Firms experiencing declining performance

Transcription

1 Earnings Forecasts of Firms Experiencing Sales Decline: Why So Inaccurate? HUONG N. HIGGINS HUONG N. HIGGINS IS an assdci^itf profes.sor at Worcester Polytechnic Institute, Department of Management, in Worcester, MA. Firms experiencing declining performance are difficult to forecast. In fact, many studies show that earnings forecasts ofpoor-pertbrm^ince ftrnis contain large errors (Kothari et al. [2005J, Brown [2001], Hwang et al. [19961, and Elgers and Lo [1994]). Why are these forecasts so inaccurate? This article discusses one possible answer to that question. A testable idea is that analysts do not correctly assess the impact of inflexible costs during sales declines, particularly when the decline is not well anticipated, resulting in large forecast errors. Essentially, firms with inflexible costs are inherently more difficult to forecast when sales fluctuate there are high swings in earnings because costs do not move to offset sales variability. In particular, sales declines are less predictable than sales increases, because many firms withhold news of bad performance [Kothari et al. 2005], so it is more difficult for analysts to produce accurate forecasts of sales-decline firms. When firms cannot nimbly shed costs in keeping with dropping sales, their earnings are hit hard. When these shocks to earnings are not anticipated, large forecast errors ensue. To investigate this idea, this article examines the Quarter 4 forecast errors of a large sample of U.S. firms in conjunction with their cost structures and control variables during This period is advantageous for this investigation because during these years. market participants were generally optimistic and did not anticipate performance dechnes well. The inflexibility of cost structure is assessed via high operating and financial leverage, determined by costs that are unavoidable and larger than the normal expectation for the considered industry. Industry benchmarks arc used because analysts typically have industry-specific training. Consistent with the investigated idea, the analyses show that forecast errors are the most optimistic and inaccurate for firms that experience sales decline and have large inflexible costs. This article addresses an important subject for the investment community, because earning forecasts are a key output hy financial analysts, and forecast accuracy is a yardstick of forecast ahility. This article serves the investment conmiunity in two ways. One, it reminds invesmient managers to evaluate analysts' forecast ahility in accordance with the particular challenges they face. Because some firnis are inherently more difficult to forecast than others, fair evaluations should consider the inherent difficulty due to different cost behaviors in revenue fluctuations. Two, by pointing out cost factors affecting forecast errors, this article helps analysts better anticipate large shocks to earnings to improve their forecast performance. Specifically, analysts should be keenly aware of a firm's true fixed cost structure to assess its impact on earnings correctly when revenues fluctuate. 26 is FORECASTS OF PIRMS EXI'ERIENCING SALES DHCLINE: WHY SO INACCUKATE? SPRING 2008

