We study the disclosure incentives for family firms by examining the characteristics of their quarterly earnings forecasts and analysts’ and investors’ responses to them. Forecasts offered before the fiscal quarter-end (guidance) by S&P 500 family firms are generally more specific and timely than those offered by S&P 500 non-family firms, particularly when they convey bad news or confirm analysts’ current expectations. Further, family firm guidance elicits a stronger response from both analysts and investors. While many of these differences largely disappear when the forecasts are offered after the quarter-end but before the earnings announcement itself (preannouncements), family firm preannouncements still tend to be more specific when they contain bad news. These more specific preannouncements also generate a significantly stronger response from analysts. Overall, our results suggest that large, visible family firms use manager-generated earnings forecasts to create a more transparent information environment, and that these forecasts are likely to be most useful in reducing information asymmetry and agency costs when they are issued as guidance.
Key Words: Management earnings forecasts, family firms, preannouncements, earnings warnings.
Data Availability: Data are available from the sources listed in the text.
Family firms are generally defined as companies that are significantly influenced by founding family members or their descendants, through large shareholdings and/or operational control.
Anderson and Reeb (2003a, 2003b) report that family members hold approximately 18% of the equity of the family firms in the S&P 500, on average, and control 45% of the CEO positions. In addition, family members often hold seats on the board of directors or are part of upper-level management in these firms (“Family Inc.”, Business Week, November 10, 2003). The structure inherent in these family firms gives rise to different agency problems than those in firms with much greater separation of ownership and control. Specifically, the family firm structure significantly limits the agency problems that arise from the separation of ownership and control (often referred to as Type I agency problems) while exacerbating those that arise in the conflict between controlling and non-controlling shareholders (often referred to as Type II agency problems, see Ali et al. 2007, Chen et al. 2007, Wang 2006 and Anderson and Reeb 2003a). It is well known that the second type of agency problem can be partially mitigated by frequent and transparent disclosure. However, it is also possible that reputational concerns may arise from the long-term nature of family members’ investment in their firm, mitigating this problem and reducing the need for more frequent and transparent disclosure (Wang 2006).
The purpose of this paper is to add to our understanding of these competing incentives for differential disclosure by examining the characteristics of quarterly earnings forecasts issued by the management of family firms and the response of sell-side analysts and investors to them. Recent accounting research that examines mandatory financial disclosures by family firms suggests that reputational concerns alone may not be sufficient: Characteristics of family firms’ mandatory financial reports are consistent with their being used to mitigate the agency problem between controlling and non-controlling shareholders. More specifically, Ali et al. (2007) and Wang (2006) show that large family firms offer higher quality financial reports as evidenced by lower discretionary accruals, greater ability of earnings to predict cash flows and larger earnings response coefficients. In addition, Ali et al. (2007) find that family firms in the S&P 500 are more likely to voluntarily issue earnings forecasts during periods of earnings declines. However, they also find that family firms are less forthcoming in their disclosures about corporate governance. In a paper that was written concurrently with ours, Chen et al. (2007) study the frequency of voluntary disclosures (earnings and non-earnings forecasts and conference calls) from a larger sample of firms that includes the S&P 500, S&P MidCap 400 and S&P SmallCap 600 in the five years before the enactment of Regulation Fair Disclosure (Reg FD). They also find that family firms are more likely to issue bad-news earnings warnings but overall make fewer forward-looking disclosures than non-family firms, and conclude that their results are consistent with family owners having a longer investment horizon and better monitoring of management, characteristics that obviate the need for greater disclosure.
This paper contributes to the growing literature on the disclosures of family firms by studying one of the most informative and common types of voluntary financial disclosures—the company’s own forecasts of its quarterly earnings per share—and sell-side analysts’ and investors’ responses to them. More specifically, we examine the characteristics of these disclosures (forecast specificity, surprise and accuracy), and the impact they have on important market indicators—professional analysts’ earnings estimates and stock prices. Thus, our analysis is designed to provide additional evidence on the relation between ownership structure and the quality of the firm’s information environment and, in particular, complements the existing empirical evidence on the characteristics and informativenesss of mandatory financial disclosures made by family and non-family firms (Ali et al. 2007 and Wang 2006).
As noted above, we focus on a particular type of voluntary disclosure, management’s forecasts of quarterly earnings per share, and do so for two reasons. First, prior research indicates that these forecasts are highly value-relevant—and more value-relevant than management forecasts of annual earnings per share (Pownall et al. 1993, Baginski and Hassell 1997). As a result, we believe that the quarterly forecasts are particularly well-suited for examining the different incentives family and non-family firms face in their attempts to control Type I and II agency problems, respectively. For example, higher quality forecasting by family firms (in terms of their forecasts being more specific, timely and accurate) is consistent with such firms creating a more transparent information environment and reducing a potentially severe Type II agency problem. Second, we are able to use a non-stock-price measure of the news in these management forecasts in our empirical work, which allows us to more effectively analyze the market’s perception of the differential information content in the forecasts made by family and non-family firms. We also separate our sample of forecasts into guidance (i.e., forecasts made prior to the end of the quarter) and preannouncements (i.e., forecasts made after the quarter ends but before earnings are released). We do this because the forecast horizon associated with preannouncements is very short, sometimes a matter of two or three weeks, and because much of the uncertainty regarding the forthcoming earnings number is resolved by the fiscal quarter end for most, if not all, firms, regardless of whether or not they are controlled by a family. Thus, the Type II agency problem in family firms, if it dominates the Type I agency problem, is more likely to be mitigated through the provision of guidance than preannouncements. This leads us to hypothesize that the characteristics of guidance, but not preannouncements, are systematically related to family-firm status, and that analysts and investors will react differently to the guidance, but not to preannouncements, issued by family firms, holding all else constant.
