• No results found

92 Amortization expense(#65)

2. Earnings Only Model

5.3. Model Estimation

5.3.1. Models Using Information from Statement of Cash Flows

5.3.1.1. In-Sample Estimations Tests

For comparison purposes and consistent with the previous studies (e.g. Barth et al, 2001; Al-Attar and Hussain, 2004), the research models of this study are evaluated using in-sample estimations, which are measured by the adjusted R-squared calculated in OLS regression. In addition, to select the best model, the present research considers Voung‟s test for the explanatory power of research models. Table 5.4 presents the summary of the results of the in-sample estimations for the research models across several prediction horizons and up to five lagged predictors which are discussed in detail below.

Cash Flow Model

The first model of the research models, the cash flow model, captures the predictive ability of current and past aggregate cash flow with respect to future cash flow. Regression summary statistics from this model are presented in Panel A of Table 5.4 across prediction horizons and up to five lagged predictors.

Panel A of the Table 5.5 shows that CFOt (0.6921 with a t-statistic of 95.78) is

significant at the 1% level to predict one-year-ahead cash flows (CFO t+1) and 58.59%

of the future cash flow variations are explained by the CFOt.

In addition, Panel B and C of the Table 5.5 show that CFOt is significant at the

1% level to predict second-year-ahead cash flows (CFO t+2) and third -year-ahead cash

flows (CFOt+3).

Furthermore, the results indicate that up to five years lags of cash flow from operations increase the adjusted R2s from 58.59% to 64.48% in predicting one-year- ahead cash flows. The results are true in predicting two-year-ahead (the adjusted R2s from 44.35% to 51.76%) and three-year-ahead (the adjusted R2s from 37.04% to 44.48%) cash flows.

Aggregate Earnings Model

The second research model assesses the predictive ability of current and past earnings with respect to future cash flows. Regression summary statistics from this model are presented in Table 5.4 across prediction horizons and up to five lagged predictors. Consistent with prior studies, the aggregate earnings are significant in predicting future cash flows. Panel A of Table 5.6 show that EBIT t explains 55.73% of variation in

predicting one-year-ahead cash flows (CFOt+1) when the coefficient of EBIT is 0.7770

with a t-statistic of 90.35.

In addition, Panel B and C of the Table 5.6 present that EBIT t is significant at

the 1% level to predict second-year-ahead cash flows (CFO t+2) and third -year-ahead

cash flows (CFO t+3), and up to three years lags of EBIT from operations increase the

adjusted R2s from 55.73% to 58.0% in predicting one-year-ahead cash flows. The results accurately predict two-year-ahead (the adjusted R2s from 41.77% to 43.95%) and three-year-ahead (the adjusted R2s from 32.78% to 34.95%) cash flows. These findings are inconsistent with Barth et al (2001) who report that “up to six lag of earnings are

significant in predicting next period cash flow”.

The Comparison of CFO and EBIT Models

As discussed earlier, CFO and EBIT are significantly and positively correlated with

future cash flows. The comparison of testing the current aggregate EBIT and the current

CFO models in predicting one-year-ahead cash flow demonstrates that the adjusted R2

for the CFO only model (58.59%) is higher than that for the EBIT only model (55.73%).

This finding is consistent with Barth et al (2001) who report adjusted R2s of 24% for the CFO only model and 15% for the EBIT only model. Habib (2010) also

documents a higher coefficient for CFO (0.82) than the EARN (0.62) and a higher

adjusted R2 for the CFO model (48%) than EARN model (40%). Nevertheless, Kim and

Kross (2005) report the average annual R2 when using the cash flow only model (from 12.9% to 46.9%) and earnings only model (from 12.8% to 52.%) and note that the explanatory power of earnings with respect to future cash flows has been increasing.

Panel B of the Table 5.4 demonstrates that Voung‟s test for cash flow only model versus earnings only model is insignificant for all prediction horizons, meaning that there is no difference between the explanatory powers of these two models.

As a result, although the CFO only model has a higher adjusted R2 than EBIT

only model, according to Voung‟s test, the explanatory power of these two models is similar.

Disaggregated Earnings Model

Table 5.7 indicates summary statistics from estimating the model, which disaggregates earnings into cash flow and aggregate accrual, across prediction horizons. As discussed in the previous section, CFO is significantly and positively correlated with future cash

flows and TACC is significantly and negatively correlated with future cash flows.