2 WHY ARE FORECASTS OF POOR-PERFORMANCE FIRMS SO INACCURATE? This article investigates whether inflexible costs are associated with large forecast errors during sales decline. The flexibility of the firms cost structure means the firm's ability to shed costs nimbly in keeping with fluctuating sales to safeguard earnings. Firms with flexihle cost structures can shed costs easily when sales decrease, leaving earnings unaflected. On the contrary, firms with inflexible cost structures cannot avoid their usual large charges, and consequently their earnings are hit hard. There are two important examples of inflexible costs: operating leverage and financial leverage. Operating leverage refers to the division between the fixed versus variable components of a firm's cost structure. High fixed costs, such as plant depreciation, are not avoidable, and therefore cause strong shocks to earnings when sales drop (Bodie et al. [20051; Lev [ 1983]). Financial leverage refers to a firm's borrowing. Because interest payments on debt must be paid regardless, they are a special fixed cost that also increases the sensitivity of earnings to falling sales (Bodie et al. [2005]). If analysts do not anticipate the effect of operating and financial leverage correctly, forecasts of earnings will contain large errors. INVESTIGATION DATA For this investigation, this article uses data ot U.S. firms during the high-flying period ofl , when market decline was generally not well anticipated. This often-dubbed bubble time frame is most advantageous for this investigation because market participants, including analysts, tended to neglect bad news, making it easier to detect the impact of unanticipated declines. Market participants have been more vigilant since 1999 as a result of the many accounting scandals, and data for periods before 1992 are less available. The main data requirements are earnings forecasts, actual earnings, and operating and financial leverage. In addition, because earnings growth and volatility typically affect forecast errors, these variables are also necessary as controls. The focus is on Quarter 4 (Q4), because the quarterly frequency is the most important frequency of reporting in the U.S. and Q4 has the largest forecast errors (Collins et al. [1984]). Further, earnings and economic growth are often quoted by the financial press as one-year, quarter-to-quarter changes. Forecast and earnings data are retrieved ftom the Thomson Financial I/B/E/S Summary file, and data to determine sales decline, operating leverage, financial leverage, and other controls are retrieved from Worldscope. The above data requirements result in a final sample of 5,180 Q4 forecasts as described in Exhibit 1. Exhibit 2 presents the summary description of the sample. As shown, the typical sample firm is characterized by an average of 19.92% Q4-to-Q4 sales increase, USS457, million in market capitalization, and 7.02 analysts following the firm. Earnings Volatility; an index computed by I/B/E/S as the mean absolute percentage difference between actual reported earnings per share and a five-year historical EPS growth trend line, averages 26.65%. Earnings Growth, an index computed also by 1/B/E/S as the average annualized earnings per share growth over the past five years, averages 20.48%. EPS magnitude averages SO.36, and only 3% of firms experience negative earnings in this quarter. Operating leverage, measured as the ratio between fixed assets and income from continuing operations, averages Financial leverage, measured as the ratio between interest expense and income from continuing operations, averages ARE POOR-PERFORMANCE FIRMS REALLY FORECASTED WITH LARGE OPTIMISTIC ERRORS? The following compares the forecast error of firms having one-year Q4-to-Q4 sales decline versus those having one-year Q4-to-Q4 sales increase. Following prior studies, forecast error is measured as the difference between actual and forecasted earnings scaled by actual earnings, and winsorized at +2 and -2. A non-zero value means the forecast is not accurate, and a negative value means the forecast is optimistic. For visual illustration, Exhibit 3 is a bar graph depicting earnings forecast errors of sample firms over the examination period. The depths (or negative heights) of the hars indicate that earnings forecast errors are negative on average i.e., the forecasts are optimistic, as actual earnings are lower than analysts' forecasts. The mean earnings forecast errors of sales-decline firms are -44%, -23%, -24%, -11%, -15%, -21%, -10%. and -18%, depicted by dark-colored bars from 1992 through 1999, respectively. The mean earnings forecast errors of sales-increase firms are respectively 9%, 7%, 1%, 3%, -2%, -3%, -A%, and -6%, depicted by light-colored bars over these respective years. As shown, the dark bars are ; 2008 THE JOURNAL OF INVESTINC; 2 7

3 EXHIBIT 1 Sample Composition Panel A: Total Sample Panel B: Total Sample Subdivided by Sales Performance Panel C: Total Sample Subdivided by Sales Performance and Cost Flexibility Sales Decline Sales Increase Year N Sales Decline Sales Inerease Iti flexible (1) Flexible (2) Inflexible (3) Flexible (4) Total too Notes: An ohsefratioti is determitied iti >(i/cs Sales Dedlrw if the oiic-year Q4-lo-Q4 penriihu^e chaii^^e in sales is negative. Cost Flexibility is. dctcrmimd based on Operating Leverage, the ratio between fixed assets and income from amiinuing operations, and ftnaiuial Leverage, the ratio between interest expense and income from continuing operations. An observer is determined inflexible if the firm has greater operating leuerage and greaterfinancialleverage than its industry. much deeper than the light bars in all eight years, indicating that the earnings forecast errors of sales-decline firms are larger and more optimistic than are those of sales-increase firms. For a more rigorous test. Exhibit 4 shows the regression analysis of forecast error against a Sales-Decline dummy. which is coded 1 as a firm experiences sales decline, and 0 otherwise. The regression includes other variables as in Exhibit 1, such as Market CapitaHzation, Analyst Following, Earnings Volatility, Earnings Growth. EPS Magnitude, and Loss, to control for their effects. The regression also includes dummy variables for distinct industries and EXHIBIT 2 Summary Description of Sample Firms Year Total [nter-temporal mean N Sales Growth (%) Market Capitalization Analyst Following Earnings Volatility (%) Earnings Growth (%) Earnings Per Share Earnings Loss Operating Leverage L Financial Lcverajie L03 Notes: Sales Growth is one-year Q4-to-Q4 percentage change in sales. Market Capitalization is market value of common shares measured in Million U.S. dollars. Analyst Following is the number of analysts following, measured as the number of estimates entering the consensus forecast. Earnings Volatilily is the I/B/E/S stability index, measured as the mean absolute percentage difference between actual reported earnings per share and a five-year historical EPS growth trend line, expressed as a percentage of trend line earnings per share. Earnings Growth is the l/b/e/s growth index, measured as the average annualized earnings per share growth over the past five years expressed as a percent. UPS Magnitude is the absolute value of Q4 earnings per share. Earnings Loss is coded I for firms reporting Q4 loss. 0 otheni'ise. Operating Leverage is the ratio between fixed assets and iticome from continuing operations. Financial Leverage is the ratio between ittterest expense and income from continuing operations. 28 EARNINCS FOREC:ASTS OF FIRMS EXPERIENCING SALES DECI-FNE: WHY SO INACCURATE? SPRING 2008

