Usefulness of Accounting Estimates

Abstract

This is an exploratory study to examine the quality or usefulness of accounting estimates of companies in China and India over time. Specifically, we examine how well the accounting estimates are able to predict future earnings and cash flows during the period 2003-2013. The results for India indicate that the out-of-sample earnings and cash flow predictions derived are more accurate and more efficient in the more recent period (2010-2013) than the earlier period (2003-2006). In contrast, the out-of-sample earnings and cash flow predictions for China are generally more biased, less accurate, and less efficient. The results indicate abnormal returns earned on hedge portfolios formed on earnings (cash flow) predictions for India in the recent period. In contrast, none of the portfolios for China earn positive returns. The results suggest that the accounting estimates in India in recent years have become better predictors of future earnings and cash flow than accounting estimates in the earlier period. However, the accounting estimates in China are not relevant for predicting earnings and cash flows over the years in the sample period.

Keywords

prediction models earnings cash flows international financial reporting standards accounting systems

Introduction

This article examines the quality or usefulness of accounting estimates of companies in China and India. Quality is measured as to how well the accounting estimates are able to predict future earnings and cash flows. Specifically, we examine the quality of earnings and cash flows as reported by Chinese and Indian firms during the period 2003 through 2013. The sample period excludes the financial crisis of 2007-2009; according to the U.S. National Bureau of Economic Research (NBER), the U.S. recession began in December 2007 and ended in June 2009, and the financial crisis appears to have ended about the same time. We separate the sample period to pre-financial crisis (2003-2006) and post-financial crisis (2010-2013).

This is an exploratory study. There is relatively little research about these emerging economies, so this would be an initial step toward understanding accounting quality in these countries. China and India are among two of the emerging economies of BRIC (Brazil, Russia, India, and China); China is the largest emerging economy, and India is the second largest. Accounting standards in India and China are in the process of convergence to International Financial Reporting Standards (IFRS), but both jurisdictions do not permit application of IFRS yet.

We use four accounting prediction models—earnings; operating cash flows; operating cash flows and accruals; and operating cash flow components. We regress earnings (operating cash flows) on previous year’s accounting variables in the four prediction models to obtain regression estimates. The estimated coefficients are used to calculate firm specific predicted values for earnings and operating cash flows in the following year. We then calculate the difference between the actual and predicted values of earnings (operating cash flows) to obtain firm specific prediction error. Prediction performance metrics are calculated for pre- and post-financial crisis periods in India/China. We determine whether accounting estimates are better at predicting future earnings and cash flows in the post-period than the pre-period.

Overall, the results for India indicate that the out-of-sample earnings and cash flow predictions derived from the four prediction models are more accurate and more efficient after 2010. However, we find opposite results in China. The out-of-sample earnings predictions for China are generally more biased, less accurate, and less efficient. We then examine whether out-of-sample earnings and cash flow predictions based on accounting data in four models contribute to predictable stock returns during the pre- and post-periods in India and China. The results indicate that accounting estimates in India in the post-period are better predictors of future earnings than accounting estimates in the pre-period. In contrast, the findings for China indicate that none of the portfolios earn positive returns; the accounting estimates in China are not relevant for predicting positive future earnings. It is interesting that although accounting standards in India and China are converging to IFRS, they differ in quality or usefulness of accounting estimates in predicting future earnings or cash flows; there is an improvement in predictive quality of accounting estimates in India, but improvement is not evident in China. The results are interesting for users of accounting information in China and India who may be interested in assessing the data for predicting future earnings and cash flows.

The remainder of the article is organized as follows. “Literature Review and Methodology” section presents a summary of prior literature and the methodology. “Sample and Data” section discusses the sample and data. The results are presented in “Results” section. Finally, “Conclusion” section concludes the article.

Literature Review and Method

We summarize prior research on accounting in China and India in the table that follows.

	Summary findings
Research on China
Wang, Chen, Lin, and Wu (2008)	Frequency and magnitude of earnings management have gone up during the post-2000 period.
Chen, Chen, Lobo, and Wang (2011)	Non-state-owned enterprises have higher earnings management (lower quality) than state-owned enterprises.
S. Qu, and Zhuang (2011)	Family ownership in China is associated with lower earnings quality.
Li, Abeysekera, and Ma (2014)	Accrual quality, earnings predictability, and earnings smoothness are significantly different between healthy and bankrupt firms, but not earnings persistence.
Research on India
Sarkar, Sarkar, and Sen (2008)	Board quality is important for reducing opportunistic earnings management but board independence has no significant relation with discretionary accruals.