We test our hypotheses on the quarterly earnings forecasts made between 1998 and 2006 by the family and non-family firms in the S&P 500 index, as identified by Business Week (November 10, 2003) and contained in the First Call Company Issued Guidance (CIG) database. There are two aspects of our sample that should be highlighted. First, our sample firms are among the largest, most stable and most visible in the U.S. As a result, our results may not generalize to smaller, less visible family firms such as those included in Chen et al.’s (2007) sample. Second, our sample period spans the implementation of Reg FD. Thus, we provide evidence that complements the pre-Reg-FD evidence in Chen et al. (2007) and the limited post-Reg-FD evidence in Ali et al. (2007).
The results of our empirical tests generally indicate that the guidance provided by family firms is of higher quality than that provided by non-family firms. In particular, after controlling for other influencing factors, we find that the family firms in our sample provide significantly more specific guidance (in terms of forecast form and narrowness of forecast range) than non-family firms, especially when conveying bad news or offering confirmatory guidance. We also find that family firms use guidance to make smaller average adjustments to the market’s estimate of the upcoming quarterly earnings than non-family firms, especially when conveying bad news. This is consistent with their being more timely in offering corrections to analysts’ estimates. More importantly, we find some evidence of a stronger and quicker response by analysts (as measured by the number of subsequent earnings estimate revisions and the speed with which they occur) to the guidance issued by family firms, and strong evidence of a significantly greater investor response (as measured by announcement-period abnormal stock returns) to the guidance issued by family firms. These findings, taken together, indicate that guidance is more informative and more useful to the market when it is issued by a family firm. They are also consistent with family firms using guidance to create a more transparent information environment, which therefore, complements the finding of higher quality financial reporting by family firms in Ali et al (2007) and Wang (2006).
Consistent with our expectations, we find little evidence of differences in the characteristics of preannouncements issued by family and non-family firms, although there is some (weak) evidence of family-firm preannouncements being more specific when they contain bad news. Also consistent with our expectations, we find no evidence of a differential stock price response to preannouncements made by family and non-family firms, although we do find that analysts response more strongly to family-firm preannouncements, especially when they contain bad news. These results, when considered with the guidance results discussed above, suggest that family firms produce higher quality earnings forecasts than non-family firms, particularly when they are offered as guidance or contain bad news, and that their guidance is more informative and useful to investors and analysts. Thus, our paper provides evidence of family firms using management-generated earnings forecasts to create a more transparent information environment.
Our paper contributes to two bodies of research: the growing literature on disclosures by family firms, as noted before, and the established literature on management forecasts. While our paper is most closely related to Ali et al. (2007), Chen et al. (2007) and Wang (2006), who examine the mandatory financial disclosures of family firms and the frequency of their voluntary disclosures, we also complement Anderson et al.’s (2006) analysis of other dimensions of disclosure transparency. Anderson et al. (2006) find that family firms are significantly more opaque than non-family firms as measured by a summary statistic that captures the effects of trading volume, the bid-ask spread, analyst following and analyst forecast errors. Taken together, the evidence in Anderson et al. (2006) and our paper suggest that certain types of transparent disclosures appear to be better suited than others to mitigating the agency problem that arises between controlling and non-controlling owners.
The literature on management forecasts is more mature and, as a result, guides much of the structure for our analysis. Consequently, we follow prior work by Ajinkya and Gift (1984), Baginski and Hassell (1990, 1997), Bamber and Cheon (1998), Baginski et al. (2002, 2004), Ajinkya et al. (2005) and others, in designing our tests. In a recent paper, Hirst et al. (2007) provide a review of this literature and propose a framework for continued research in this area. They observe that choices concerning the characteristics of management earnings forecasts are not yet well understood and suggest that additional work addressing this issue is needed. Our contribution to the literature on management forecasts is to analyze the differential impact of Type I and Type II agency problems on the characteristics of management earnings forecasts provided by family and non-family firms, including the time of their release, as well as the market and analyst reactions to them. Thus, we add to the initial evidence on the underlying reasons for providing management forecasts in different forms and with different specificity—and on their impact of the stock prices of family and non-family firms. Finally, our results on confirmatory guidance support and extend the results in Clement et al. (2003).
The rest of the paper is organized as follows. In Section 2, we review of the relevant literature and develop hypotheses. In Section 3, we describe our sample and data, and in Section 4, we present the empirical tests. We offer concluding remarks in Section 5.
Family firms are defined in the academic literature as firms in which founders or their descendants exercise control either because they are significant shareholders or because they are part of top management or the board of directors. Not only are family firms common in Europe and Asia (see, for example, LaPorta et al. 1999, Claessens et al, 2000, Gomez-Mejia et al. 2001 and Faccio and Lang 2002), they comprise approximately one-third of the S&P 500 in the U.S. (Anderson and Reeb 2003a). Further, family members’ ownership stakes are significant: Anderson and Reeb (2003a) report that in the S&P 500, family members hold, on average, 18% of the voting shares in their companies.
A large literature on family firms has recently developed in accounting and finance, much of it focused on the differences in agency problems that arise in family and non-family firms. Of particular interest to us are the agency problems arising from (1) the separation of ownership and control, and (2) the conflict between controlling and non-controlling shareholders. The papers that examine these conflicts generally argue that (1), referred to as the “Type I” agency problem in Ali et al. (2007), is less important for family firms because of the unusually close alignment of owners and management in those firms when compared to non-family firms (e.g., Ali et al. 2007, Chen et al. 2007, Wang (2006). They also argue that the tight linkage between some owners and control in family firms exacerbates (2), referred to as the “Type II” agency problem in Ali et al. (2007), in which family members transfer wealth to themselves to the detriment of other shareholders. As is well known, such agency problems can be partially mitigated by frequent and transparent disclosure, suggesting that family firms are more likely to offer a variety of mandatory and voluntary disclosures whose implications are clearer to market participants. In contrast, Wang (2006) suggests that family firms may not face a more severe Type II agency problem if the long-term nature of their investment is well understood by the market. In essence, he argues that long-term investors are less likely to exploit agency problems for short-term gain—thus, family firms may not need to resort to greater frequency or transparency of disclosures.