4 ExH r B IT 3 Earnings Forecast Errors of Firms Characterized by Sales Performance years, to control for their fixed effects. The coefficient of the Sales-Decline dummy is expected to be negative, consistent with prior findings that forecast errors of poor-performance firms are more optiniistic than others. Market Capitalization, Earnings Growth, and Analyst Following are expected to be positive because large firms, growth firms, and firms followed by many analysts tend to have rich disclosures tbat improve analyst forecasts and reduce forecast optimism. Earnings magnitude is also expected to be positive, because large earnings magnitude results in reduced error and optimism. Firms that are volatile or that incur earnings losses are difficult to forecast, leading to more negative errors; consequently. Earnings Volatility and Earnings Loss are expected to be negative. The regression result in Exhibit 4 shows that all control variables are significant in their expected signs, except Earnings Volatility and Analyst Following, which are insignificant. More importantly, Sales-Decline is significandy negative (p-value < ), indicating that the earnings forecast errors of sales-decline firms are more negative than are those of sales-increase firms. Overall, the results in Exhibits 3 and 4 show that forecast errors of sales-decline firms are more negative than those of sales-increase firms. Indeed, results from this sample confirm prior research that forecasts of poor-performance are larger and more optimistic than are those of strongperformance firms. The results are consistent with prior findings that poor-performance firms are more difficult to follow {Kothari et al. [20()5 ; Brown [2001]; Hwang et al. [1996], and Elgers and Lo [1994]). ARE INFLEXIBLE COSTS ASSOCIATED WITH OPTIMISTIC FORECAST ERRORS? To investigate this question, this article assesses the flexibility ofthe firm's cost structure based on operating and fmancial leverage. As described in Exhibit 2, operating leverage is measured as fixed assets scaled by income from continuing operations, and financial leverage as interest expense on debt over the same scale. An inflexible firm is defined as one having greater operating leverage and greater financial leverage than its industry. On the contrary, a flexible firm is defined as one having smaller operating leverage and/or smaller financial leverage than SPRING 2008 THE JOURNAL OF INVEST[NC; 29

5 EXHIBIT 4 Regression of Forecast Error Definition Intercept Sales-Decline Log of Market Capitalization Analyst Following Earnings Volatility Earnings Growth EPS Magnitude Earnings Loss Dummy variables for Years Dummy variables for industries Expected sign Included Included Coefficient Estimated Value p-value (Two-tailed) Number qfobsermlions: 5,180; Pr > Model F: : Adj-Rsquare %. Notes: Forecast Error is measured as ihe difference between acituil and forecast eartiitij^s,.scaled by actual earinii(^.i, and win.^orized at +2 and -2. Sales Declin is coded I for sales decline, 0 otheni'ise. Market Cap is llie market \'ahie of llie company conmioii shares. Analyst Follou'inf> is the number of analysts followiin;, measured as the number of estimates enterinji the consensus forecast. Barriings Volatility is the I/B/H/S stability index, measured as ihe mean absolute percentage difference between aaual reported earnings per share and a five-year historical EPS growth trend line, expressed as a percentage of trend line earnings per share. Earnings Growth is the I/B/E/S growth index, measured as the average atmualized earnings per share growth over the past five years expressed as a percent. EPS Magnitude is the absolute value of EPS. Earnitigs Loss is coded I for firms reporting Q4 loss, 0 othenvise. its industry. Leverage of an industry is measured as the industry-year-adjusted median, which is the median of all firms belonging to the firm's industry in the same year. For this analysis, the sample is characterized by sales performance and cost structure. Specifically, sample firms are divided into two performance groups, Sales-Decline and Sales-Increase, both of which are further divided into two cost groups. Inflexible and Flexible. The resulting four sub-samples are labeled as Inflex-Decline, Flex- Decline, Inflex-Increase, and Flex-Increase. These subsamples are presented in Panel C of Exhibit 1, Columns 1 4, respectively. Exhibit 5 is a bar graph to visualize the comparison of forecast errors among the sub-samples. The depths of the bars are examined to compare the mean earnings forecast errors ofthe sub-samples. Exhibit 5 shows that the first bar (Inflex-Decline) is deeper than the second bar (Flex- Decline) in seven of eight years, while the third bar (inflex- Decline) is deeper than the fourth bar (Flex-Increase) in all eight years, consistent with the notion that cost inflexibility increases earnings forecast error. Furthermore, consistent with a stronger effect of cost inflexibility' on sales-decline firms than on sales-increase firms, the first bar (Inflex- Decline) is deeper than the third bar (Flex-Decline) in all eight years. Overall, Exhibit 5 shows that inflexible firms incur more negative earnings forecast errors than do flexible firms, especially for periods of sales decline. For a more rigorous test. Exhibit 6 shows the regression analysis of Forecast Error against two variables denoting inflexible costs. The first is a dunmiy variable, Inflex-Decline, coded 1 for firms with inflexible cost and sales-decline, and 0 otherwise. The second is a dummy variable, Inflex-Increase, coded 1 for firms with inflexible cost and sales-increase, and 0 otherwise. It cost inflexibility leads to large negative forecast errors, both Inflex-Decline and Inflex-Increase are expected to be negative. Furthermore, if declines are less anticipated than increases, cost inflexibility should aflect sales-decline firms more strongly than sales-increase firms, and Inflex- Decline should be more negative than Inflex-Increase. The control variables and their expected effects are as in Exhibit EARNINGS FORECASTS OF FIRMS EXPERIHNCING SALES DECUNE: WHY SO INACCURATE? SPRING 2[1O«