Since April 2010, the Securities Exchange Board of India (SEBI) has provided an option to listed entities having subsidiaries to submit their consolidated financial results either in accordance with the accounting standards specified in Section 211(3C) of the Companies Act, 1956, or in accordance with IFRS (with required reconciliations). The IFRS converged Indian Accounting Standards (referred to as Ind AS) have been issued but the effective application date of these standards has been deferred to April 2016 for larger firms and April 2017 for other firms.

China has not adopted IFRS. China’s new accounting standards of 2006 have become substantively convergent with IFRS (Y. Ding & Su, 2008). Firms that issue A-shares (which can only be owned and traded by Chinese citizens) are required to comply with Chinese domestic accounting standards that have gone through changes to converge with IFRS. Firms that issue B-shares (which can be owned and traded by foreigners, Chinese citizens, or both) are mandated to comply with International Accounting Standards (IAS)/IFRS. Those that issue both A- and B-shares are required to issue two sets of annual reports, one with Chinese standards and the other with IFRS (Peng, Tondkar, van der Laan Smith, & Harless, 2008). We select only A-share listing firms for the Chinese sample because these firms are required to comply with Chinese domestic accounting standards. This sample will reduce confounding effects from using different sets of accounting standards.

Prediction Models

We examine usefulness of accounting estimates as the ability to predict future earnings and cash flows. We use four prediction models similar to Lev, Li, and Sougiannis (2010) and Li and Sougiannis (2014). The models are earnings, operating cash flows, operating cash flows and accruals, and operating cash flow components.

E A R N_{t + 1} = β_{0} + β_{1} E A R N_{t} + ε_{t},

E A R N_{t + 1} = β_{0} + β_{1} C F O_{t} + ε_{t},

E A R N_{t + 1} = β_{0} + β_{1} C F O_{t} + β_{2} A C C R U A L S_{t} + ε_{t},

EAR N_{t + 1} = β_{0} + β_{1} CF O_{t} + β_{2} Δ A R_{t} + β_{3} Δ IN V_{t} + β_{4} Δ A P_{t} + β_{5} {DP}_{t} + β_{6} OTHE R_{t} + ε_{t},

where EARN = earnings before extraordinary items; CFO = net cash flow from operations; ACCRUALS = EARN−CFO; ΔAR = change in accounts receivable; ΔINV = change in inventory; ΔAP = change in accounts payable; DP = depreciation and amortization expenses; $OTHER = other accruals defined as EARN - (CFO + Δ AR + Δ INV - Δ AP - DP) .$

All variables are scaled by beginning total assets. We run regression of these models to obtain sample estimates.

Out-of-Sample Prediction of Earnings (Cash Flows)

From the regression models above, we use the estimated coefficients to calculate firm-specific predicted values for operating cash flow and earnings in the following year. We then calculate the difference between the actual and predicted values of operating cash flow (earnings) to obtain firm specific prediction error. The following is an example of the prediction of earnings for year 2005 using Model 1.

Estimate the following regression for each country:

EAR N_{2004} = β_{0} + β_{1} EAR N_{2003} + ε_{t} .

Use the country-specific estimated coefficients (β₀ and β₁) to predict earnings, Est(EARN), for each firm in the country:

Est (EAR N_{2005}) = Est (β_{0}) + Est (β_{1}) \times EAR N_{2004} .

Determine prediction error (PE) for each firm in a given country:

P E_{2005} = EAR N_{2005} - Est (EAR N_{2005}) .

We repeat the procedure for every firm and sample year.

We use the metrics in Lev et al. (2010) and Li and Sougiannis (2014) to evaluate the out-of-sample prediction performance. MPE is the average prediction error, the difference between actual and predicted values, indicating prediction bias. MAPE (mean absolute prediction error) is the average absolute difference between actual and predicted values, indicating the accuracy of the prediction. RMSE (root mean square prediction error) also measures prediction accuracy. We obtain ALPHA (the intercept), BETA (the slope coefficient), and the adjusted R² from the Mincer and Zarnowitz (1969) regressions of actual values on predicted values. ALPHA indicates prediction bias, BETA indicates the correlation between actual and predicted values, and the adjusted R² measures how well the predicted values are related to the actual values. Theil’s U statistic, defined as the square root of Σ(Actual − Predicted)² / Σ(Actual)², measures overall accuracy of the forecast. A lower U statistic indicates better prediction.

