تاثیر سیاست های پولی در حباب های بازار سهام و رفتار تجاری: شواهدی از آزمایشگاه

We investigate the effect of monetary policy on stock market bubbles and trading behavior in experimental asset markets. We introduce the possibility of investing in interest bearing bonds to the widely used laboratory asset market design of Smith et al. (1988). Treatment groups face a variable interest rate policy which depends on asset prices, while control groups are subjected to a constant interest rate. We observe a strong impact of our interest rate policy on liquidity in the stock market but only a small impact on bubbles. However, we find that announcing the possibility of reserve requirements significantly reduces bubbles.

Recently, financial markets have experienced the expansion and burst of major asset price bubbles, for example, the dot-com bubble around 2000 or the housing bubble around 2007. The appropriate response of monetary policy to these bubbles is an issue that is under hot debate among central bankers and monetary economists. The collapse of the dot-com bubble, as an example, led a number of authors to argue that the central banks' monetary policy should not only aim for inflation targets, but also incorporate asset price developments and other financial system stability indicators (Borio and Lowe, 2002 and Renaud, 2003). Other authors are skeptical of enforcing such a monetary policy. Simply raising interest rates in the hope of containing a speculative asset price bubble might result in exactly the economic collapse one was trying to avoid (see, e.g., Assenmacher-Wesche and Gerlach, 2010 and Posen, 2006).1 The theoretical literature remains unclear on the feasibility of using an active interest rate policy to deflate asset market bubbles, in particular of stock markets. Much of this literature focuses on stochastic asset price bubbles (see, e.g., Bernanke and Gertler, 2001, Cecchetti et al., 2002 and Gruen et al., 2005). The bottom line of this literature seems to be that the results hinge on the particular stochastic assumptions regarding the asset price (as well as other shocks that might provide a fundamental explanation for the asset price movements) and, above all, on the information available to the policy maker. The idea of our paper is to extract the essence of the mentioned arguments and test them in a controlled laboratory experiment setting. Advantageously in comparison with empirical research, the information content of monetary policy and investors' expectations can be to some extent controlled, and one can analyze with the use of control experiments how the bubble would have evolved without the policy intervention. We introduce a portfolio alternative to trading in stocks to the classical experimental design of Smith et al. (1988): an interest bearing bond. As a consequence, subjects have to decide how to split their capital between the stock and the bond market. Interest rate policy changes opportunity costs, thus possibly affecting prices. We raise the interest rate in treatment groups while control groups face a fixed interest rate. An alternative instrument of monetary policy is reserve requirements. In a separate experiment we thus reduce liquidity directly using reserve requirements. We are interested in several questions: Is it possible to influence bubbles, in particular to reduce them based on one of the measures we propose? Which monetary instruments are best? Does interest rate policy affect portfolio choices of participants and thus liquidity in the stock market? Given that no one has yet studied this research question, we also want to examine how our results compare to those of earlier experiments. Of course, the flip side of being able to isolate certain effects in a laboratory experiment is that it is hard to model all possible consequences, such as, for example, the effects on the economy's real sector. However, our experiments give at the very least new insights into the effectiveness of monetary policy in taming speculative behavior in asset markets and additionally contribute to the research on experimental asset markets. Our main findings are as follows. We observe a significant impact of our interest rate policy on liquidity in the stock market, however, the effect on trading prices is limited. Conversely, we find a strong impact of announcing the possibility of reserve requirements on prices. The actual fulfillment of such reserve requirements does not play a major role. The rest of the paper is organized as follows. Section 2 gives a short overview of research on experimental asset markets and describes our study in relation to previous research. Section 3 describes our experiment in more detail. Section 4 presents the research questions and defines measures which we use in our analysis. Section 5 analyzes the data and Section 6 concludes.

This is the first study to investigate whether active monetary policy can mitigate bubbles in an experimental asset market. Specifically, we analyze the impact of interest rate policy and reserve requirements on stock market prices, liquidity, and trading behavior in the widely used experimental asset market design of Smith et al. (1988). Our results indicate that active interest rate policy has a clear impact on subjects' asset allocations: Subjects considerably reduce their investment for trading and, therefore, liquidity is reduced in the stock market. Reduced liquidity in the stock market in turn showed an effect on stock market prices. This effect of interest rate policy on prices is, however, only significant if one does not account for the fundamental value being modified by interest policy interventions. We conclude that the effect of interest rate policy is limited in reducing bubbles. Apart from that, interest rate policy showed only a small effect on trading volume and stock market volatility. Concerning reserve requirements, however, we observe a strong effect on prices if the central bank announces the possibility of minimum reserve holdings. Whether such reserve requirements are actually imposed is not important. Consequently, simply announcing the possibility of reserve requirements seems to be a much more effective way to reduce stock market bubbles in the laboratory than using active interest rate policy. In our experimental setup at least, central banks thus seem to have some power by simply threatening the market (even without subsequent action). Of course, we do not want to overstate the impact and external validity of our experimental findings. For example, we cannot exclude the possibility that interest rate policy might be more effective if rates are changed more frequently and/or in smaller steps. The fact that our results are generally stronger in later periods gives some indication that this might indeed be the case. Also, given the recent criticism of Kirchler et al. (2012) concerning the Smith et al. (1988) experimental design, it would be interesting to test the robustness of our findings in alternative asset market designs, for example with a flat or increasing fundamental value (FV). The question to what extent an active monetary policy would be effective in these settings is not easy to answer since even the occurrence of bubbles in such settings is not clear. Giusti et al. (2012) show that bubbles only appear if interest rates are fixed at a level such that the FV is decreasing, but not if the FV is increasing. Bostian et al. (2005) are able to create bubbles in a design with a flat FV, but Kirchler et al. (2012) are not. The latter, however, did offer an interest-paying cash account. In Hommes et al., 2005 and Hommes et al., 2008 bubbles occurred in a flat FV setting where participants' task was not to trade but to forecast asset prices (and prices depend on forecasts using a standard asset pricing model). Whether an active interest rate policy will reduce bubbles in such a setting depends on whether coordination between subjects is affected. Given the fact that coordination is observed to be very strong we assume bubbles still to emerge if an active interest rate policy is in place. Also concerning a setting as in Bostian et al. (2005) we expect an active interest rate not to be able to eliminate bubbles completely. This is because our findings show that participants well realize changes in interest rates as they considerably adjust their asset allocations, but, interestingly, our subjects somehow incorporated such changes more or less “accurately” as our sophisticated bubble was largely unaffected by the interest rate policy. Furthermore, the main driver for bubbles seems to be the non-existence of common knowledge, not (only) the confusion about the decreasing FV process (Cheung et al., 2012). However, given the somewhat ambiguous results on bubble existence in flat FV settings, these hypotheses might better be explored in a separate study. Concerning external validity, lab experiments typically cannot incorporate all aspects of the real world. In our case, some impacts of monetary policy on the economy, such as the effect of interest rate policy on GDP or inflation, could not be modeled. Nonetheless, our results should at the very least give some intuition about the possibility of reducing stock market bubbles by monetary policy interventions and how trading behavior, liquidity, and prices are affected. Given our results, it seems hard to calibrate an interest rate policy that deflates bubbles in a timely and controlled manner, even in a modestly complex setting as in our experiment