6 EXHIBIT 5 Earnings Forecast Errors of Firms Characterized by Sales Performance and Cost Structure The regression results in Exhibit 6 show that Market Capitalization, Earnings Growth, Earnings Magnitude, and Earnings Loss are significant in their expected signs. More importantly, Inflex-Decline is significantly negative as expected. Inflex-Increase is negative, but its effect lacks statistical significance. Furthermore, an F-test to compare the two shows that Inflex-Declirie is significantly more negative than Inflex-lncrease, consistent with the notion that cost inflexibility impacts sales-decline firms more strongly than sales-increase firms. Overall, the results from Exhibits 5 and 6 show that cost inflexibility is associated with larger and more optimistic forecast errors, especially of sales-decrease firms. The results are consistent with the idea that analysts do not assess the impact of inflexible costs correctly in periods of sales dechne. OTHER VALIDITY TESTS The article's findings are validated through a battery of replications based on alternative measurements of key variables, and additional controls such as international diversification, special charges, and earnings loss. The regression analyses reported in Exhibit 4 are replicated using alternative measures of forecast error, and operating and financial leverage. The first replication is based on winsorizing earnings forecast error at +1 and 1, instead of +2 and 2. The second is based on measuring earnings forecast error as a percentage of stock price instead of earnings, and winsorization at The third replication is the same as the second, except that the stock price is at the beginning ofthe year. Fourth, operating leverage is defmed asfixedassets over total assets, and financial leverage as interest expense over total sales. And finally, operating leverage is defined as the three-year cumulative depreciation and amortization expense over the three-year cumulative net income, and financial leverage as the three-year cumulative interest expense over the three-year cumulative net income. All these replications yield results consistent with the main results. The regressions are then replicated with international diversification, special charges, and earnings loss. These are potentially confounding factors because they may affect the earnings forecast errors of sample firms, thus causing this article's conclusions to be erroneous. International diversification tends to increase earnings i 2008 OF INVESTING 31