We test for the significance of the change in MPE and MAPE from the pre- to post-periods using t statistics. We apply a bootstrapping approach in testing for differences in RMSE, ALPHA, BETA, and R² between the pre- and post-periods. For example, to test whether the change in R² is significantly different from zero, we randomly select, with replacement, observations from each subsample to generate representative samples and compute the measure. The procedure is repeated 1,000 times to obtain the empirical distributions of the difference between R² between two periods, and t statistics are calculated.

Hypothesis 1: Accounting estimates perform better in predicting future earnings (cash flows) in the recent period (2010-2013) than the earlier period (2003-2006) as accounting standards converge with IFRS.

Sample and Data

We obtain the data from Datastream database. The sample firms are actively listed on the main stock exchanges in India or China during 2003-2013. The main stock exchanges in India are Bombay Stock Exchange and Indian National Stock Exchange. The main stock exchanges in China are Shanghai Stock Exchange and Shenzhen Stock Exchange. Table 1, Panel A shows our sample selection procedure. We start with 44,253 and 25,245 observations from India and China, respectively. We exclude sample with missing data and exclude firms that do not use local accounting standards.¹ We then eliminate observations that are financial service, real estate, and insurance companies from the sample because these firms have to comply with specific rules and regulations. All variables are winsorized at the top and bottom one percentile of each variable to reduce the effects of outliers. We require our sample to have data available in all years during the study periods. Our final sample consists of 2,266 firm-year observations (206 firms) from India and 7,425 firm-year observations (675 firms) from China. To avoid confounding effects from the financial crisis, we exclude the financial crisis period (2007-2009) from the sample years. Hence, our sample periods cover 2003-2006 (pre-financial crisis) and 2010-2013 (post-financial crisis).

Table 1.

Sample.

Panel A: Sample Selection.
	India	China
	Firm–year observation	Firm–year observation
Initial sample	44,253	25,245
Less: Missing data	29,387	8,856
Less: Accounting standards are unknown or not local GAAPs	59	618
Less: Financial service, real estate, and insurance	561	1,077
Less: Firms not available in all years	11,980	7,269
Final sample	2,266	7,425
Final sample (No. of firms)	206	675

Panel B: Sample Distribution by Industry.
Industry	India		China
	Firm–year observation	% of total	Firm–year observation	% of total
Automobile	198	8.7	385	5.2
Basic resources	209	9.2	836	11.3
Chemical	264	11.7	781	10.5
Construction	209	9.2	429	5.8
Food and beverage	77	3.4	484	6.5
Health care	209	9.2	660	8.9
Industrial goods	407	18.0	1,617	21.8
Media and entertainment	55	2.4	77	1.0
Oil and gas	99	4.4	44	0.6
Peripheral	286	12.6	407	5.5
Retail	0	0.0	616	8.3
Technology	132	5.8	396	5.3
Telecommunication	22	1.0	22	0.3
Travel and leisure	55	2.4	187	2.5
Utility	44	1.9	484	6.5
Total	2,266	100.0	7,425	100.0

Note. GAAP = Generally Accepted Accounting Principles.

Table 1, Panel B shows our sample distribution by industry. We classify all firms into 15 industries based on Datastream classification. The leading industries in India are industrial goods, peripheral, chemical, basic resources, construction, and health care industries, which together make up 69.9% of the sample. The leading industries in China are industrial goods, basic resources, chemical, health care, and retail industries, which together make up 60.8% of the sample.

Table 2 provides descriptive statistics in pre- and post-financial crisis. Panel A shows the descriptive statistics of the Indian sample. The averages of all variables in the post-period are significantly lower than in the pre-period, except the average accruals (ACCRUALS). We find no significant difference in the average other accruals (OTHER) between pre- and post-periods. Table 2, Panel B reports descriptive statistics of Chinese sample. In general, the averages of variables in post-period are significantly higher than in pre-period. However, the average cash flow from operations and depreciation (DP) are significantly reduced over the post-period. We find no statistical difference in the average of changes in accounts payable (ΔAP) and changes in inventory (ΔINV) between pre- and post-crisis periods.

Table 2.

Descriptive Statistics.

