Security analysts, cash flow forecasts, and turnover

Abstract

We examine the relationship between security analyst turnover and the relative accuracy of their annual earnings and cash flow forecasts. Controlling for self-selection in an analyst’s decision to issue a cash flow forecast, we find that relatively more accurate earnings and cash flow forecasts reduce the probability of turnover. Relative earnings forecast accuracy decreases the probability of turnover more than relative cash flow forecast accuracy. We conduct two cross-sectional tests. We find that relative cash flow forecast accuracy is more important in the analyst’s career outcome when cash flow forecasts are potentially more useful to investors. We find that relative cash flow forecast accuracy is more heavily weighted in the career outcome when the number of other analysts providing cash flow forecasts for the firm is larger. This finding is consistent with economic intuition that relative performance evaluation is more effective when larger groups of individuals are compared.

Introduction

Traditionally, sell-side security analysts have provided earnings forecasts and stock recommendations for the firms they cover. More recently, analysts have included additional pieces of quantitative information, primarily cash flow forecasts and target prices, in their reports (see, e.g., Asquith et al., 2005, Bradshaw and Brown, 2006, Brav and Lehavy, 2003, DeFond and Hung, 2003). We investigate whether and when relative cash flow forecast accuracy affects analysts’ career outcomes (as measured by their turnover).

Examining analysts’ incentives to produce accurate cash flow forecasts is important for at least two reasons. First, understanding the effect of cash flow forecasts on analysts’ job prospects allows us to contribute to the labor economics literature on the way in which principals (brokerage employers) weight multiple output signals (earnings and cash flow forecasts) produced by agents (sell-side analysts). Second, if an analyst is not rewarded for issuing accurate cash flow forecasts, she is unlikely to put effort into generating accurate cash flow forecasts. On the other hand, a brokerage house should compensate an analyst for publishing accurate cash flow forecasts if forecast accuracy helps to differentiate the analyst’s relative ability and effort.

To investigate these issues, we examine whether the relative accuracy of cash flow forecasts is part of an analyst’s performance evaluation criteria. Because analyst compensation data are not publicly available, we use analyst turnover as our observable analyst performance evaluation metric. We believe that analysts are averse to turnover, and investigate whether analysts are more likely to experience turnover if their cash flow forecasts are less accurate than those of their peers, other analysts following the same firm in the same year. We also control for analysts’ relative earnings forecasting accuracy in our analyses, because prior research (e.g., Hong and Kubik, 2003, Mikhail et al., 1999) has documented an inverse relationship between relative earnings forecast accuracy and turnover.

First, we study the association between an analyst’s relative cash flow forecast accuracy and turnover. Call, Chen, and Tong (2009) also provide evidence on the association between analyst turnover and relative cash flow forecast accuracy, but do not control for self-selection in an analyst’s decision to issue cash flow forecasts. Such a self-selection bias can occur because it is unlikely that analysts will choose to release cash flow forecasts randomly. An analyst’s decision to issue a cash flow forecast is conditional on the perceived costs and benefits to her of providing a cash flow forecast in her report. Akin to an omitted variable, self-selection causes biased parameter estimates; the ordinary least squares estimator of the variance–covariance matrix is also biased, undermining statistical inference (Greene, 2005). Therefore, a failure to control for self-selection in the forecasting task may alter the conclusions of prior work. To address these concerns, we model the cash flow forecast issuance decision and turnover using a seemingly unrelated bivariate probit estimation.

Based on a large sample of sell-side analysts who provided earnings and cash flow forecasts (or earnings forecasts alone) over the period 1993 to 2005, we find an inverse relationship between relative cash flow forecast accuracy and turnover. This inverse relationship is weaker than that between relative earnings forecast accuracy and turnover. Thus, a proficiency in forecasting cash flows appears to be important to an analyst’s career outcome, but not as important as proficiency with forecasting earnings. Thus, Call et al.’s (2009) conclusions are not affected after controlling for self-selection. In our cash flow forecast issuance model, we find, similarly to Ertimur and Stubben (2006), that analysts from larger brokerages and analysts who revise their earnings forecasts more frequently are more likely to issue cash flow forecasts. We also find that analysts who follow more firms and analysts who forecast closer to the annual earnings announcement date are more likely to issue cash flow forecasts.

Second, we examine cross-sectional variation in the weight placed on relative cash flow forecast accuracy, a factor which has not been investigated previously. If brokerage houses are rational, they should reward analysts more for accurate cash flow forecasts when: (i) cash flow forecasts are more useful to investors; and (ii) cash flow forecasts are a more precise signal of analyst effort or ability. These cross-sectional analyses yield insights into the efficiency with which brokerage firms evaluate analysts.

To examine the weight placed on the accuracy of cash flow forecasts, conditional on their usefulness to investors, we use DeFond and Hung’s (2003) firm-level determinants of the cash flow forecast issuance decision. DeFond and Hung (2003) find that cash flow forecasts are more helpful in predicting the performances of firms with relatively: larger total accruals, more heterogeneous accounting choices compared to industry norms, higher earnings volatility, higher capital intensity (deflated by sales volume), and poorer financial health (measured by the Altman $Z$-score). They find that analysts are more likely to issue cash flow forecasts in such cases, to aid investors in the interpretation of earnings.

We separate firm-years into above and below median partitions for each of these variables. Our results indicate that relative cash flow forecast accuracy is significantly associated with an analyst’s career outcome in those cases in which DeFond and Hung (2003) conclude that analysts are more likely to issue a cash flow forecast. Further analysis shows that the impact of these factors on the association between accuracy and turnover is statistically significant for four of the five DeFond and Hung (2003) metrics; the results for ALTMAN_Z are not statistically significant. We interpret these cross-sectional findings as brokerage firms placing an increased emphasis on relative cash flow forecast accuracy in the turnover outcome when a cash flow forecast is potentially more useful to investors.

To test whether a greater emphasis is placed on cash flow forecast accuracy in cases in which it is a more precise signal of analyst effort or ability, we investigate the weight placed on relative cash flow forecast accuracy as the number of other analysts issuing cash flow forecasts increases. Following the economic intuition of Gibbons and Murphy (1990), we expect that relative cash flow forecasting accuracy will represent a more precise estimate of analyst effort or ability when the number of other analysts issuing cash flow forecasts increases. Thus, brokerage houses should put more weight on relative cash flow forecasting accuracy when the number of analysts issuing cash flow forecasts increases. Our findings are consistent with this prediction. We interpret this cross-sectional finding as brokerage firms placing an increased emphasis on relative cash flow forecast accuracy when it is more reflective of analyst forecasting ability.

Collectively, our cross-sectional results are consistent with brokerages placing an increased emphasis on the accuracy of analysts’ cash flow forecasts when cash flow forecasts are useful to investors in assessing particular types of firms, and when brokerages are assessing an analyst relative to her peers. This paper proceeds as follows. In Section 2, we develop our hypotheses. Section 3 describes the sample. We provide the primary empirical work in Sections 4 Empirical work: analyst turnover, 5 Empirical work: cross-sectional analyses. Section 6 contains robustness checks and Section 7 concludes.