Ali et al. (2007) and Wang (2006) empirically test these competing predictions by comparing aspects of the accounting disclosures made by family and non-family firms. Both find that earnings quality is higher for family firms, especially when a founder CEO is in place. Thus, both provide some evidence consistent with family firms mitigating their Type II agency problems—or responding to the demands of the users of financial statements—with higher quality disclosures. More specifically, Ali et al. (2007) document lower discretionary accruals and greater earnings persistence for S&P 500 family firms compared to S&P 500 non-family firms. In addition, they find that the association between earnings and stock returns is higher for the family firms. Similarly, Wang (2006) finds that S&P 500 founding family firms have lower abnormal accruals, greater earnings informativeness and less persistence in transitory loss components in earnings. He extends this analysis by considering the effect of the percentage of common stock owned by family members on the magnitude of the Type II agency problem. Interestingly, he finds that the relation is nonlinear: When founding family ownership is above (approximately) 60%, the quality of the earnings reported by non-family firms exceeds that of family firms. Ali et al. (2007) also provide some evidence inconsistent with family firms mitigating their more severe Type II agency problem through the use of disclosures: They observe that family firms are less forthcoming about their corporate governance practices and that when they employ a dual class share structure, earnings quality is lower relative to when they do not have such a structure.
Another method for testing whether family firms mitigate the potentially more severe Type II agency costs—or respond to financial statement users’ demand for high quality accounting information—through greater frequency and transparency of disclosures is to examine the issuance of management earnings forecasts by family and non-family firms. Complicating this is the litigation argument proposed by Skinner (1994) and Kasznik and Lev (1995) which suggests that the use of earnings warnings will vary positively with the litigation risk that the firm faces, and inversely with the severity of the firm’s Type I agency problem (Ali et al. 2007). However, since the Type II agency problem is expected to be more severe and the Type I agency problem less severe in family firms (Ali et al. 2007), family firms would be expected to provide management forecasts to mitigate both types of agency problems, holding litigation risk constant. The relative severity of the Type II agency problem further suggests that family firms’ earnings forecasts will be of higher quality (i.e., more specific, timely and accurate), and that market participants (e.g., sell-side analysts and investors) will respond more strongly to them.
Ali et al. (2007) provide initial evidence in favor of this hypothesis when they observe that family firms are more likely to provide earnings warnings (i.e., guidance that warns of a forthcoming earnings decline) than non-family firms. In a more recent paper, however, Chen et al. (2007) provide evidence that family firms make fewer voluntary disclosures than non-family firms. They collect ownership and founding family information from several sources to identify family firms in the S&P 1500 and find that family firms are (1) 8.1% less likely to provide management forecasts of all kinds (i.e., annual and quarterly earnings, revenues, cash flows, etc.), and (2) less likely to hold conference calls as well. They also find, however, that family firms are more likely than non-family firms to issue bad-news earnings warnings. Chen et al. (2007) conclude that these results, when considered collectively, indicate that family firms’ owners prefer less disclosure because of their long investment horizon and effective monitoring of managers, but that their concern with reducing litigation costs results in an increased likelihood of bad news earnings warnings.
In this paper, we hope to add to our understanding of the relative importance of the competing incentives studied in previous work by examining (1) the characteristics of management forecasts of quarterly earnings per share (both guidance, which is offered prior to the end of the quarter, and preannouncements, which are offered after quarter-end but before the actual earnings announcement) of family and non-family firms, and (2) the response of sell-side analysts and investors to those forecasts. In particular, we hope to add to our understanding of the disclosure choices of family firms by determining whether their own earnings forecasts are more specific, timely and accurate, consistent with family firms providing higher quality disclosures—and whether those forecasts are viewed as being of higher quality by market participants as measured by their response to the disclosure. We also separate our forecasts into guidance and preannouncements under the assumption that any family-firm effect will be more likely to be observed in guidance because of the longer horizon over which the forecasts can be made. More specifically, in the case of preannouncements, there is a very short forecast horizon (e.g., a few weeks beyond the end of the quarter) and so we do not expect large differences in timeliness of the preannouncements between family and non-family firms. Further, because much of the uncertainty about the earnings numbers is resolved by quarter-end, differences in the specificity of preannouncements between family and non-family firms, if any, are likely to be small. Finally, motives to provide preannouncements are likely to be dominated by the litigation argument proposed by Skinner (1994) and Kasznik and Lev (1995). If this is the case, differences in characteristics of voluntary earnings forecasts, and in market participants’ responses to them, are likely to be concentrated in guidance.
As in prior research, we recognize that because of competing forces, whether the guidance of family firms is of higher quality is an empirical question. Thus, our formal hypotheses regarding guidance are non-directional, as in Chen et al. (2007) and Wang (2006):
H1: The specificity, timeliness and content of earnings guidance is systematically related to whether the firm is classified as a family firm.
H2: Sell-side analysts’ and investors’ responses to earnings guidance is systematically related to whether the issuing firm is classified as a family firm.
Our sample is comprised of 4,130 management quarterly earnings guidance announcements issued between 1998 and 2006 by the family and non-family firms in the S&P 500 as identified by Business Week in its November 10, 2003, issue. Business Week defines a family firm as “…any company where founders or descendants continue to hold positions in top management, on the board, or among the company’s shareholders.” To identify family firms, Business Week relies on the methodology developed by Anderson and Reeb (2003a, 2003b) as well as their advice and the help of Spencer Stuart as they “…examined regulatory filings, company Web sites and corporate histories” to ensure significant family involvement in the company. (For details, see “Defining Family,” Business Week, November 10, 2003, p. 111.) Before proceeding, we want to highlight certain aspects of our sample. First, because the Business Week classification pertains to only S&P 500 firms, the firms in our sample are among the largest, most stable and most profitable companies in the U.S. As a result, our findings might not extend to mid- or small-cap companies. Second, our reliance on the Business Week classification means that we do not form a new sample of family and non-family firms each year. However, as Ali et al. (2007) note, family firm status is sticky, and thus misclassifications due to changing firm status will most likely bias against our finding significant results. Third, Business Week’s classification scheme is designed to identify firms that are controlled by a family without relying on a single proxy for control, such as ownership share. As a result, it captures features of family firms, beyond simply having large blockholders, that are likely to exacerbate Type II agency problems. Fourth, by using Business Week’s classification, which is based on the “standard” developed by Anderson and Reeb, our results are more easily compared to many prior results. Finally, while we recognize that Business Week might not accurately classify every firm, both types of classification errors (i.e., misclassifying firms without significant family control as family firms, and misclassifying firms with significant family control as non-family firms) limit our ability to detect differences in the forecasts of family and non-family firms and therefore bias against our finding significant results.