Variable	Pre-period		Post-period		M difference
	M	Median	M	Median	Pre − Post
Panel A: India
EARN	0.0969	0.0803	0.0763	0.0592	0.0206***
CFO	0.1151	0.1025	0.0809	0.0731	0.0342***
ACCRUALS	−0.0184	−0.0190	−0.0059	−0.0080	−0.0125***
ΔAR	0.0310	0.0191	0.0205	0.0179	0.0104***
ΔAP	0.0284	0.0197	0.0148	0.0111	0.0136***
ΔINV	0.0285	0.0173	0.0243	0.0144	0.0042*
DP	0.0447	0.0420	0.0334	0.0317	0.0114***
OTHER	−0.0040	−0.0037	−0.0034	−0.0038	−0.0006
Panel B: China
EARN	0.0320	0.0271	0.0370	0.0287	−0.0050***
CFO	0.0684	0.0624	0.0499	0.0452	0.0185***
ACCRUALS	−0.0372	−0.0350	−0.0125	−0.0149	−0.0247***
ΔAR	0.0124	0.0077	0.0270	0.0119	−0.0147***
ΔAP	0.0165	0.0087	0.0181	0.0086	−0.0016
ΔINV	0.0274	0.0125	0.0273	0.0105	0.0001
DP	0.0337	0.0302	0.0321	0.0277	0.0016***
OTHER	−0.0257	−0.0197	−0.0161	−0.0113	−0.0097***

Note. All variables are deflated by beginning total assets and are defined as follows: EARN = earnings before extraordinary items; CFO = net cash flow from operations; ACCRUALS = defined as EARN−CFO; ΔAR = a change in accounts receivable; ΔAP = change in accounts payable; ΔINV = a change in inventory; DP = depreciation and amortization expenses; OTHER = other accruals defined as EARN− (CFO +ΔAR +ΔINV−ΔAP−DP).

Statistical significance at 10% level. ***Statistical significance at 1% level.

Results

Table 3 presents Pearson correlation coefficients among variables for India and China in Panels A and B, respectively. Most of the correlation coefficients are statistically significant. Earnings are more persistent (have higher correlation coefficients with future earnings) than cash flow from operations (have lower correlation coefficients with future cash flows) in the pre- and post-periods. Concurrent cash flow from operations and accruals are negatively correlated. The correlation coefficient between cash flows and next period earnings is higher than that between accruals and next period earnings, consistent with the lower persistence of accruals relative to cash flows as reported in prior studies (e.g., Li & Sougiannis, 2014; Richardson, Sloan, Soliman, & Tuna, 2005; Sloan, 1996).

Table 3.

Correlations.

Panel A: India.
	EARN	CFO	ACCRUALS	ΔAR	ΔAP	ΔINV	DP	OTHER	EARN_t ₊₁	CFO_t ₊₁
India (Pre)
EARN	1
CFO	.652***	1
ACCRUALS	.227***	−.565***	1
ΔAR	.168***	−.098***	.332***	1
ΔAP	.098***	.174***	−.093***	.509***	1
ΔINV	.165***	−.081**	0.294***	.226***	.411***	1
DP	.005	.248***	−.342***	−.018	−.014	.028	1
OTHER	.054	−.290***	.426***	−.379***	−.098***	−.194***	−.111***	1
EARN_t₊₁	.770***	.564***	.084**	.038	−.004	.072**	.009	.008	1
CFO_t₊₁	.534***	.581***	−.192***	−.020	−.032	−.057	.189***	−.116***	.614***	1
India (Post)
EARN	1
CFO	.693***	1
ACCRUALS	.201***	−.531***	1
ΔAR	.195***	−.073**	.352***	1
ΔAP	.142***	.173***	−.049	.419***	1
ΔINV	.210***	−.028	.322***	.214***	.449***	1
DP	−.020	.176***	−.283***	.023	.055	.082**	1
OTHER	.024	−.230***	.316***	−.425***	−.025	−.231***	−.045	1
EARN_t₊₁	.733***	.603***	.075**	.115***	.110***	.134***	−.015	−.033	1
CFO_t₊₁	.568***	.541***	−.052	.055	−.004	.014	.172***	−.049	.687***	1
Panel B: China.
	EARN	CFO	ACCRUALS	ΔAR	ΔAP	ΔINV	DP	OTHER	EARN_t ₊₁	CFO_t ₊₁
China (Pre)
EARN	1
CFO	.423***	1
ACCRUALS	.196***	−.794***	1
ΔAR	.208***	−.118***	.255***	1
ΔAP	.169***	.146***	−.066***	.301***	1
ΔINV	.218***	−.135***	.264***	.181***	.426***	1
DP	.212***	.378***	−.271***	.048**	.061***	.013	1
OTHER	−.006	−.339***	.384***	−.428***	−.072***	−.452***	−.043**	1
EARN_t₊₁	.624***	.373***	.011	.084***	.075***	.087***	.191***	−.049**	1
CFO_t₊₁	.227***	.291***	−.164***	.029	−.003	−.022	.367***	−.071***	.395***	1
China (Post)
EARN	1
CFO	.372***	1
ACCRUALS	.265***	−.782***	1
ΔAR	.228***	−.912***	.260***	1
ΔAP	.170***	.113***	−.005	.459***	1
ΔINV	.163***	−.173***	.292***	.222***	.430***	1
DP	.025	.274***	−.268***	.022	.036*	−.042**	1
OTHER	.032*	−.346***	.363***	−.423***	−.172***	−.505***	.035*	1
EARN_t₊₁	.562***	.321***	.043**	.120***	.067***	.046***	−.009***	−.062***	1
CFO_t₊₁	.224***	.405***	−.270***	−.013	−.045**	−.133***	.224***	−.093***	.432***	1