Based on Panel A of Table 5.7, the regression of one-year-ahead future cash flows on the current cash flow with aggregate accruals for the entire sample demonstrates that CFOt (0.8336 with a t-statistic of 107.77) and TACCt (0.4090 with a

t-statistic of 35.32) are significant to predict next year cash flows (CFOt+1) and 65.27%

of future cash flow variation is explained by this model. The results suggest that CFO

and TACC provide a better explanation of the variation of future cash flows.

In addition, the coefficient of CFO (0.8336) is more than that of TACC

(0.4090). Thus, CFO has more effect in explaining future cash flows than TACC. This

result also indicates that aggregate total accruals have incremental information content in predicting future cash flows and the aggregate TACC adds to ability of CFO in

predicting future cash flow by increasing the coefficient of CFO in cash flow model

0.0395 in the cash flow model to 0.0339 in the disaggregated earnings model. As a result, these findings provide evidence that accrual accounting improves cash flow predictions and is better predictor of future cash flow than cash accounting. These results are consistent with Barth et al (2001) who document that aggregate accruals adds significantly to ability of CFO in predicting future cash flows.Furthermore, Panel B and

C of Table 5.7 show that CFO t and TACC t are significant at the 1% level to predict

second-year-ahead cash flows (CFOt+2) and third-year-ahead cash flows (CFOt+3).

The results also indicates that up to four years lags of CFO and TACC increase

the adjusted R2s from 65.27% to 68.91% in predicting one-year-ahead cash flows. Most lags of CFO and TACC are insignificant when more than three lags are used in

estimation models.

Comparison of the Disaggregated Earnings Model with CFO and EBIT Models

The comparison of the adjusted R2 of the disaggregated earnings model (65.27%) with

CFO (58.59%) and EBIT (55.73%) models for one-year-ahead prediction of future cash

flows, demonstrates that disaggregating earnings into cash flow and aggregate accruals adds to predictive ability of the model. Decomposing earnings also increases the coefficient of CFO from 0.6921 in the cash flow model to 0.8336 in the disaggregated

earnings model and reduces the intercept from 0.0395 in the cash flow model to 0.0339 in the disaggregated earnings model. Testing the equality of the CFO and TACC

coefficients indicates that when TACC is added as variable increases the explanatory

power the model. Accordingly, CFO and TACC provide a better explanation of the

variation of future cash flows. Thus, the conclusion to be drawn is that the aggregate accrual is incremental to CFOt to predict future cash flows. These results are consistent

with Barth et al (2001) who document that disaggregating earnings increases the adjusted R2 of their CFO model from 24% to 27% in disaggregated earnings model for

one-year-ahead prediction of future cash flows.

Panel B of Table 5.4 shows the results of Voung‟s test, which implies that the explanatory power of the disaggregated earnings model is higher than both the CFO

only model and aggregate earnings model. These results are true for all prediction horizons.

As a result, according to the adjusted R2 and Voung‟s test the disaggregated earnings model is a better predictor of future cash flows than the CFO only model and

aggregate earnings model.

Full Disaggregation Model (Cash Flow with Accruals components)

Table 5.8 indicates summary statistics from estimating the model, which disaggregates earnings into cash flow and accrual components, across the prediction horizons.

The results show that disaggregating earnings into current cash flows from operations and the current components of accruals (Adj.R2 69.98%) for one-year-ahead prediction of future cash flows further increases ability to predict future cash flows compared to disaggregating earnings into current cash flows from operations and aggregate accruals (Adj.R2 65.27%). The results are true for two-year-ahead and three- year-ahead prediction of future cash flows.

In addition, the table demonstrates that the accrual components, with the exception of depreciation and amortisation, are significant in predicting future cash flows with the predicted sign. Inconsistent with Barth et al (2001) depreciation and amortisation are not significant predictor of future cash flows. Nevertheless, in the

prediction of future cash flows for two-years-ahead and three-years-ahead, depreciation and amortisation are significant predictors of future cash flows with the predicted sign. These results are consistent with Al-Attar and Hussain (2004).

Furthermore, although the adjusted R2 of this model in predicting future cash flows for all prediction horizons (when adding lags of variables to the model) increases compared to disaggregating earnings into current cash flows from operations and aggregate accruals, most of the accrual components are not significant predictors of future cash flows.

Panel B of Table 5.4 demonstrates the result of Voung‟s test, which shows that the explanatory power of the full disaggregation model is higher than other models in one-year-ahead prediction of future cash flows.

As a result, according to the adjusted R2 and Voung‟s test the full disaggregation model is a better predictor of future cash flows than other models.