We form our sample by first gathering all forecasts of quarter-ahead earnings made between 1998 and 2006 by the S&P 500 as of June 2003 from the First Call Company Issued Guidance (CIG) database. We lose 1,994 of the original 7,694 observations because of unavailability of (1) necessary Compustat and CRSP data, (2) actual earnings per share and other analyst forecast data from First Call, and (3) observations with multiple actual earnings per share numbers. After deleting stale forecasts (those made before the prior quarter’s earnings announcement date), we retain all “guidance” observations (forecasts made at the same time as or after the prior earnings announcement and at or before the quarter end, N = 4,332). We trim the sample to mitigate the effect of outliers as follows. First, we eliminate the top and bottom one-half percent of the management forecast errors in each sample, the top and bottom one-half percent of the forecast surprises in each sample, the top and bottom one-half percent of the three-day cumulative abnormal returns in each sample and finally, the top and bottom one-half percent of return volatility ratios in each sample—and retain the union of the remaining observations. (These variables are defined in the Appendix and will be discussed in detail later.) We then eliminate 62 firm quarter observations whose stock price is less than $5 as of the beginning of the quarter. This results in a final sample of 4,130 guidance announcements. One-hundred-and-forty six of the 177 family firms identified by Business Week (82.5%) provide guidance during our sample period as compared to 240 of the 323 non-family firms in the S&P 500 (74.3%). 
Before turning to the empirical analysis, we note for the reader that the management guidance we gather from the CIG database is not split-adjusted whereas the analysts’ estimates and reported earnings per share in the main First Call file are (further, they are rounded to the nearest penny). An I/B/E/S unadjusted data file is available but unfortunately, we would lose a significant number of observations if we were to use it. Consequently, to keep the sample size as large as possible and still allow for comparability, we split-adjust the management guidance from the CIG file using the split-adjustment procedures used for the analysts’ estimates and reported earnings per share in the First Call file.
We present descriptive statistics for the guidance announcements, firm-specific characteristics and variables relating to analysts and stock returns in Table 1. We also include the results of two-sample t-tests and Wilcoxon signed rank sum tests for each variable. As noted before, we provide a list of variables and their definitions in the Appendix.
We begin with forecast characteristic metrics designed to help us understand the differences, if any, in the specificity, timeliness, frequency and content of the earnings forecasts offered by the management of family and non-family firms. We present descriptive statistics first for the form of the forecast (an indicator of specificity) as measured by Forecast Form. As is well known, forecasts in the CIG database take one of several forms, which we code in the following manner: If the forecast is a specific earnings per share number (a point forecast), it is coded as 4; if it is a range of possible earnings per share numbers (a range forecast), it is coded as 3; if it consists of a one-sided directional forecast (either a maximum or minimum forthcoming earnings per share number), it is coded as 2; and if it contains no quantitative information (a qualitative forecast), it is coded as 1. Note that our coding scheme is designed so that a higher value of Forecast Form indicates a more specific forecast. To further examine forecast specificity, we focus next on Forecast Width for range forecasts, which measures the difference between the maximum and minimum earnings per share figures offered in the forecast. (A narrower width indicates a more specific forecast.) In later tests, we include point forecasts as forecasts with a width of zero. To examine forecast timeliness, we use Forecast Horizon which is the number of calendar days from the management forecast date until the end of the quarter. More days in the forecast horizon indicate more timely forecasts. Finally, we form Annual Frequency and Quarterly Frequency variables, which measure the number of annual and quarterly management forecasts for each of our sample firms in the CIG database from 1994 through 2006, scaled by the total number of possible forecasting years (for Annual Frequency) or quarters (for Quarterly Frequency) to date.
The descriptive statistics and statistical tests for Forecast Form provide initial evidence consistent with family firms issuing significantly more specific guidance than non-family firms. In particular, Forecast Form has slightly higher numerical values, on average, for family firms (p = .028, using the Wilcoxon test). To further explore the potential differences, we examine the frequency distributions of the forms that guidance takes, as presented in Figure 1. As is obvious from the figure, range forecasts are by far the most common form of guidance for both family and non-family firms, making up nearly two-thirds of all guidance in our sample. Further, both family and non-family firms offer approximately 89% of their guidance as point or range forecasts. However, family firms offer relatively more of the more specific point forecasts (28% versus 23% for non-family firms) and relatively fewer of the less specific range forecasts (61% versus 66% for non-family firms). Conversely, guidance in the form of qualitative statements or minimum/maximum earnings per share numbers is unusual in our sample, regardless of the type of firm examined. The small number of qualitative forecasts in our First Call sample is inconsistent with Hutton et al. (2003) and Miller (2002), who find a substantially larger number of such forecasts when hand-collecting their samples than are included in the First Call database. (Anilowski et al. 2006 also suggest that First Call is more likely to include quantitative forecasts than qualitative ones.) This suggests that our sample is most likely incomplete and most representative when only quantitative forecasts are considered. For these reasons and because many tests require that we restrict attention to point and range forecasts, we will generally focus our discussion on point and range forecasts only.
As just noted, range forecasts are the most common type of guidance in our sample. While it is clear from Figure 1 that non-family firms issue more range forecasts as guidance than family firms, Table 1 indicates that those issued by family firms are significantly narrower, as measured by Forecast Width (p = .000 for both the Wilcoxon and the two-sample t tests). This finding, when considered with the preliminary evidence of greater usage of point forecasts by family firms, suggests that guidance issued by family firms is generally more specific than that issued by non-family firms, consistent with H1.