Note. Pearson correlation coefficients are presented. Variable definitions are under Panel B of Table 2. EARN = earnings before extraordinary items; CFO = net cash flow from operations; ACCRUALS = defined as EARN−CFO; ΔAR = a change in accounts receivable; ΔAP = change in accounts payable; ΔINV = a change in inventory; DP = depreciation and amortization expenses; OTHER = other accruals defined as EARN− (CFO +ΔAR +ΔINV−ΔAP−DP).

Statistical significance at 10% level. **Statistical significance at 5% level. ***Statistical significance at 1% level.

Out-of-Sample Prediction

Table 4 provides results from the out-of-sample prediction of 1-year ahead earnings, EARN_t₊₁. Panel A of Table 4 reports results for India during pre- and post-periods. MAPE and RMSE are lower in the post-period indicating an increase in the overall accuracy. However, Theil’s U is higher in the post-period indicating lower prediction accuracy. MPE has mostly negative values across all models in both periods except for Model 2 in the pre-period; actual earnings are less than predicted earnings, that is, the models give an optimistic bias. The bias is greater in the post-period than the pre-period. ALPHA values are slightly higher in the post-period, indicating more bias. BETA is in the range close to 1.0 indicating that the predicted value is close to the actual value. The efficiency of forecast is higher after crisis as suggested by the higher adjusted R² in Models 2 to 4. Overall, the out-of-sample forecasts derived from the four models indicate that accounting numbers in India may be more accurate and more efficient but more biased in more recent years (post-period) compared with earlier years (pre-period).

Table 4.

Out-of-Sample Prediction, Forecast of EARN_t₊₁.

Panel A: India.
Model	MAPE	MPE	RMSE	ALPHA	BETA	ADJ R²	Theil’s U
India (Pre)
1	0.0411	−0.0079	0.0621	−.0026	0.9534	.6142	0.4207
2	0.0562	0.0040	0.0799	.0094	0.9495	.3623	0.5367
3	0.0421	−0.0067	0.0640	−.0009	0.9489	.5902	0.4329
4	0.0458	−0.0026	0.0683	.0104	0.8841	.5345	0.4640
India (Post)
1	0.0382	−0.0102	0.0596	−.0029	0.9315	.5526	0.5777
2	0.0503	−0.0112	0.0697	−.0095	0.9708	.3868	0.6550
3	0.0357	−0.0078	0.0531	−.0080	1.0313	.6434	0.5095
4	0.0392	−0.0133	0.0571	−.0097	0.9738	.5892	0.5488
Pre – Post
1	0.0028	0.0022	0.0025***	.0003***	0.0218***	.0616***	−0.1570***
2	0.0059	0.0151***	0.0102***	.0189***	−0.0213***	−.0245***	−0.1182***
3	0.0064**	0.0011	0.0109***	.0071***	−0.0824***	−.0533***	−0.0767***
4	0.0066*	0.0108**	0.0112***	.0202***	−0.0898***	−.0547***	−0.0848***
Panel B: China.
Model	MAPE	MPE	RMSE	ALPHA	BETA	ADJ R²	Theil’s U
China (Pre)
1	0.0273	0.0054	0.0394	.0044	0.9458	.4313	0.7349
2	0.0336	−0.0010	0.0481	−.0059	1.1433	.1551	0.8363
3	0.0279	0.0042	0.0395	.0015	1.0135	.4307	0.7305
4	0.0296	0.0025	0.0418	.0005	1.0215	.3620	0.7531
China (Post)
1	0.0294	−0.0020	0.0443	.0023	0.9287	.2844	0.7840
2	0.0349	−0.0045	0.0494	−.0068	1.0853	.1078	0.8178
3	0.0294	−0.0019	0.0437	.0017	0.9315	.3036	0.7657
4	0.0298	−0.0021	0.0446	.0043	0.8341	.2752	0.7592
Pre − Post
1	−0.0021	0.0074***	−0.0049***	.0022***	0.0171***	.1469***	−0.0491***
2	−0.0013	0.0035*	–0.0014*	.0008***	0.0579***	.0473***	0.0185***
3	−0.0015	0.0061***	−0.0042***	−.0001***	0.0820***	.1272***	−0.0352***
4	−0.0002	0.0046***	−0.0027***	−.0038***	0.1874***	.0868***	−0.0060***

