FORECASTING CASH FLOWS: A COMPARISON OF PREDICTION MODELS BETWEEN INDUSTRIES. Brooke Young. University of Nebraska at Kearney

Transcription

1 FORECASTING CASH FLOWS: A COMPARISON OF PREDICTION MODELS BETWEEN INDUSTRIES by Brooke Young University of Nebraska at Kearney

2 2 Forecasting Cash Flows: A Comparison of Prediction Models Between Industries I. Introduction Financial reporting should provide information to help investors, creditors, and others assess the amounts, timing, and uncertainty of prospective net cash inflows related to the enterprise (FASB 1978). FASB asserts that current earnings are a better predictor of future cash flows than current cash flows. In recent years, earnings discrepancies have created a renewed interest in cash flow analysis. The ability to accurately predict future cash flows is undeniably important for a company s sustainability and aiding creditor and investing decisions. Additionally, a relationship has been shown to exist between cash flow and the valuation of securities. The most recent use of cash flows involves assessing a firm s worth by the present value of its future cash flows. Accurately predicting future cash flows is important on several levels. However, the best measure by which to predict future cash flows is less than concrete. FASB s assertion has generated research examining the superiority of earnings over cash flow measures when forecasting future cash flows. Early findings by Greenburg, Johnson, and Ramesh (1986) supported FASB s assertion that current earnings are superior to current cash flows when predicting future cash flows. However, Bowen, Burgstahler, and Daley s (1986) results did not confirm that accrual earnings measures had greater predictability to forecast cash flows. More recent studies have progressed to test various measures of cash flows and earnings as predictors of future cash flows. Additionally, some research has studied the predictive ability of cash flows

3 3 being industry specific. Jordan and Waldron (2001) provide seemingly conclusive evidence that a mixed measure of earnings and cash flows best predicted future cash flows within the petroleum industry. The purpose of this research is to test and compare the ability of current earnings and cash flow measures to predict future cash flows between the high-tech industry and the specialty retail industry. This study will contribute to the body of research examining the ability of earnings and cash flows to predict future cash flows by investigating whether superior cash flow prediction methods are industry specific. Further comparing the results of this study to those of the petroleum industry study may provide some answers to the discrepancies in prior research that did not attempt to isolate industries when testing the predictive abilities of earnings and cash flows. Finally, results will be used to support or reject FASB s statement regarding the prediction of future cash flows within the high-tech and specialty-retail industries. II. Literature Review Prior research investigating FASB s claim that current earnings measures are better forecasters of future cash flows than current cash flow measures has progressed through three major stages. Initial research examined the ability of current net income before extraordinary items and discontinued operations (NIBEI) and cash flows from operations (CFO) to forecast future CFO. The next phase of research began to consider several measures of cash flow to compare to earnings. Lastly, some preliminary work has begun to test whether cash flow prediction models may be industry specific. Greenburg, Johnson, and Ramesh (1986) separately tested linear relationships between the independent variables, current NIBEI and CFO, and the dependent variable, future CFO, using a regression model. Generally, this study indicated earnings measures are better predictors

4 4 of future cash flows than current cash flows, and confirmed FASB s statement concerning the prediction of cash flows. A limitation of these results involves the sample of companies taken from varying industries. Equity compositions and varying measurements of accruals that are industry specific may impair conclusions. Bowen, Burgstahler, and Daley (1986) also tested the predictive ability of current earnings and cash flows to predict future cash flows. Dissimilarly, this study included a continuum of cash flow measurements to test as forecasters of the same range of cash flow variables. Univariate models using NI and cash flow measurements were used to predict the range of cash flows. A comparison of the models median prediction errors finds earnings to have the greatest error for predicting each cash flow variables. Therefore, findings disagree with FASB s statement regarding currents earning s superior predictive ability. However, prediction errors of the majority of cash flow variables in this study were not significantly less than those of earnings. Barth, Cram, and Nelson (2001) studied the role of accruals in cash flow prediction. The study finds that disaggregating components of earnings provides information specific to different measures of cash flow. These findings indicate that weighting accrual components of earnings greatly enhances the predictive ability of earnings to predict future cash flows. Results were consistent when compared across industries. Therefore, comparing the ability of earnings measures and cash flow measures to predict future cash flows between industries becomes even more interesting. Jordan and Wallace (2001) examined the predictability of a continuum of cash flow measures ranging from NI to change in cash (CC) to predict future CFO within the petroleum industry. Net income before depreciation (NIDPR) was found to have the strongest correlation to

