Ricerca | Progetti

SURVIVORSHIP BIAS WITH OVERCONFIDENT INVESTORS

To make formal inference on data we need to feed models with a continuous stretch of data. Researchers are aware that in commercial time series there can be outliers and missing values (e.g. an exchange halts a security trading due to anomalous order imbalance) and, to face these situations a number of statistical techniques exist trying to interpolate missing values without altering the empirical sample proprieties. Anyway, an "ordinary" data set contains performance related data only with respect to operating firms, that is, we normally do not find data concerning those firms which have defaulted. Even if the data set contains information for defaulted firms, it makes no sense to interpolate the missing values, as the firms have gone out of business. In fact, focusing on survivors means skewing the sample toward those funds which have performed better. As it follows, extending the performances and characteristics of the sample to the whole universe is not the soundest approach. Given these issues, the study will start by defining analytically the two forms of financial bias: survivorship bias and overconfidence bias. The first is a bias due to the data sample, which normally favours the firms which have performed better, therefore overestimating the sample performance. The second is a behavioural bias due to investors, here fund mangers, overestimating the information they possess or their ability to process it. Specifically, The study will be concerned with the overestimation of local information, which might imply home bias on behalf of the investors. The two biases are related as the first prevents a proper explanation of the second. To address overconfidence in domestic information, the study will use an approach based on discriminating geo-focused funds which are managed by local managers and domestic funds which are managed by foreign managers. By merging the two cluster, we devise a preliminary ratio measure to gauge the degree of local investment commitment. The main ideas concerning the methodology to manage survivorship bias are outlined below: 1 Mutual funds are sorted on the basis of monthly returns; 2 Returns are based on funds' market values; 3 Mutual funds are grouped into quantile portfolios; 4 Quantiles are such that the lowest quantile has the lowest return and the highest quantile has the highest return; 5 Portfolios are rebalanced monthly so that whenever a fund disappears weights' are adjusted; This yields a time series of monthly returns on each quantile portfolio. To investigate the relationship between overconfidence and home bias we will use the tracking error of funds quantiles with respect to a domestic benchmark. In general, a low tracking error should imply that funds are passively replicating the market portfolio, which means that there will be a relatively low international diversification. Large tracking error values imply that funds whether are actively focusing on specific domestic stocks or are internationally diversifying. A tracking error both meaningful and high would require therefore further investigations.