اعتماد بیش از حد فرهنگی و خودتخصیصی جانبدارانه

We use a stock-market game and predictions of examination marks to examine differences between overconfidence and biased self-attribution (BSA) of British and Asian students. Although different overconfidence measures show little correlation, Asians are consistently more overconfident than the British. All are equally prone to BSA.

Behavioural finance models frequently assume the existence of ‘irrational characteristics’ that are regularly found in psychology experiments, such as ambiguity aversion, loss aversion and narrow framing. A developing strand of the finance literature (for example, Benos, 1998, Daniel et al., 1998 and Odean, 1998) is based on another such characteristic, overconfidence. It suggests that if investors are overconfident then financial markets will be characterised by greater trading activity and higher price volatility. Although there is some investigation in the finance literature itself for the existence of overconfidence (for example Malmendier and Tate, 2005, who investigate the existence of managerial overconfidence, or Glaser et al., 2005 who provide evidence on the degree of overconfidence of professional traders and investment bankers), most of the evidence for it comes from the results of experiments reported in the psychology literature. Within this psychology literature a number of different types or traits of overconfidence have been identified and, more recently, the relationships among them explored. Evidence for one type – commonly referred to as miscalibration – comes from studies which ask a participant to answer a question and state how confident they are about their answer, or which ask them to give a range that they feel includes the correct answer (see, for example, Klayman et al., 1999 and Yates et al., 2002). The evidence for its existence and prevalence appears very strong, although some writers, for example Griffin and Varey (1996), Ayton and McClelland (1997), and Soll (1996), have suggested that some of this evidence is contaminated by experimental issues, such as the choice of question type by the researchers. Miscalibration can be seen as one example of a broader type of overconfidence which we shall refer to as ‘stand-alone’ overconfidence, since it is a form of self-appraisal that does not require comparison with others. Another broad type of overconfidence is overconfidence about one's abilities relative to other people. The smaller volume of work on this type explores tendencies such as the ‘better-than-average’ effect, whereby people generally assess themselves as having greater ability than average (see for example Svenson, 1981, Alicke et al., 1995, Larrick et al., 2007 and Glaser and Weber, 2007, who distinguish between miscalibration, too tight volatility estimates, and the better-than-average effect2). In what follows we describe this type of overconfidence as ‘referential’. The degree of correlation between different types of overconfidence is an issue of general interest but especially so for the finance literature. It is common in this literature to define overconfidence as a mistaken belief in the precision of one's information (as in Daniel et al., 1998, for example). Although this appears to relate directly to the miscalibration literature, it implicitly assumes that the precision is being judged relative to the precision of the information held by others. That is, it is a referential judgement. It is therefore important to investigate the degree to which the two categories of overconfidence are related. A number of studies have explored this issue. For example, one argument for a positive correlation between stand-alone and referential overconfidence put forward in Stankov and Crawford (1997), is that they might both be the result of high levels of self-esteem; similarly, Griffin and Brenner (2004), suggest that optimistic overconfidence – the tendency to overestimate the likelihood that one's favoured outcome will occur – might arise from the better-than-average effect, unrealistic optimism and the illusion of control. However, the evidence for such correlations is mixed. Régner et al. (2003), find no correlation between miscalibration and ‘positive illusions’, such as the belief that one is above average or the illusion of control; while Glaser et al. (2005), argue that “some facets of overconfidence that are often assumed to be related (miscalibration and the better-than-others effect) are actually unrelated, which casts doubt on the optimistic overconfidence perspective as a framework for explaining judgemental overconfidence.” (p. 27). One final issue is whether there are significant differences between individuals in the various traits of overconfidence and whether any such differences are stable. The evidence (see, for example, Jonsson and Allwood, 2003 and Glaser et al., 2005) seems to suggest that for any particular trait there are indeed significant differences across individuals which are stable over time, whereas correlations across traits are much weaker. In this paper we use data from an Internet-based stock-market game played by undergraduates, together with their exam results, to address some of these issues. We examine the extent to which the participants exhibit stand-alone and referential overconfidence, and whether there is correlation between these two characteristics. Having found strong evidence for both types of overconfidence we test whether referential overconfidence is stable over time3 and whether men and women differ in these characteristics. Since 16% of our subjects are East Asians (defined as from China, Hong Kong, Malaysia or Taiwan, and henceforth referred to as Asians, for brevity), we also test whether there are cross-cultural differences in overconfidence. Another central aspect of some of the overconfidence-related finance literature that we consider is biased self-attribution (BSA), the tendency to view events that confirm one's views or superior ability as evidence of one's skill, while ascribing events contradicting them to sabotage or noise. For example, Gervais and Odean (2001), model a circular process where BSA feeds overconfidence; and Daniel et al. (1998), model ‘outcome dependent confidence’. While there is strong evidence for the existence of BSA (see Fischoff, 1982, for example), our experiment particularly focuses on the co-existence of BSA and overconfidence, and the extent to which the two are related. While it is plausible that one feeds on the other (one might, for example, describe overconfidence as forward-looking – predicting that one will outperform other people – while biased self-attribution is backward-looking—self-congratulation on having outperformed other people), it is important to note that the two do not have to occur together. For example, overconfidence might be the effect of an overly optimistic outlook, as we suggest below. Our principal findings are that both forms of overconfidence and BSA exist; that the two types of overconfidence do not appear to be correlated; and that there are marked and reasonably stable differences amongst individuals. Whilst there are no cross-cultural or gender differences in the degree of BSA, we find, in line with Yates and Lee (1996), and Yates et al., 1996, Yates et al., 1997 and Yates et al., 2002, that Asians are more confident that the British. We therefore conjecture that high Asian overconfidence is more likely to be due to underweighting of downside risk (i.e. unrealistic optimism, as in Roll, 1986, Kahneman and Lovallo, 1993 and Camerer and Lovallo, 1999, and Heaton, 2002), rather than a higher sense of self-worth.4 This view is supported by the work of Ito (1990), Kleinhesselink and Ross (1991), Weber and Hsee (1998), and Keown (1989),5 and is also consistent with data on the Hofstede, 1980, Uncertainty Avoidance Index6 (UAI): the UAI scores for Hong Kong and China are amongst the lowest in the world, which indicates that these societies are least averse to risk-taking. The paper proceeds as follows. In Section 2 we describe the game, discuss our measures of overconfidence and describe our tests. We present and discuss our results in Section 3, and conclude in Section 4.