Note. Model 1: earnings only, Model 2: cash flow from operations only, Model 3: cash flow from operations and accruals, Model 4: cash flow from operations, change in accounts receivable, change in accounts payable, change in inventory, depreciation and amortization expenses, and other accruals. EARN = earnings before extraordinary items; MAPE = mean absolute prediction error; MPE = mean prediction error, with prediction error calculated as (actual-forecast); RMSE = root mean square prediction error; ALPHA = the intercept from the Mincer and Zarnowitz (1969) regressions of actual values on predicted values; BETA = the slope coefficient from the Mincer–Zarnowitz regressions of actual values on predicted values; ADJ R² = the adjusted R² from the Mincer–Zarnowitz regressions of actual values on predicted values; Theil’s U = Theil’s U statistic, defined as the square root of Σ(Actual − Predicted)² / Σ(Actual)². The Mincer–Zarnowitz regressions are run for each year and the averages of the intercept, the regression coefficient, and adjusted R² are reported for each period. Theil’s U statistic is calculated for each year, and the averages are reported for each period.

Statistical significance at 10% level. **Statistical significance at 5% level. ***Statistical significance at 1% level.

The out-of-sample prediction results in China during pre- and post-periods are reported in Table 4, Panel B. RMSE and Theil’s U are higher in the post-period indicating a decrease in the overall accuracy. We find no difference in MAPE between pre- and post-periods. The sign for MPE is opposite in the pre- and post-periods. It has mostly positive values across all models except for Model 2 in the pre-period; actual earnings are higher than predicted earnings. The sign for MPE is negative in post-period; actual earnings are less than predicted earnings. The bias is pessimistic in pre-period and optimistic in post-period. ALPHA values are slightly higher in the post-period, indicating more bias. BETA is lower in the post-period indicating that lower correlation between the predicted value and the actual value. The efficiency of forecast is lower after crisis as suggested by the lower adjusted R². Overall, the out-of-sample forecasts derived from the four models do not show improvement in accuracy, bias, and efficiency of accounting numbers in China in more recent years (post-period) compared with earlier years (pre-period).

Table 5, Panel A provides the results from the out-of-sample prediction of CFO_t₊₁ for India. MPE and ALPHA values are lower in the post-period compared with the pre-period, indicating less bias over time. The decline in MAPE and RMSE for all models after 2010 suggests an increase in the overall accuracy. However, Theil’s U is higher in the post-period indicating lower prediction accuracy. The efficiency of forecast is higher after 2010 as can be seen from the higher adjusted R². In sum, the results of the prediction of CFO_t₊₁ indicate that accounting estimates are better at predicting future earnings and cash flows in the post-period in India.

Table 5.