The next two forecast characteristics that we consider are forecast horizon, an indicator of forecast timeliness, and annual and quarterly forecasting frequency, an indicator of prior forecasting intensity (or, alternatively, an indicator of the establishment of a forecasting history). Table 1 shows that Forecast Horizon is significantly longer for family firms using both the Wilcoxon and the t-tests. Thus, family firms appear to issue their guidance sooner than non-family firms—and thus appear to be more timely disclosers. Turning next to the prior forecasting frequency variables, family firms have a stronger history of providing forecasts (both annual and quarterly are significantly greater for family firms with p-values < .01). Thus, Table 1 provides evidence of family firms being more timely in their issuance of guidance—and of having established stronger, more consistent histories of issuing guidance than non-family firms.
To assess the content of the forecasts issued by family and non-family firms, we next examine the Forecast Error (actual earnings per share minus the guidance number) and Forecast Surprise (the guidance number minus the consensus analyst forecast at the time of the guidance announcement) variables. We note that we calculate forecast errors and surprises for point and range forecasts only, and that we use the midpoint of the range as the forecasted number for the range forecasts. We also note that the forecast error variable is, in some sense, a “forward looking” measure in that it assesses how the management forecast fares relative to the actual earnings per share number that the firm releases in the near future. In contrast, the forecast surprise variable is a “current” variable in that it measures the deviation of management’s forecast from the current market expectation of earnings as measured by the analyst consensus forecast prevailing at the time of the management forecast.
The results of both the Wilcoxon and two-sample t tests indicate that forecast errors do not differ significantly between family and non-family firms—a finding that provides initial evidence of family firms not offering forecasts that are differentially accurate when compared to the actual earnings per share number that they later release. Again, this result is not very surprising since there is little reason to believe that family firms would face differential incentives for such short-term forecasting. However, when forecast surprises are considered, some of the results change. Although the forecast surprises arising from point forecasts in guidance are not significantly different between family and non-family firms, those arising from range-forecast guidance are significantly smaller for family firms (p < .01 for both the Wilcoxon and two-sample t-tests). This suggests that when managers use this most common form of guidance, those affiliated with family firms adjust the market consensus for smaller deviations from their own forecast. These univariate results, when considered with the generally narrower range estimates offered by family firms, provide additional, preliminary support for H1 that family firms offer more specific and timely forecasts than non-family firms. However, we offer this conclusion with the caveat that we have not yet controlled for other factors that affect the form that these forecasts take.
The remaining rows in Table 1 contain descriptive statistics and univariate tests for differences in variables associated with (1) analysts’ earnings estimates and their earnings estimate revision activity surrounding the management forecasts in our sample, (2) the stock price reactions to the management forecasts in our sample, and (3) characteristics of the family and non-family firms in our sample. Many of these variables will serve as control variables, and some will serve as dependent variables, in the multivariate tests in the next section and will be discussed in more detail there. However, before concluding this section, we will briefly discuss the univariate tests associated with the remaining variables that will serve as dependent variables in the multivariate tests.
We use two metrics to measure analyst revision activity following the management forecasts in our sample: (1) Fraction of Estimate Revisions, which is the number of analyst revisions between the guidance and earnings announcement dates scaled by the number of analysts following the firm, and (2) Average Revision Delay, which is the average number of days between the guidance date and the dates of analyst revisions, scaled by the number of days between the guidance and earnings announcement dates. Both metrics provide some indication of sell-side analysts’ perception of the credibility of—or information in—the forecasts issued by management: The first captures the intensity of earnings estimate revision activity after a management forecast, and the second captures the speed with which analysts respond to it. Table 1 indicates no statistical differences between the Fraction of Estimate Revisions for point-forecast guidance issued by family and non-family firms, but significantly shorter Average Revision Delay for those same point forecasts when they are issued by family firms (p-values < .05). Average Revision Delay continues to be weakly smaller for family firms when range-forecast guidance is considered (p = .094 for the Wilcoxon test)—but, in contrast to what we observe for point forecasts, the Fraction of Estimate Revisions is also significantly smaller for the range-forecast guidance of family firms (p-values < .01). While we cannot be confident in interpreting these univariate results because of the lack of controls for other factors that might affect analyst responses, overall, they provide some evidence of analysts responding more quickly to some of the guidance issued by family firms, consistent with H2.
Turning next to the 3-Day CAR variable (the cumulative abnormal stock return over a three-day event window, centered on the guidance announcement date), contrary to H2, we find no strong evidence of differential event-period stock price response for guidance across family and non-family firms—however, once again, the lack of controls for other factors, especially the forecast surprise, is again a problem, and so we return to this variable in the next section when we discuss the multivariate tests.
The final section provides evidence of systematic differences in the characteristics of the family and non-family firms in our sample, all of which might be expected to affect disclosure policy, regardless of family firm status, and therefore are used as control variables in the multivariate tests. In particular, the family firms have significantly lower stock prices and are smaller in terms of total assets than the non-family firms; their average book-to-market ratio is significantly smaller; and they are more often members of industries that face higher litigation risk as measured by Litigation Indicator, defined as 1 or 0 for each firm on the basis of membership in the computer, electronics, pharmaceutical/biotechnology or retail industries (based on Ajinkya et al 2005 and Francis et al. 1994). They do not differ significantly from non-family firms in terms of how frequently they report losses (Loss Indicator).
In this section, we report the results of multivariate analyses, designed to control for firm-specific and other factors that prior research indicates are likely to influence the provision of guidance and market participants’ response to it, so that we can better isolate the impact of family-firm status on voluntary disclosure policy. We perform the analysis (generally) separately for good news, bad news and confirming (neutral news) forecasts.
Specifically, we begin by estimating the following ordered probit model for Forecast Form:
Forecast Form = F (ß1Family Firm Indicatori + ß 2 Size i + ß 3 BTMi + ß 4Litigation Indicatori
+ ß 5 Loss Indicator + ß 6 Reg FD Indicatori + ?i )
where all variables are as defined in the Appendix.
We then examine Forecast Width by estimating the following OLS regression:
Forecast Widthi = 0 + 1Family Firm Indicatori + 2 Size i + 3 BTMi
+ 4 Litigation Indicatori + 5 Loss Indicatori
+ 6 Reg FD Indicatori +i
where, again, all variables are as defined in the Appendix.
