عواقب اقتصادی مالیات توبین؛ تجزیه و تحلیل تجربی

The effects of a Tobin tax on foreign exchange markets have long been disputed. We present an experiment with currency trading on two markets, where either none, one, or both markets are taxed. Our results confirm the hitherto undisputed issues: a tax reduces trading volume, shifts market share to untaxed markets, and leads to negligible tax revenues if tax havens exist. Concerning the controversial issues we find that (i) volatility effects depend on the existence of tax havens and on market size, (ii) market efficiency decreases in taxed markets when tax havens exist, and (iii) short-term speculation is reduced.

In this paper we present an experimental test of the economic consequences of a Tobin tax. Such a transactions tax on foreign exchange markets was advocated by James Tobin in the early 1970s, and it has been controversial among economists and politicians ever since.1 Of course, the actual implementation of a Tobin tax on real-world foreign exchange markets would resolve the controversies over its alleged consequences on volatility, efficiency, and short-term speculation, to name but a few of the disputed issues. Since a Tobin tax has not been implemented on any real foreign exchange market so far, however, we use the method of experimental economics to assess the effects of a Tobin tax. In the political debate, the Tobin tax has gained popularity as a candidate instrument to fight speculation and stabilize foreign exchange markets. Its intended effects (according to many of its proponents) include a decrease in volatility and an increase in market efficiency. These expected benefits of a Tobin tax have been the reason for the Canadian House of Commons to speak out for a Tobin tax in recent years and for several political proposals in the U.S. to introduce a securities transaction tax (Bloomfield et al., 2009). Although the tax revenues are often downplayed as “side-effects”, expected fiscal benefits obviously also increase the political appeal of a Tobin tax. For instance, when taking over the EU-presidency in 2006, the Austrian Federal Chancellor Wolfgang Schüssel proposed the introduction of a Tobin tax to provide a stable revenue basis for the EU budget. In the academic debate, the Tobin tax has often been linked to the more general issue of how a transaction tax affects financial markets (e.g., Stiglitz, 1989, Summers and Summers, 1989, Schwert and Seguin, 1993, Jones and Seguin, 1997, Subrahmanyam, 1998 and Dow and Rahi, 2000). The economics literature has reached a consensus on several issues such as the negative effects of a Tobin tax on trading volume or market shares (see, e.g., the contributions in Haq et al., 1996 and Weaver et al., 2003). However, some other issues are still disputed, e.g., the impact of a Tobin tax on market efficiency and volatility. Parts of the controversy concerning the latter issues are probably due to different modeling approaches concerning the coverage of the tax, either uniformly across all markets or applying only to a subset of markets. Assuming full coverage of the tax across all markets, Kupiec (1995) relies in his analysis partly on the empirical evidence concerning a transaction tax on stocks in Sweden (Umlauf, 1993). Kupiec then argues that a Tobin tax would increase mispricing, i.e., decrease informational efficiency, and lead to lower liquidity. The latter result is also established in a model with only one market by Subrahmanyam (1998). Palley (1999) presents a microeconomic model with two groups of risk-neutral traders (fundamentalists and noise traders). He shows that noise traders (speculators) cause inefficiencies and higher costs for fundamentalists. Therefore, anything that reduces the volume of noise trading without harming fundamentalists would be considered positive. Palley then argues that although a Tobin tax would hit fundamentalists and noise traders alike with respect to a single transaction, noise traders would be affected more heavily due to their higher trading frequency. As a consequence, a Tobin tax would reduce noise trading and, so he claims, increase market efficiency, contrary to the conclusions by Kupiec (1995). More recent models by Ehrenstein (2002) and Westerhoff (2003) also predict that a Tobin tax will increase informational efficiency by reducing the degree of mispricing (i.e., the difference between market prices and fundamental values). Cipriani and Guarino (2008) focus on the effects of a transaction tax on informational cascades, and hence market efficiency in incorporating information in market prices, in a laboratory financial market. While theory would predict the transaction tax to reduce informational efficiency, they find in the experiment no significant effect, which is due to less irrational behavior in the presence of transaction costs. Summing up the evidence on transaction taxes and market efficiency, it seems fair to say that the literature is still inconclusive. Turning to the effects of a Tobin tax on price volatility we start by noting that Kupiec (1995) does not arrive at a clear-cut prediction for the influence of a Tobin tax on volatility, because a possible reduction in volatility might be wiped out by an increase in liquidity premia. Many other papers (see, e.g., Frankel, 1996, Westerhoff, 2003 and Ehrenstein et al., 2005) expect a decrease in price volatility. However, an empirical study by Aliber et al. (2003) provides conflicting evidence. They consider the Tobin tax as a particular type of transactions costs on currency markets. Therefore, they investigate the impact of the size of transactions costs on trading volume and volatility. Using an innovative approach to derive transactions costs from futures prices, they show that higher transactions costs are associated with higher volatility and lower trading volume on foreign exchange markets. 