Behavioural finance models are increasingly based on the assumption that financial decision makers have irrational characteristics. We have presented evidence on one of these broad characteristics—overconfidence. We have distinguished between two types of overconfidence, namely stand-alone and referential overconfidence. Our evidence comes from the responses of students to a multi-period stock-market game. Our principal findings are these. First, our results strongly confirm the finding of other studies mentioned in Section 1 concerning the existence of overconfidence: we find strong evidence of both stand-alone and referential overconfidence. But we also find evidence to support the view expressed in Glaser et al. (2005), quoted earlier: the different measures of overconfidence do not appear to be correlated. This finding confirms the argument of Glaser and Weber (2007), that it may be misleading for models of financial behaviour to use evidence of one type of overconfidence to justify the existence of another. Our data allow us to test for stability in participants’ overconfidence: our results suggest that for the majority of participants their levels of overconfidence are stable, although there is a sizeable minority whose confidence levels appear to vary over time. We also find significant differences between participants’ levels of overconfidence. In particular, we find strong evidence for both measures that Asians are more overconfident than British participants. Our tests on biased self-attribution suggest that the increased confidence of the Asians is not because of inflated views of their own abilities, since there is no difference between the BSA characteristics of the British and the Asians. One possibility is that Asian overconfidence arises from over-optimism and/or underweighting of downside risk. As mentioned in Section 1, the finance literature suggests that where participants are overconfident financial markets are likely to have higher price volatility and trading activity. So, whatever the reason for it, if our finding that Asians have a greater tendency to overconfidence is correct, then, ceteris paribus, Asian financial markets are more likely to exhibit price volatility and high trading activity.