Research

Publications

Behavioral bias in number processing: Evidence from analysts’ expectations
with Patrick Roger (University of Strasbourg) and Alain Schatt (HEC Lausanne)
Published in Journal of Economic Behavior & Organization 149 [CNRS cat. 2/HCERES], May 2018.

2015 Hillcrest Behavioral Finance Award (Finalist)
Strasbourg Place Financière 2016 Best Paper Award

Abstract:
Research in neuropsychology shows that individuals process small and large numbers differently. Small numbers are processed on a linear scale, while large numbers are processed on a logarithmic scale. In this paper, we show that financial analysts process small prices and large prices differently. When they are optimistic (pessimistic), analysts issue more optimistic (pessimistic) target prices for small price stocks than for large price stocks. Our results are robust when controlling for the usual risk factors such as size, book-to-market, momentum, profitability and investments. They are also robust when we control for firm and analyst characteristics, or for other biases such as the 52-week high bias, the preference for lottery-type stocks and positive skewness, and the analyst tendency to round numbers. Finally, we show that analysts become more optimistic after stock splits. Overall, our results suggest that a deeply-rooted behavioral bias in number processing drives analysts’ return expectations.

The coverage assignments of financial analysts
Published in Accounting and Business Research 48 [CNRS cat. 3/HCERES], January 2018.

Abstract:
Previous studies document that forecast accuracy impacts analyst career outcomes. This paper investigates the influence of forecast accuracy on coverage assignments. I show that brokerage houses reward accurate analysts by assigning them to high-profile firms and penalise analysts exhibiting poor accuracy by assigning them to smaller firms. The coverage of high-profile firms increases the potential for future compensation linked to investment banking and trading commissions. In addition, covering such firms increases analysts’ recognition from buy-side investors, which, in turn, increases the likelihood of obtaining broker votes and votes for the I/I star ranking. Overall, my results indicate that high forecast accuracy leads to increased future compensation.

Idiosyncratic volatility and nominal stock prices: Evidence from approximate factor structures
with Patrick Roger (University of Strasbourg) and Alain Schatt (HEC Lausanne)
Published in Finance Bulletin 1, March 2017.

Abstract:
Approximate factor structures defined by Chamberlain (1983) allow to test whether a given quantitative firm characteristic (the nominal stock price in this paper) is a determinant of the idiosyncratic volatility of stock returns. Our study of 8,000 U.S stocks over the period 1980-2014 shows that small price stocks exhibit a higher idiosyncratic volatility than large price stocks. This relationship is persistent over time and robust to variations in the number of common factors of the approximate factor structure. Moreover, this small price effect does not hide a small-firm effect because it is still valid when we analyze the tercile of large firms. Our result confirms that small price stocks have lottery-type characteristics and, therefore, it is not in line with the efficient market hypothesis.

Reporting errors in the I/B/E/S earnings forecast database: J. Doe vs. J. Doe
Published in Finance Research Letters 20, [CNRS cat. 3/HCERES], February 2017.

Abstract:
This paper provides evidence of systematic errors in the way I/B/E/S reports analyst earnings forecasts. Analysis of the I/B/E/S earnings forecast database over the 1982-2014 period pinpointed a lack of consistency in the identification of financial analysts, a number of whom are consequently (1) identified by several different codes, and (2) erroneously attributed forecasts that were issued by namesakes. The present empirical investigation reveals that over 10% of the analyst codes in the database are subject to such reporting errors. These reporting errors impact the evaluation of analysts’ characteristics, and may bias empirical studies that rely on tracking analysts.

When behavioral portfolio theory meets Markowitz theory
with Marie Pfiffelmann (University of Strasbourg) and Olga Bourachnikova (University of Strasbourg)
Published in Economic Modelling 53 [CNRS cat. 2/HCERES], February 2016.