5 5 future CFO. Jordan and Wallace (2001) further examined predictability of the continuum of measures by comparing the means and medians of percentage prediction errors (PPE). Results of this analysis also indicated NIDPR was the best predictor of future cash flows within the petroleum industry. By limiting the study s focus to the petroleum industry, Jordan and Wallace (2001) eliminated the sample composition error of the Greenburg et al. (1986) and Bowen et al. (1986) studies. The current study will expand upon Jordan and Wallace s research by comparing the predictive abilities of a continuum of cash flow measures between two additional industries. This study will specifically examine the high-tech and specialty retail industries. The reasoning behind this comparison is twofold. First, the balance sheet compositions and accruals differ substantially due to the nature of individual businesses within the industries. High-tech firms typically carry either small inventories or no inventory at all, while specialty retail industry companies carry large inventories. Secondly, the industries differ by their stage of maturity. The high-tech industry is a relatively new industry and young in comparison to the established specialty retail industry. Extreme differences between these two industries make them an ideal pair for comparison in this study. III. Methodology This research will address the following questions: 1. Are current earnings or current cash flow measures superior when predicting future cash flows with respect to the specialty-retail industry?` 2. Are current earnings or current cash flow measures superior when predicting future cash flows with respect to the high-tech industry?

6 6 3. Are the predictive abilities of earnings and cash flow measures in the specialty retail industry equal to the predictive abilities of earnings and cash flow measures in the high-tech industry? Answering these questions will provide further evidence to support or reject FASB s general statement regarding the predictive ability of cash flows. Furthermore, addressing the third question will provide a possible explanation of discrepancies in the previous studies, and find superior cash flow prediction methods to be industry specific. To address these questions, two samples were obtained from the Compustat database listing companies who filed with the SEC from The first sample was comprised of specialty retail firms identified by SIC codes: 5600, 5621, 5651, 5661, 5700, 5940, 5944, 5945, and After eliminating companies with.com names (since their structures may emulate high-tech firms), the sample consisted of 289 companies. The second sample consisted of high-tech firms from the telecom or high-tech industry including mostly companies with.com names. The base of companies was taken from the SIC codes 7370, 731, 7371, 7373, 7374, and 7377 was narrowed to a sample size of 192 firms using a purely random number selection. Similar to the Jordan and Wallace (2001) study, five independent variables along a continuum of cash flows were used to predict future CFO are define as followed: NI=net income before extraordinary items and discontinued operations NIDPR=NI plus depreciation WCFO=CFO plus/minus adjustments related to changes in earnings to current assets and current liabilities CFO=Cash flow from operations as reported on statement of cash flows CC=net change in cash and cash equivalents during the year.

7 7 Ten linear regressions (five for each industry) were developed to test the ability of each of the variables to singularly predict CFO one-year period ahead of the current year. A total of ten years of data from was tested. Industries were kept separate such that high-tech variables only predicted cash flows for high-tech companies and specialty retail variables only predicted cash flows of specialty retail firms. Predictive abilities of the variables were analyzed by comparing the R 2. Those variables with higher R 2 were determined to have greater predictive ability. Consistent with the Jordan and Wallace study, an analysis of the companies residuals examines the percentage prediction errors (PPEs) of the variables. PPEs were calculated as follows: PPE=(Actual CFO-Predicted CFO) / Actual CFF. Paired samples t-tests were developed such that mean PPE for the variable that best predicted future CFO was compared to the mean PPEs of the remaining variables for each industry. Lower mean PPEs indicate a variable is a better predictor of CFO. IV. Results Table 1 shows the statistics of the regression models for each industry. Within the specialty retail industry, CFO had the highest r square which indicates it has the best fit for predicting future CFO. NI, NIDPR, and WCFO also had very high R 2 in comparison to CC. All variables were significant in the specialty retail industry at the.001 level with the exception of CC. The variable with the highest R 2 value for the high-tech tech industry was WCFO. This variable also had the highest R 2 value when comparing the two industries. CFO was the only other variable within the high-tech industry that had the ability to predict future CFO in the hightech industry. Both of these variables were also significant at the.001 level.

8 8 Table 1: Cash Flow Regression Models Variable R 2 Std Error of the Est. F-value Sig. Industry* NI NIDPR WCFO CFO CC # # NI=net income before extraordinary items and discontinued operations NIDPR=NI plus depreciation WCFO=CFO plus/minus adjustments related to changes in earnings to current assets and current liabilities CFO=Cash flow from operations as reported on statement of cash flows CC=net change in cash and cash equivalents during the year. * 0=High-tech industry, 1=Specialty retail industry # Bolding indicates variable with high R 2 for each industry Table 2.0 displays the PPE statistics for the specialty-retail industry. The mean PPE of the variable CFO is smaller than the mean PPEs for all other variables except NIDPR. The two tailed sample test also reveals that the predictive ability of all variables except NIDPR is significantly different than that of CFO. Therefore, the predictive ability of CFO holds superior only to NI, WCFO, and CC in the specialty retail industry. To address first question of whether earnings or a certain cash flow measure is a better predictor of future CFO, the best prediction method seems to fall within the continuum of measures. However, it is unclear whether CFO or NIDPR is the better forecaster of future CFO because they are not significantly different from one another and each variable tested superior to the other in one method.

