Navigating Stock Price Crashes by SSRN
H/T The Idea Farm
University of California, Berkeley – Haas School of Business
University of California at Berkeley – Haas School of Business
June 1, 2015
This paper analyzes procedures for forecasting and avoiding stock price crashes. First, we identify the underlying events that cause stock prices to crash. Second, we synthesize previous academic research on the prediction of stock price crashes and construct a parsimonious model for forecasting stock price crashes. Third, we examine how positioning a portfolio to reduce exposure to stocks with high crash risk improves investment performance. Our research should help investors to construct equity portfolios with fewer stock price crashes, higher returns and lower volatility.
Navigating Stock Price Crashes – Introduction
Stock price crashes are dreaded events for active investors. A single stock price crash can erase an otherwise strong quarter of investment performance. Moreover, an active investor with a large position in a stock that suffers a well-publicized crash can suffer a loss of reputational capital. Unfortunately, however, stock prices are quite prone to such crashes. It has long been established that the distribution of stock returns is leptokurtic, meaning that extreme outcomes are more common than for a normal distribution (see Fama, 1965). To put some numbers on this phenomenon, over 10% of stocks have at least one daily return lower than -20% during a typical year.
Despite the significance of stock price crashes, there is little practical guidance to aid investors in avoiding crashes. In this paper, we identify (i) the causes of stock price crashes; (ii) information that can help investors to anticipate and avoid stock price crashes and (iii) the gains to investment performance that result from positioning an investment portfolio to avoid stock price crashes.
We begin by defining and measuring stock prices crashes from the perspective of an investment practitioner. Existing academic research defines stock price crashes relative to the ex post distribution of stock returns. But since the events causing stock price crashes often change other characteristics of the distribution of stock returns, this approach misclassifies some crashes. Instead, we recommend that crashes be defined with respect to the ex ante distribution of stock returns. In other words, we define a stock price crash as a large and abrupt negative stock return relative to the distribution of returns leading up to the crash.
We next examine the events that cause stock prices to crash. While previous research has identified earnings announcements as one common cause of stock price crashes (see Skinner and Sloan, 2001), there is no systematic evidence. Our analysis confirms that earnings announcements are the most common cause of stock price crashes, accounting for around 70% of all crashes. Other common events precipitating stock price crashes include earnings preannouncements and the outcome of clinical trials (for healthcare stocks).
We then turn to the central topic of forecasting stock price crashes. Previous research has identified a number of characteristics that forecast stock price crashes, including abnormally high share turnover, low book-to-market ratio, high short interest, low accounting quality and high growth expectations. We distill this research to identify a parsimonious set of crash predictors.
Finally, we design a practical strategy for avoiding stock price crashes. The strategy not only reduces the incidence of future stock price crashes, but also generates higher future stock returns with lower risk.
Our research design proceeds in two stages. In the first stage, we discuss the definition and measurement of stock price crashes. In the second stage, we describe the variables used to forecast crashes.
Defining and Measuring Stock Price Crashes. A stock price crash is an unusually large and abrupt drop in the price of a stock. Existing academic literature in this area uses two different measures of crashes. Beginning with Chen, Hong and Stein (2001), one line of literature measures realized stock price crashes in terms of the negative skewness in the distribution of daily stock returns, computed using the sample analog of the negative coefficient of skewness (NCSKEW):
where Ri,t denotes the sequence of demeaned daily stock returns to stock i during period t and n is the number of daily stock returns in the period. This measure indicates whether the left tail of the distribution of stock returns is either longer or fatter than the right tail of the distribution. Note that a negative sign is placed in front of the expressions, meaning that a larger positive value implies a larger stock price crash. The logic underlying the use of this measure is that a stock price crash will result in an extreme left-tail outcome. This measure, however, is subject to two limitations. First, a crash is defined as a large negative return (i.e., a long left tail), while negative skewness can also be caused by several less extreme negative returns (i.e., a fat left tail). Second, this measure eliminates stocks that are prone to both crashes and jumps (i.e., large and abrupt increases in stock returns).
See full PDF below.