2 Similar results are presented in Hau (2006). Hence, there is no general agreement on the consequences of a Tobin tax on price volatility, although two recent contributions may be able to resolve the contradictions. Haberer (2006) presents a model with a U-shaped relationship between volatility and market volume. The reduction of market volume due to the introduction of a Tobin tax can then have different consequences for volatility, depending on the relative market volume. Taxing relatively large markets may decrease volatility, whereas a tax on relatively small markets may increase volatility. This will be one of the key findings of our experiment. Pellizzari and Westerhoff (2009) have investigated in computer simulations the impact of market microstructure on the effects of a transaction tax on market volatility. They have found that different trading institutions – either a continuous double auction or a dealership market – yield different effects of a transaction tax on market volatility. While there is no significant effect in a continuous double auction (where the tax reduces liquidity), the introduction of a transaction tax reduces volatility in a dealership market (where abundant liquidity is provided by specialists and the tax crowds out speculative orders). Interestingly, the implications of tax havens have only recently been explicitly modeled. Mannaro et al. (2008) and Westerhoff and Dieci (2006) analyze models with two markets where traders can choose on which market to trade and where a Tobin tax is either implemented on both markets or on just one of them, leaving the other market as a tax haven. Both papers show that introducing the tax on only one market leads to a strong decrease in trading volume on the taxed market. Whereas Mannaro et al. (2008) expect an increase in volatility on the taxed market, Westerhoff and Dieci (2006) claim that volatility decreases on the taxed market, but increases on the untaxed market. The latter paper stresses that the interplay between liquidity and volatility (via the price impact of orders) is difficult to assess in practice, so Westerhoff and Dieci (2006) explicitly call for an experimental analysis of the question. Bloomfield et al. (2009) run a controlled laboratory experiment to study trading behavior on markets when a securities transaction tax (STT) is introduced. They are particularly interested in the effects of a STT on three different types of traders whom they call informed traders, liquidity traders, and noise traders. Their experimental results suggest that a STT leads to less noise trading, which then increases informational efficiency. Market volume is driven down by the tax, whereas market volatility is hardly affected. A limitation of the setting used in Bloomfield et al. (2009) is its restriction to only a single market. In such a setting, it is impossible to examine how one market is affected by a Tobin tax if there are other markets that remain untaxed, i.e., if there are tax havens.3 In our experiment we let subjects trade currencies on two distinct markets. Initially, there is no tax on any of these markets, but then a transactions tax is either introduced on one of the two markets or on both. In order to study whether some effects of the tax persist even after its abolishment – an aspect which has not been explored in the literature so far – we consider also a scenario where the tax is abolished again after its introduction. We let 480 participants trade in a continuous double auction for 18 trading periods. We had 7 different treatments, defined by the sequence of taxing none, one or both markets and by the prevalent tax rate. Our key findings for the case of a unilateral introduction of the tax are that (i) the tax causes a dramatic shift in trading volume to the untaxed market, (ii) tax revenues are negligible, (iii) volatility on the taxed market may decrease or increase, depending on market size, while (iv) volatility on the untaxed market is reduced significantly as a consequence of an increase in liquidity, and (v) market efficiency decreases in the taxed market. If a Tobin tax is introduced simultaneously on both markets, we find that (i) overall trading volume is reduced, (ii) price volatility remains unchanged, and (iii) market efficiency remains unchanged as well. Through an analysis of individual trading patterns we can examine the microfoundations of these aggregate effects on the market level. Taking two different measures of speculation, we find that a Tobin tax reduces speculative trading. Although this was presumably one of the motivations for James Tobin’s proposal (Tobin, 1978 and Eichengreen et al., 1995), it has to be stressed, though, that the effect of this reduction in short-term speculation on volatility can go in either direction. Hence, individual trading patterns may be a misleading indicator for aggregate market effects. The rest of the paper is organized as follows: In Section 2, we present our market model and the experimental design. Sections 3 and 4 report the experimental results. Section 5 concludes the paper by relating our results to previous findings and by discussing the practical implications of our results.