We control for firm size per the arguments in Kasznik and Lev (1995), among others; book-to-market per the arguments in Bamber and Cheon (1998); litigation risk per Ajinkya et al. (2005), Kasznik (1995), Francis et al. (1994), and Skinner (1994); and negative earnings in the forecast quarter per the arguments in Ajinkya et al. (2005) based on the findings in Brown (2001) and Hayn (1995). We also include a control for whether the forecast is made after the implementation of Regulation Fair Disclosure (Reg FD) because the decision to offer public guidance—and the analyst and investor reactions to it—are likely to have changed after the ban on selective disclosure (i.e., private guidance) imposed by Reg FD (e.g., Bagnoli et al. 2005, Bailey et al. 2003, Heflin et al. 2002).
The OLS regressions for Forecast Horizon, Forecast Error and Forecast Surprise are similar to those for Forecast Width with one exception. We add controls for annual and quarterly management forecast history (Prior Forecast Frequency Annual and Prior Forecast Accuracy Qtr) since both are likely to affect the timing of the current forecasts and how surprising/informative they are. We add further controls for the Forecast Surprise (absolute value in the analyst response regressions) and whether the forecast is concurrent with the prior quarter’s earnings announcement (Concurrent with Earnings Announcement indicator variable) in the OLS regressions for the analyst response metrics (Fraction of Estimate Revisions and Average Revision Delay) and investor response metric (3-Day CAR). We also add interaction terms between (1) the Family Firm Indicator variable and Forecast Surprise (absolute value in the analyst response regressions), and (2) the Concurrent with Earnings Announcement indicator variable and the prior quarter’s earnings surprise (Prior Qtr Earnings Surprise, absolute value in the analyst response regressions). Both H1 and H2 predict significant coefficients for the Family Firm Indicator variable and the family firm interaction term when applicable.
We present the results of the estimations in Tables 2 through 4, and begin by discussing the ordered probit regressions for Forecast Form and the OLS regressions for Forecast Width in Panels A and B of Table 2. As Panel A of Table 2 shows, family firm status is not associated with Forecast Form when all guidance is considered, but is significantly, positively associated with it when only point and range forecasts are considered (p = .001 for that subsample). Further, that significant relation appears to be driven by neutral and bad forecast news: The coefficient for the Family Firm Indicator variable is positive and significant at the .025 level for neutral news and at the 0.005 level for bad news in the point and range forecast subsample; it is not significant for good news. These results provide relatively strong evidence that family firms in the S&P 500 are more likely to provide more specific guidance than their non-family firm counterparts—especially when conveying bad news or confirming the outstanding consensus estimate (neutral news), consistent with H1.
Although we do not focus on Reg FD in this paper, the coefficients for the Reg FD Indicator variable in the ordered probits in Panel A of Table 2 offer new insight into guidance offered by S&P 500 firms after its enactment. Specifically, when the entire sample is considered, guidance containing either good or bad news appears to be more specific in the post-Reg-FD period, while the opposite appears to be true for neutral guidance (the Reg FD Indicator variable is significantly positive for the first two cases and significantly negative for the third). When only point and range forecasts are considered, however, there is consistent, strong evidence of a post-Reg-FD decline in specificity, regardless of the type of news contained in the forecasts (the Reg FD Indicator variable is always significantly negative). Thus, it appears that in the years after Reg FD, there was a reduction in the use of guidance forms that are most likely to require subsequent “tweaking.” The same ordered probits also provide some evidence that the probability of a more specific guidance announcement is increasing in litigation risk. Specifically, the coefficient for the Litigation Indicator variable is not universally positive and significant in Panel A of Table 2. The inconsistent results for litigation risk are most likely due to the fact that our sample contains forecasts from only S&P 500 firms that are among the largest and most stable firms in the U.S.
We present the OLS results for Forecast Width, a different measure of specificity, for the guidance sample in Panel B of Table 2. Because we include only point and range forecasts in this analysis, only one set of results is presented. (We assign a forecast width of zero to the point forecasts. We calculate the forecast width of the range forecasts as the difference between the maximum and minimum values provided by management.) The regressions provide strong, consistent evidence of family firms offering narrower, and thus more specific, forecasts than non-family firms, regardless of the good, bad or neutral nature of the news. The coefficient for the Family Firm indicator variable is always negative and highly significant (p = .000 in all cases but that of neutral news, and in that case, p = .005). However, because family firms provide more point forecasts as guidance than non-family firms, we also examine the width of range forecasts separately. In this subsample (results not tabulated), the coefficient for the Family Firm Indicator variable continues to be negative and highly significant, regardless of the news. Thus, the evidence continues to support the conclusion that family firms offer more specific guidance than non-family firms in the sense that they offer more point forecasts and narrower range forecasts, consistent with H1.
We turn next to the Forecast Horizon regressions in Panel C of Table 2. In contrast to the strong univariate results, the OLS regressions indicate that only in the case of confirmatory guidance (neutral news) do family firms have consistently longer forecasting horizons than non-family firms. In the case of bad news when all guidance is considered, forecast horizons are significantly shorter for family firms but this result disappears when only point and range forecasts are considered. Thus, it appears that family firms are relatively less timely in providing qualitative guidance when it is considered bad news. Otherwise, once other influencing factors are controlled for, we find no other evidence of S&P 500 firms offering their guidance earlier in the quarter than non-family firms. Interestingly, both the Litigation Indicator and Reg FD Indicator variables are consistently positively associated with Forecast Horizon —a finding that suggests that litigation risk and the restrictions imposed by Reg FD improve the timeliness of guidance provided by S&P 500 firms.