9 9 Table 2.0: Specialty Retail Percentage Prediction Errors N Mean PPE Std. Dev. Sig. (2-tailed) Pair 1 CFOPPE NIPPE Pair 2 CFOPPE NIDPRPPE Pair 3 CFOPPE WCFOPPE Pair 4 CFOPPE CCPPE Table 2.1 presents the PPE result for the high-tech industry. The mean PPE of WCFO was the smallest mean PPE of all the variables which further indicates its superiority over other measures used to predict future CFO within the high-tech industry. This model also compared the mean PPE of the WCFO, the best predictor when comparing the R 2, to the PPEs of NI, NIDPR, CFO, and CC. Findings show significant differences between the mean PPE of WCFO and the mean PPE errors of NI and CC. To answer the second question of superiority, the variable WCFO seems to be the best predictor of future CFO in the high-tech industry, because it is superior to all other variables in both prediction methods. Table 2.1: High-tech Percentage Prediction Errors N Mean PPE Std. Dev. Sig. (2-tailed) Pair 1 WCFOPPE NIPPE Pair 2 WCFOPPE NIDPRPPE Pair 3 WCFOPPE CFOPPE Pair 4 WCFOPPE CCPPE

10 10 The third question of whether the predictive abilities of earnings and cash flow measures are equal in the specialty retail and high-tech industries is addressed by examining Tables 1, 2.0, and 2.1. A comparison of the R 2 in Table 1 shows that the predictive ability of NI is very different between the high-tech and specialty retail industry. NI has some predictive ability in the specialty retail industry, but no predictive ability in the high-tech industry. This circumstance remains true for the variable NIDPR with an additional implication that increases the difference of predictive ability. As aforementioned, within the specialty-retail industry, the mean PPE of NIDPR was lower than the mean PPE of CFO, the variable with greatest predictive ability according to the regression model. While the predictive abilities vary, WCFO and CFO have at least some ability to forecast cash flows in both industries. The greatest similarity between the industries is that the variable CC does not have any ability to predict future CFO. Table 3 presents the statistics of an independent t-test comparing the differences of the five different mean PPEs between the high-tech and specialty retail industries. The only variables for which the mean PPEs are significantly different between industries are WFO and CFO. Considering this statistic offers a possible explanation to why the results of the Greenburg et al. study found earnings to the stronger predictor of future cash flows. The difference between the mean PPEs for NI was the smallest of all the variables. Table 3: Comparison of Percentage Prediction Errors Between Industries t Sig. (2-tailed) Mean Difference NI PPE NIDPR PPE WFCO PPE CFO PPE CC PPE

11 11 V. Conclusion Since FASB s assertion regarding the earnings superiority to cash flows to predict future cash flows, several studies have tested the validity of this statement. Prior research has produced mixed results as to which measure is a better predictor. These studies, with the exception of the work by Jordan and Waldron (2001), have also failed to separate their samples by industries. The purpose of this paper was to compare the predictive ability of earnings and cash flows along a continuum of measures between the specialty retail and high-tech industries. These two industries were selected for comparison because they are very diverse in regard to their maturity and balance sheet composition. The results of this study found that the best predictor of future CFO was a measure found along the continuum of cash flow measurements. The best predictor of future CFO in the hightech industry was WCFO, while CFO and NIDPR were the superior forecasters in the specialty retail industry. The results produced by the models for the specialty retail industry are similar to the findings for the petroleum industry by Jordan and Waldron (2001) who found NIDPR to be the clear cut superior predictor of future cash flows. Interestingly, CFO was not nearly as predictive as NIDPR in the petroleum industry study as it was for the specialty retail industry. The differences between the abilities of earnings and cash flow measures between the high-tech and specialty retail industry indicate that superior cash flow prediction methods are industry specific. Future research examining cash flow prediction in additional industries would enhance the findings of this study. Groups of industries could be compared according to maturity, similar core operations, etc. to see which, if any, of these defining industry characteristics specifically contribute to the difference of superior cash flow prediction methods between industries.

12 12 Additionally examining cash flow prediction throughout the business life of individual companies may provide further insight to cash flow prediction. Perhaps superior cash flow prediction methods follow a pattern through different stages of maturity. The ability to accurately predict a company s future cash flows is important to executives, investors, and creditors. The usefulness of cash flows in the valuation of securities and a company s worth suggest the recent interest in cash flows will not soon come to an end; making research in the area of cash flow prediction very valuable.

13 13 REFERENCES Barth, M., Cram D., Nelson, K Accruals and the Prediction of Future Cash Flows. The Accounting Review 76: Bowen, R., Burgstahler, D., and Daley, L Evidence on the Relationships Between Earnings and Various Measures of Cash Flow. The Accounting Review 61: Financial Accounting Standards Board. Statement of Financial Accounting Concepts No.1, Objectives of Financial Reporting by Business Enterprises. Standford, Conn.: FASB, Greenburg, R., Johnson, G., Ramesh, K Earnings Versus Cash Flow as a Predictor of Future Cash Flow Measures. Journal of Accounting, Auditing, and Finance 1: Jordan, C., Waldron, M Predicting Cash Flow from Operations: Evidence on the Comparative Abilities for a Continuum of Measures. Journal of Applied Business Research 17:87-94.