James Tobin has triggered a lively debate about the pros and cons of a transaction tax on foreign exchange markets. In this paper, we have examined in a controlled experiment many of the disputed issues. While one may consider it a big leap from the laboratory to the real world, the experimental approach seems justified because of the lack of empirical evidence from real foreign exchange markets. Of course, the actual implementation of a Tobin tax on real-world foreign exchange markets would ultimately resolve the controversies over the tax’s alleged consequences. In the meantime, experimental studies on a Tobin or transactions tax on foreign exchange markets can provide insights into the consequences of such taxes. Our experimental results confirm many of the theoretically expected effects of a Tobin tax, while at the same time questioning some of its alleged effects. The results on issues where there is broad consensus could have been easily anticipated, of course: Trading volume is negatively affected when all markets are subjected to a Tobin tax, and the tax reduces the number of transactions. The large degree of the shift of trading and transactions from the taxed to the untaxed market – if a tax haven exists – may also be regarded as self-evident, as it is the outcome of massive tax avoidance, leading to almost negligible tax revenues in the presence of tax havens. If these experimental results applied to the real world, this would clearly question many politicians’ expectation of using the Tobin tax as a stable basis for tax revenues. The more interesting issues are those where the results would have been difficult to anticipate. For instance, the interaction of market size and Tobin tax on market activity seems less straightforward. In fact, trading volume and trading activity are much more affected if the Tobin tax is levied on the larger of the two markets. The different positioning of both markets on subjects’ screens has turned out to yield strong differences in market size (with an approximate ratio of 2:1 for the LEFT market when there is no tax), which has made it possible to detect these intricate interaction effects of market size and Tobin tax. The stronger influence of the tax on the larger market seems to be driven by drying up the hitherto very liquid large market. When the tax is introduced on the small market, this relatively illiquid market recovers its rather low level of liquidity when the tax is abolished again. The latter effects has not been dependent on the tax rates applied on our experimental financial markets. One disputed key issue in the debate on the Tobin tax has been market volatility. We find in our experiment no reduction in volatility due to the introduction of a Tobin tax if the tax is encompassing. This result is clearly in conflict with the hopes of the supporters of a Tobin tax. When the tax is introduced unilaterally on the larger (smaller) of the two markets, volatility decreases (increases). Noting that market size in our experiments is closely linked to liquidity, this result confirms theoretical results by Haberer (2004). Furthermore, it is important to note that a Tobin tax on one market has been found to decrease volatility on the untaxed market, which is mainly caused by a shift in trading volume. This effect has not been documented in the literature so far. However, as the results of Pellizzari and Westerhoff (2009) suggest, this finding may depend on the trading institution prevalent on a market. One avenue for future research would be to examine this conjecture by running experimental markets with different trading institutions. Another key issue besides volatility is the question of how a Tobin tax affects market efficiency. Confirming earlier experimental findings of Bloomfield et al. (2009) and Cipriani and Guarino (2008) we observe no impact of the tax on informational efficiency when both markets are taxed. If only one market is taxed, inefficiency on the taxed market increases significantly. We have also been able to document the microeffects of a Tobin tax on individual trading patterns. As intended by the supporters of a Tobin tax, it affects in particular the trading activities of those traders who might be classified as short-term speculators. The frequency of switching between buying and selling is adversely affected by the introduction of a tax, as is the traders’ willingness to issue market orders. This result is clearly in line with earlier findings of Bloomfield et al. (2009) who have shown that a securities transaction tax limits the activities of noise traders. In contrast to Bloomfield et al. (2009) we have been able to document these effects in a broader range of settings that includes both the uniform taxation of all foreign exchange markets and the parallel existence of taxed and untaxed markets. The latter case seems to be the one that is more likely – provided some politicians deem the (economic) consequences of a Tobin tax desirable and implement such a tax, whereas others abstain from it in order to benefit from shifts in trading volume towards tax havens. Our results show that speculators will mostly evade the tax, hence short-term speculation will shift to tax havens rather than vanish. If our results – in particular those in Section 4.5.2 – would hold on real financial markets, then the most important political implication would be that the distortions caused by introducing a Tobin tax not worldwide, but on a subset of markets, may not be undone completely by later on abolishing the tax again. This would suggest that politicians should think twice before they use the financial markets in their countries for a real-time field experiment on the economic consequences of a Tobin tax. Of course, our paper can only be considered a first step into analyzing the effects of a Tobin tax using experimental economics. More research is clearly needed. A next step – also worthy of future investigation – would be to examine the behavioral reactions of actual foreign-currency traders on the introduction of a Tobin tax on a controlled experimental financial market. This would contribute to the issue of the external validity of our experiment. Interestingly, a paper by Haigh and List (2005) shows that professional traders do not better (and partly worse) than university students in an investment task that examines myopic loss aversion. It remains to be seen whether the same holds true for trading behavior in foreign exchange markets. If it does, this would be no good news for the (political) supporters of a Tobin tax.