The OLS regression results for Forecast Error and Forecast Surprise are presented in Panel D of Table 2. Again, we present the results for only point and range forecasts because of the need to identify a numerical forecast error or surprise. The regression results for the forecast errors are consistent with the univariate results: There is no significant association between forecast errors and family firm status. This suggests that incentives to provide accurate estimates of the upcoming earnings are similar for family and non-family firms. However, when the forecast surprise in guidance is considered (Panel D of Table 2), we find a significant, positive relation between the surprise and family firm status that is driven by the guidance that contains bad news. (The coefficient for the Family Firm Indicator variable is not significant for good news, and we cannot estimate the regression for forecast surprises of zero, which as noted earlier we define as neutral news.) Since bad news is made up of forecast surprises with negative signs, the significant, positive coefficient for the Family Firm Indicator variable in the bad news regressions indicates that family firms offer guidance that contain bad news for smaller forecast surprises (smaller deviations from the prevailing analyst consensus estimate) than do non-family firms—a finding that is consistent with some of the conclusions drawn from the univariate tests. That is, for bad news, family firms warn when the consensus analyst forecast is closer to their own expectation of the upcoming earnings number than do the non-family firms, and thus provide the information in a more timely manner. Again, considered collectively, family firms offer more specific guidance than non-family firms and, for bad news, offer guidance when the market’s estimate is relatively close to management’s estimate, consistent with H1.
To complete the multivariate analysis, we turn to analyst and investor reactions to the management forecasts in our sample. Tables 3 and 4 contain the OLS regressions for Fraction of Estimate Revisions, Average Revision Delay and 3-Day CARs. Once again, we present the results for only point and range forecasts because of the need to calculate the forecast surprise. We focus first on the regressions designed to explain the fraction of analysts’ earnings estimates that are revised in response to guidance announcements (left half of Table 3). As is evident from the table, when all point and range forecasts are considered (the leftmost column of numbers), the interaction term between the Family Firm Indicator variable and Forecast Surprise (absolute value) is significantly positive (p = .057), suggesting that more analysts revise their estimates in response to a larger forecast surprise when the guidance is issued by a family firm. However, that same coefficient is insignificant in the bad- and good-news subsamples, indicating some weakness or sensitivity in the overall result. Further, the significant, negative coefficient for the Family Firm Indicator variable itself for the entire sample indicates that, on average, analysts make fewer revisions to the guidance issued by family firms. Interestingly, this result appears to be concentrated in the confirmatory guidance subsample—the Family Firm Indicator variable is significantly negative in that subsample but is indistinguishable from zero in the other two subsamples. Thus, so far the regression results provide some evidence of greater estimate revision activity for guidance issued by family firms, consistent with H2. When Average Revision Delay is considered (right half of Table 3), the regression results show an overall quicker response by analysts to a larger forecast surprise when the guidance is issued by a family firm: The interaction term between the Family Firm Indicator variable and Forecast Surprise (absolute value) is significantly negative when all forecasts are considered (p = .051). Once again, however, this same coefficient is not statistically significant in the bad- and good-news subsamples, indicating weakness or sensitivity in the overall result. Finally, in contrast to what we observe in the Fraction of Estimates Revised regressions, the Family Firm Indicator variable itself is never statistically significant in the Average Revision Delay regressions. As a result, we conclude that overall for our sample of S&P 500 firms, the data provide some evidence that analysts respond more strongly and more quickly, in terms of revising their earnings estimates, to guidance when it is issued by a family firm, consistent with H2.
When investor response to guidance is considered, we again find evidence of a stronger response when the forecast is offered by a family firm. In particular, Table 4, which contains event-study (3-Day CAR) regressions for guidance, shows that the interaction term between the Family Firm Indicator variable and Forecast Surprise is always significant and positive; and that the Family Firm Indicator variable itself is generally positive and highly significant (the exception is the point forecast subsample for which it is statistically indistinguishable from zero). While overall these results provide strong indications that guidance offered by family firms is more informative to investors, the fact that the Family Firm Indicator variable is not statistically significant in the point forecast subsample regression is interesting—especially when considered with the fact that many of the control variables that are significant in the other two regressions are not in that same regression. These observations suggest that when guidance is offered in its most specific form (a point estimate), many, but not all, of the differences between the issuing firms lose their influence on the investor response to the news in the guidance. They also suggest that range forecasts issued by a family firm are particularly useful to investors, consistent with their tendency to be more specific than range forecasts issued by non-family firms. Overall, we conclude that the evidence in Table 4 provides strong evidence that the market perceives greater value in earnings forecasts offered by the management of an S&P 500 firm before the fiscal period end, when they come from a family firm.
Before concluding, we want to compare our findings with those of other researchers who examine the management forecasts of family firms. Ali et al. (2007), using a sample of S&P 500 firms as we do, find that family firms are more likely to issue a forecast of the current quarter’s earnings if the firm experiences a decline in earnings in that quarter (that is, if the change in earnings from that of the same quarter in the previous fiscal year is negative). Our findings complement and extend theirs by showing that not only are S&P 500 family firms more frequent forecasters, they also offer higher quality forecasts, especially when those forecasts contain bad or neutral news and are offered as quarterly guidance. We further extend Ali et al.’s analysis by showing that quarterly guidance elicits a stronger response from analysts and investors when it is offered by a family firm. Thus, our findings suggest that S&P 500 family firms provide more specific earnings forecasts, especially when providing guidance, as a means of reducing the potentially more severe agency costs inherent in family firms—or as a means of meeting the demand for more transparent, useful financial disclosures by non-family shareholders.
Our results also complement those in Chen et al. (2007), who use forecasts from a sample of firms that includes the S&P MidCap 400 and SmallCap 600 and focus on the frequency of management forecasts of all kinds and conference calls. As noted earlier, Chen et al. (2007) find that their family firms provide fewer forecasts and host fewer conference calls in general than non-family firms. The one exception to this is that their family firms tend to provide more short-run bad-news earnings forecasts than their non-family firms. Our results, when considered with theirs, suggest that the relative severity or importance of certain agency problems and the costs of reducing them are likely to be different for the large family firms that we examine when compared to the broader set of family firms in Chen et al.’s sample. Our studies together also suggest that the type of voluntary disclosure matters when studying the incentives of family and non-family firms to disclose. In particular, we restrict our sample to quarterly earnings forecasts whereas Chen et al. employ an annual (year t +1) voluntary disclosure indicator that equals one if the firm issued a management forecast of any kind (annual or quarterly about earnings, cash flows, revenue, etc.) and another that equals one if the firm holds at least one conference call in a given year. Thus, differences in our findings may be concentrated in the non-earnings subset of forecasts, a result worthy of additional analysis.