7 EXHIBIT 6 Regression of Forecast Error Definition Coefficient Intercept Log of Market Cap Analyst Following Earnings Volatility Earnings Growth Earnings Magnitude Earnings Loss Inflex-Deciine Inflex-Increase Dummy variables for Years Dummy variables for industries Expected sign Included Included Estimated Value P-value (Two-tailed) Number of observations = 5,180; Pr> Model F: <.OOOI: AdJ-Rsquare = 18.95%; P-value of F test to compare Injiex-Decline and Itijiex-Increase Notes: Forecast Error is measured as the difference between actual and forecast earnings, scaled by actual earnings, and unnsorized at +2 and 2. Market Cap is the market value ofthe company common shares. Analyst Following is the number of analysts following, measured as the number of estimates entering the consensus forecast. Earnings Volatility is the I/B/E/S stability index, measured as the mean absolute percentage difference between actual reported earnings per share and a Jive-year historical EPS growth trend line, expressed as a percentage of trend line earnings per share. Earnings Growth is the I/B/E/S growth index, measured as the average annuahzed earnings per share growth over the past five years expressed as a percent. EPS Magnitude is the absolute value of EPS. Earnings Loss is coded 1 for firms reporting Q4 loss. 0 othmme. infiex-dechne is coded! for inflexible sales-decline firms, 0 otherwise. Injiex-Increase is coded 1 for i}i{\exible sales-increase firms.o otheni'ise. forecast error because it complicates the forecasting task (Duru and Reeb [2002]). To assess this factor's effect on the articles results, sample firms are divided into purely domestic (with zero foreign sales) versus internationally diversified (with positive foreign sales). Replications of the main analyses on these two separate sub-samples yield consistent fmdings. Special charges dramatically reduce earnings and therefore often lead to large earnings forecast errors (Basu et al. [2000]). Replications using only firms that do not report special items are consistent with the main fmdings. Although earnings loss is a control variable in Exhibits 4 and 5, it may not be adequate because loss is a strong determinant of forecast error (Brown [2001J, Hwang et al. [1996]), and the loss distribution may be problematic for statistical testing (Gu and Wu [2003]). Nonetheless, replications after excluding all loss observations yield results that are consistent with the main results, indicating that earnings loss is not driving the results. SUMMARY AND APPLICATION OF FINDINGS This article demonstrates that firms with inflexible cost structures are generally forecasted with larger and more optimistic errors than are firms with flexible cost structures, especially in sales decline. The results suggest that it is difficult to correctly assess the impact of inflexible costs during sales drops. The results are useful to the investment community in two ways. First, it helps investment managers evaluate analysts' forecast ability by highlighting the greater challenge of following firms with inflexible costs, as analysts must deal with wider earnings fluctuations because costs do not vary to offset revenue variability. For firms with performance declines, the additional challenge 32 EARNINGS FORECASTS OF FIRMS EXPERIENCING SALES DECLINE: WHY SD INACCURATE? SPRING 2(JO8

8 to analysts is that niatiagers tend to delay communication of bad news (Kothari et al. [20051). Second, this article helps analysts improve their forecasts by better anticipating the impact of inflexible costs on earnings. Analysts should be keenly aware of the firm's true cost structure beyond financial statements' numbers to identify costs that are large, unavoidable, and/or already committed. One notable example is operating leverage, which is related to fixed assets. For practical purposes, analysts should pay specific attention to depreciation, depletion, and amortization, which is typically reported with other selling and general administrative (SGA) items. When sales do not change much, SGA may be estimated roughly as a pro-rated quantity of revenues. However, when sales change drastically, pro-rating leads to large error due to the fixed portion from depreciation. Therefore, analysts should estimate the impact of this cost separately from other SGA items. Another example is financial leverage, which is related to debt. Analysts should be aware of the firm's debt structure to estimate interest expense separately from other expenses. ENDNOTE The author gratefully acknowledges the contribution of Thomson Financial I/B/E/S for providing earnings per share forecast data. These data have been provided as part of a broad academic program to encourage earnings expectation research. REFERENCES Basu, S., L. Hwang, and C. Jan. "Auditor Conservatism and Analysts' Fourth Quarter Earnings Forecasts." Working Paper, liaruch College and California State University-Hay ward, January 2U00. Brown, L.D. "A Temporal Analysis of Earnings Forecast Errors; Profits versus Losses." Journal of Accounting Research, 39, (September 2001), pp Collins, W., W. Hopwood, and J. McKeown. "The Predictability of Interim Earnings Over Alternative Quarters." Journal of Accounting Research, 22 (Autunm 1984), pp Duru, A., and D. Reeb. "Internarional Diversification and Earnings Forecast Characteristics." 'Hie Accounting Rcvieu^, 77 (April 2002). pp Elgers, P.T, and M.H. Lo. "Reductions in Analysts" Annual Earnings Forecast Errors Using Information in Prior Earnings and Security Returns." _/i»»m(j/ if Accounting Research, 32 (Auaimn 1994), pp Gu, Z., and S. Wu. "Earnings Skewness and Analysts' Forecast Bias" Journal of Accounting and Economics, 35 (April 2003), pp Hwang, L., C.Jan, and S. Basu. "Loss Firms and Analysts' Earnings Forecast ETTOTS'' Journal of Financial Statement Analysis, 1 (1996), pp Kothari, S.P., S. Shu, and P. Wysocki. "Do Managers Withhold Bad News?" MIT Sloan Research Paper No , Lev, B. "Some Economic Determinants of Time-Series Properties of Earnings." Journal of Accounting and Economics, 5 (April 1983), pp To order reprints of this article, please contact Dewey Palmieri at dpalmieri@iijournals.com or Bodie, Z., A. Kane, and A. Marcus. Essentials of Investments. Boston: McGraw-Hill Irwin, 2U05. SPRING 200S THE JOURNAL OF INVESTING 33

9