Out-of-Sample Prediction, Forecast of CFO_t₊₁

Panel A: India.
Model	MAPE	MPE	RMSE	ALPHA	BETA	ADJ R²	Theil’s U
India (Pre)
1	0.0758	−0.026	0.0972	−.009	0.8775	.2960	0.6211
2	0.0732	−0.016	0.0940	−.010	0.9638	.3409	0.5898
3	0.0714	−0.022	0.0917	−.011	0.9228	.3733	0.5834
4	0.0700	−0.022	0.0896	−.011	0.9196	.4015	0.5715
India (Post)
1	0.0582	−0.0404	0.0818	.003	0.9150	.3198	0.6825
2	0.0581	−0.004	0.0810	−.008	1.0418	.3337	0.6726
3	0.0544	−0.002	0.0767	−.004	1.0274	.4018	0.6363
4	0.0545	−0.006	0.0755	−.001	0.9294	.4192	0.6349
Pre − Post
1	0.0176***	−0.022***	0.0154***	−.012***	−0.0375***	−.0237***	−0.0614***
2	0.0151***	−0.012*	0.0130***	−.003***	−0.0780***	.0072***	−0.0828***
3	0.0170***	−0.020***	0.0150***	−.008***	−0.1047***	−.0285***	−0.0529***
4	0.0155***	−0.016***	0.0141***	−.010***	−0.0098***	−.0177***	−0.0634***
Panel B: China.
Model	MAPE	MPE	RMSE	ALPHA	BETA	ADJ R²	Theil’s U
China (Pre)
1	0.0547	0.0066	0.0735	.0014	1.0738	.0815	0.7095
2	0.0547	0.0027	0.0726	−.0228	1.3881	.1014	0.7038
3	0.0535	0.0047	0.0721	−.0067	1.1809	.1151	0.6968
4	0.0512	0.0034	0.0700	.0078	0.9382	.1680	0.6733
China (Post)
1	0.0575	0.0019	0.0785	.0112	0.7823	.0336	0.8458
2	0.0545	0.0012	0.0736	−.0135	1.1998	.1499	0.8123
3	0.0544	0.0021	0.0739	−.0069	1.1021	.1435	0.8142
4	0.0538	0.0015	0.0736	.0039	0.8990	.1507	0.8072
Pre − Post
1	−0.0027	0.0048	−0.0050***	−.0098***	0.2915***	.0478***	−0.1363***
2	0.0002	0.0015	−0.0009***	−.0092***	0.1883***	−.0486***	−0.1085***
3	−0.0009	0.0027	−0.0018***	.0003***	0.0789***	−.0284***	−0.1174***
4	−0.0026	0.0019	−0.0036***	.0038***	0.0392***	.0173***	−0.1339***

Note. Model 1: earnings only, Model 2: cash flow from operations only, Model 3: cash flow from operations and accruals, Model 4: cash flow from operations, change in accounts receivable, change in accounts payable, change in inventory, depreciation and amortization expenses, and other accruals. CFO = net cash flow from operations; MAPE = mean absolute prediction error; MPE = mean prediction error, with prediction error calculated as (actual-forecast); RMSE = root mean square prediction error; ALPHA = the intercept from the Mincer and Zarnowitz (1969) regressions of actual values on predicted values; BETA = the slope coefficient from the Mincer–Zarnowitz regressions of actual values on predicted values; ADJ R² = the adjusted R² from the Mincer–Zarnowitz regressions of actual values on predicted values; Theil’s U = Theil’s U statistic, defined as the square root of Σ(Actual − Predicted)² / Σ(Actual)². The Mincer–Zarnowitz regressions are run for each year, and the averages of the intercept, the regression coefficient, and adjusted R² are reported for each period. Theil’s U statistic is calculated for each year, and the averages are reported for each period.

Statistical significance at 10% level. **Statistical significance at 5% level. ***Statistical significance at 1% level.

Table 5, Panel B reports the out-of-sample predictions of CFO_t₊₁ for China. RMSE and Theil’s U are higher in the post-period indicating a decrease in the overall accuracy. The sign for MPE is positive in both periods; actual earnings are higher than predicted earnings; that is, the forecast bias is pessimistic. ALPHA values are slightly higher in the post-period, indicating more bias. BETA is lower in the post-period indicating lower correlation between the predicted value and the actual value. The results for efficiency of forecast are mixed as indicated by mixed levels of adjusted R² across the four models between pre- and post-periods; adjusted R²s are higher for Models 2 and 3 and lower for Models 1 and 4 in the post-period. Overall, the out-of-sample forecasts derived from the four models do not show improvement in accuracy, bias, and efficiency of accounting numbers in China in more recent years (post-period) compared with earlier years (pre-period).

Portfolio Analysis

We further examine whether out-of-sample earnings and cash flow predictions based on the accounting data in the four models contribute to predictable stock returns during the pre- and post-periods in India and China. We rank the sample based on the earnings (cash flow) predictions each period and form five portfolios based on the predicted earnings (cash flow). We then calculate market-adjusted returns from holding these portfolios over 90, 180, 270, and 365 days after the fiscal year end. We form zero-investment portfolios by investing (going long) in the top earnings (cash flow) portfolio and selling (shorting) the bottom portfolio and calculate the abnormal returns earned on hedge portfolios. If accounting estimates perform better in predicting future earnings (cash flows) in the post-period than pre-period, then post-period portfolios would earn higher returns than pre-period portfolios.