Family firms have a unique structure in that founding family members or their descendants have significant control either because they hold significant shares of the firm’s stock or because they are actively involved in management or, at a minimum, sit on the firm’s board of directors. Academic interest in this unique structure has resulted in a significant body of research, much of it focused on the consequences of the differences in agency problems that arise in family firms and firms with a more severe separation of ownership and control problem. In this paper, we study the quarterly earnings forecasts issued by the management of large (S&P 500) family and non-family firms between 1998 and 2006. Our purpose is to determine whether the quality and timeliness of these voluntary disclosures, as well as the market response to them, are consistent with family firms using them to reduce the relatively severe agency costs arising from the conflict between controlling and non-controlling shareholders inherent in family firms—or whether the longer investment horizon of family members and the close alignment of owner and manager interests in such firms mitigates the need to differentiate themselves in terms of disclosures (Ali et al. 2007, Chen et al. 2007, Wang 2006, Jensen and Meckling 1976).
We divide our sample of forecasts into those released before the end of the quarter, which we refer to as guidance, and those released after the end of the quarter but before the earnings announcements, which we refer to as preannouncements. We expect differences in quality (as measured by forecast specificity, timeliness and accuracy) and analysts’ and investors’ responses between forecasts issued by family and non-family firms, if any exist, to be concentrated in the guidance sample. We do not expect differences for preannouncements because they are offered during the short period of time between the end of the fiscal quarter and the earnings announcement date—and because most of the uncertainty regarding the forthcoming earnings number has been resolved by the end of the quarter, regardless of whether the firm is controlled by a family or not.
Our results can be summarized as follows. The family firms in our sample provide significantly more specific guidance in terms of forecast form and forecast width for point and range forecasts, especially when offering bad or confirmatory news. They also tend to offer guidance for a smaller deviation of their expectations from the outstanding consensus analyst forecast, especially when offering bad news. Further, we find some evidence of analysts reacting more strongly and quickly to family firm guidance, and strong evidence of investors reacting more strongly to it. These findings, taken together, are consistent with large, visible family firms voluntarily providing higher quality disclosures to the investing public and provide additional support for the conclusion drawn in Ali et al. (2007) and Wang (2006) that family firms provide higher quality financial disclosures. Consistent with our expectations, we do not observe the same difference in quality in preannouncements, although we do find that analysts react more strongly to preannouncements containing bad news when they are offered by a family firm. Overall, our paper provides additional evidence of family firms using financial disclosures to create a more transparent information environment. However, the evidence also shows that the voluntary disclosures that we examine are most differentially informative when they are made before the fiscal quarter end.
Forecast Form (a higher value indicates a more specific forecast form) Point forecast, coded 4
Range forecast, coded 3
One-sided, maximum or minimum forecast, coded 2
Qualitative forecast – no numerical guidance, coded 1
Forecast Horizon: The number of calendar days from the forecast date to the end of the quarter (fiscal period end).
Forecast Width: The difference between the maximum and minimum values in the range forecast.
Prior Forecast Frequency (Annual): The number of annual management forecasts captured by First Call since the database’s inception in January 1994, scaled by the total number of possible forecasting periods to date.
Prior Forecast Frequency (Quarterly): The number of quarterly management forecasts captured by First Call since the database’s inception in January 1994, scaled by the total number of possible forecasting periods to date.
Forecast Error: The realized EPS less the management forecast. For point forecasts, the management forecast is the earnings per share specified by management. For range forecasts, we use the midpoint of the range as the management forecast.
Forecast Surprise: The management forecast less the mean analyst estimate on the date of the management forecast. For point forecasts, the management forecast is the earnings per share specified by management. For range forecasts, we use the midpoint of the range as the management forecast.
Good, Bad, and Neutral News: Classification for point and range estimates is based on the sign of the Forecast Surprise, the difference between the management forecast and the mean analyst earnings estimate on the date of the management forecast. Classification for the one-sided directional forecasts is based on the sign of the difference between the minimum or maximum earnings per share specified in those forecasts and the mean analyst earnings estimate on the date of the management forecast. Classification for qualitative forecasts is based on First Call’s classification of the news as good, bad or neutral.
Reg FD Indicator: An indicator variable that equals one when the management forecast is released after the effective date of Regulation Fair Disclosure (October 2000) and zero otherwise.
Analyst Following: The number of analysts following the firm in the current quarter.
Fraction of Estimate Revisions: The number of analyst estimate revisions either between the date of the management forecast and the earnings announcement date or between the two management forecast dates, scaled by the number of analysts following the firm in the current quarter.
Magnitude of Estimate Revision: The absolute value of the difference between the analyst’s earnings estimate immediately before and after the management forecast.
Average Revision Delay: The average number of days between the date of the management forecast and the dates of analyst forecast revisions, scaled by the number of days between the management forecast date and the earnings announcement date.
3-Day CAR: The cumulative abnormal (market-adjusted) return over one day before and after the management forecast date.
Size: Total assets as of the beginning of the current forecast quarter.
BTM: The book to market ratio as of the beginning of the current forecast quarter.
Litigation Indicator: An indicator variable that equals one if the firm is in one of the following industries: pharmaceutical/ biotechnology (SIC codes 2833-2836, 8731-8734), computer (3570-3577, 7370-7374), electronics (3600-3674), or retail (5200-5961), and zero otherwise.
Loss Indicator: An indicator variable that equals one when the firm reported losses in the current quarter and zero otherwise.
Prior Qtr Earnings Surprise: The difference between the actual earnings per share in the prior quarter and the mean analyst forecast of earnings per share.
Concurrent with Earnings Announcement Indicator: An indicator variable that equals one if the management forecast is issued concurrently with the prior quarter’s earnings announcement, and zero otherwise.
We will send an essay sample to you in 2 Hours. If you need help faster you can always use our custom writing service.Get help with my paper