markets where traders benefit directly from overall market efficiency.30 We also examine endogenous income redistribution by having citizens vote on tax policies in a two-candidate majoritarian election. This allows us to gain insight into the mutual influences of redistribution and market efficiency. Consistent with median voter theories, we find that candidates woo the poor majority with very high taxes and citizens vote according to their pecuniary self-interest. As a result, after-tax incomes are virtually equal. Moreover, the effects we observe with exogenous full redistribution translate to behavior in the DA with endogenous redistribution, although they are weaker since (anticipated) winning taxes are on average somewhat smaller than one. This result does not change when, in addition to voting, citizens can influence tax policies by transferring money to the candidates. Our results show that the “one person, one vote” principle defeats that of “one dollar, one vote”, but at the loss of some trading efficiency. While we focus on the DA and a two-candidate majoritarian election, the advantage of our laboratory design is that, depending on the specific question at hand, it can be easily extended to accommodate different markets, elections, and ways in which they interact with each other. For example, an interesting next step could be to provide the rich with outside options such as tax migration or leisure opportunities, which might lower the poor's support of income redistribution and raise trading efficiency. Other interesting variations could be to study intermediate exogenous tax rates, rich majorities, and costs and reservation values obtained through real effort (as in Cason et al., 2011). Finally, one could facilitate coordination among participants to reach implicit or explicit agreements (as in Großer et al., forthcoming). Overall, our experiment can be seen as a first step to gain better knowledge of how income redistribution influences market behavior and efficiency. More generally, it provides insights into how the coexistence and interaction of markets and elections affect their functioning per se and the generation and distribution of wealth. A basic result of the present paper is that, all else held constant, redistribution negatively impacts trading behavior and efficiency, something that policymakers need to consider when choosing redistributive policies. For example, even if the labor supply is inelastic and therefore increasing taxes will not markedly reduce participation, high taxation could lead to inefficiencies because of price distortions that are driven by noisier trading.