The results (not reported) support Hypothesis 1 that accounting estimates in India in the more recent period are better predictors of future earnings (and future cash flows) than accounting estimates in the earlier period; post-period returns are significantly higher than pre-period returns for all four models. However, these results do not support Hypothesis 1 for China. We find no significant difference in returns between pre- and post-periods for most of the other models and holding periods.

Conclusion

This is an exploratory study to examine the quality or usefulness of accounting estimates of companies in China and India in the periods 2003-2006 (pre-period) and 2010-2013 (post-period), excluding the financial crisis years of 2007-2009. The results for India indicate that the out-of-sample earnings and cash flow predictions derived from the prediction models are more accurate and more efficient in the post-period. In contrast, the out-of-sample earnings predictions for China are generally more biased, less accurate, and less efficient in the post-period.

We then examine whether earnings and cash flow predictions based on accounting data contribute to stock returns during the pre- and post-periods in India and China. The results indicate abnormal returns earned on hedge portfolios formed on earnings (cash flow) predictions in the post-period for India. We find that the accounting estimates in India in the post-period are better predictors of future earnings than accounting estimates in the pre-period. In contrast, none of the portfolios for China earn positive returns; the accounting estimates in China are not relevant for predicting future returns earnings in the pre- and post-periods.

This article provides a preliminary understanding of the usefulness of accounting estimates for firms in China and India. Accounting estimates in India are useful in predicting future earnings and cash flows, but accounting estimates in China are not. The results in this article are consistent with the findings in La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1998); Ball, Robin, and Wu (2003); and Li and Sougiannis (2014) that accounting quality varies with accounting systems and legal enforcement. Accounting standards in China and India are converging to IFRS, but both countries have not fully implemented IFRS. Future research may extend this research to examine quality of accounting estimates in China and India after both countries have fully implemented IFRS.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Ball

Robin

J. S.

(2003). Incentives versus standards: Properties of accounting income in four East Asian countries. Journal of Accounting & Economics, 36, 235-270.

Chen

Lobo

Wang

(2011). Effects of audit quality on earnings management and cost of equity capital: Evidence from China. Contemporary Accounting Research, 28, 892-925.

Ding

Zhuang

(2011). Accounting properties of Chinese family firms. Journal of Accounting, Auditing & Finance, 26, 623-640.

Ding

(2008). Implementation of IFRS in a regulated market. Journal of Accounting and Public Policy, 27, 474-479.

La Porta

Lopez-de-Silanes

Shleifer

Vishny

(1998). Law and finance. Journal of Political Economy, 106, 1113-1155.

Lev

Sougiannis

(2010). The usefulness of accounting estimates for predicting cash flows and earnings. Review of Accounting Studies, 15, 779-807.

Abeysekera

(2014). The effect of financial status on earnings quality of Chinese-listed firms. Journal of Asia Pacific-Business, 15, 4-26.

Sougiannis

(2014). Mandatory IFRS adoption and the usefulness of accounting information in predicting future earnings and cash flows (Working paper). Chicago: The University of Illinois at Chicago.

Mincer

Zarnowitz

(1969). The evaluation of economic forecasts. In Mincer

(Ed.), Economic forecasts and expectations (pp. 1-46). New York, NY: National Bureau of Economic Research.

10.

Peng

Tondkar

R. H.

van der Laan Smith

Harless

D. W.

(2008). Does convergence of accounting standards lead to the convergence of accounting practices? A study from China. The International Journal of Accounting, 43, 448-468.

11.

Richardson

Sloan

Soliman

Tuna

(2005). Accrual reliability, earnings persistence and stock prices. Journal of Accounting & Economics, 39, 437-485.

12.

Sarkar

Sen

(2008). Board of directors and opportunistic earnings management: Evidence from India. Journal of Accounting, Auditing & Finance, 23, 517-551.

13.

Sloan

(1996). Do stock prices fully reflect information in accruals and cash flows about future earnings?The Accounting Review, 71, 289-315.

14.

Wang

Chen

S. K.

Lin

B.-X.

(2008). The frequency and magnitude of earnings management in China. Applied Economics, 40, 3213-3225.