Abstract:
The Behavioral Portfolio Theory (BPT) developed by Shefrin and Statman (2000) is often set against Markowitz’s (1952) Mean Variance Theory (MVT). In this paper, we compare the asset allocations generated by BPT and MVT without restrictions. Using U.S. stock prices from the CRSP database for the 1995–2011 period, this paper is the first study that empirically determines the BPT optimal portfolio. We show that Shefrin and Statman’s (2000) optimal portfolio is Mean Variance (MV) efficient in more than 70% of cases. However, our results also indicate that the BPT portfolio exhibits a high level of risk, high returns and positively skewed returns. We show that the risk aversion coefficient of the BPT portfolio is up to 10 times lower than the risk aversion degree shown by typical MV investors. Even if the asset allocations may coincide, typical MV investors will not usually select the BPT optimal portfolios. These results underline that MVT and BPT cannot be used interchangeably.

What drives the herding behavior of individual investors?
with Maxime Merli (University of Strasbourg)
Published in Finance 34(3) [CNRS cat. 2/HCERES], December 2013.

Award for the Best Article published in 2013 using EUROFIDAI data
Award for the Best Article published in Finance (the academic journal of the French Finance Association) in 2013

Abstract:
We introduce a new measure of herding that allows for tracking dynamics of individual herding. Using a database of nearly 8 million trades by 87,373 retail investors between 1999 and 2006, we show that individual herding is persistent over time and that past performance and the level of sophistication influence this behavior. We are also able to answer a question that was previously unaddressed in the literature: is herding profitable for investors? Our unique dataset reveals that the investors trading against the crowd tend to exhibit more extreme returns and poorer risk-adjusted performance than the herders.

Working Papers

A re-examination of analysts’ differential target price forecasting ability
with Patrice Fontaine (CNRS – Eurofidai)
Reject & Resubmit in Finance [CNRS cat. 2/HCERES]

Abstract:
Previous studies find persistent differences in target price accuracy and argue that financial analysts possess differential abilities to forecast stock prices. We first show that persistent differences in accuracy across analysts are driven by the characteristics of the firms the analysts follow. Those who cover easier-to-forecast firms appear more accurate.  We then control for firm-characteristics by using alternative measures of target price performance. We introduce a new measure of target price forecast quality and we adapt two relative measures of accuracy from the earnings forecast literature. We then apply five different statistical tests of persistence to each measure of performance. In contrast with the previous literature, we do not find convincing evidence of differential target price forecasting abilities across analysts.

Another law of small numbers: Patterns of trading prices in
experimental markets
with Wael Bousselmi (University of Montpellier), Patrick Roger (University of Strasbourg) and Marc Willinger (University of Montpellier)
Reject & Resubmit in Experimental Economics [CNRS cat. 1/HCERES]

2017 AFSEE Award (French Experimental Economics Association)

Abstract:
Conventional finance models indicate that the magnitude of stock prices should not influence portfolio choices or future returns. This view is contradicted by empirical evidence. In this paper, we report the results of an experiment showing that trading prices, in experimental markets, are processed differently by participants, depending on their magnitude. Our experiment has two consecutive treatments, one in which the fundamental value is a small number (the small price market), and a second treatment, in which the fundamental value is a large number (the large price market). Small price markets exhibit greater overpricing than large price markets. We obtain this result both between-participants and within-participants.  Our findings show that price magnitude influences the way people perceive the distribution of future returns. This result is at odds with standard finance theory but is consistent with: (1) a number of observations in the empirical finance and accounting literature; and, (2) evidence in neuropsychology on the use of different mental scales for small and large numbers.

The e ffect of price magnitude on analysts’ forecasts: evidence from the lab
with Wael Bousselmi (University of Montpellier), Patrick Roger (University of Strasbourg) and Marc Willinger (University of Montpellier)
Submitted to Revue Economique [CNRS cat. 2/HCERES]

Abstract:
Recent research in finance shows that the magnitude of stock prices influences analysts’ price forecasts (Roger et al., 2018). In this paper, we report the results of a novel experiment where some of the participants in a continuous double auction market act as analysts and forecast future prices. Participants engage in two successive markets: one where the fundamental value is a small price and one where the fundamental value is a large price. Our results indicate that analysts’ forecasts are more optimistic in small price markets compared to large price markets. We also find that analysts strongly anchor on past price trends when building their price forecasts. Overall, our findings support the existence of a small price bias.