The prediction of future operating cash flows using accrual-based and cash-based accounting information: Empirical evidence from Viet Nam

arth et al., 2001; Stammerjohan & Nasiripour, 2001; Farshadfar et al., 2008). Some authors support the argument that no significant difference is found in terms of prediction powers of historical earnings and cash flows (Arnold et al., 1991; McBeth, 1993; Pfeiffer et al., 1998). Furthermore, the majority of studies of future cash flow from business operations are conducted in countries with developed capital markets, for instance the United States (Barth et al., 2001; Dechow et al., 1998; Finger, 1994; Greenberg et al., 1986; Murdoch & Krause, 1989, 1990; Stammerjohan & Nasiripour, 2001), Australia (Farshadfar et al., 2008; Percy & Stokes, 1992), England (Al-Attar & Hussain, 2004). Up to date, there are a few studies in developing countries, where the commodity market is small-sized such as Egypt (Ebaid, 2011) and Vietnam (Anh, 2013). In addition, most of studies used data prior to 2010 to predict cash flows. To demonstrate, the study of Barth et al. (2001) covers the period from 1987 to 1996 and Greenberg et al. (1986) from 1963 to 1982. Moreover, Ebaid (2011) used data from 1999 to 2007. Due to the scarcity of studies using data beyond 2010, the research of cash flow prediction has not been updated lately. Therefore, this paper fills the gap by using accrual-based and cash-based accounting information from business operations, which is highly important to determine factors stated in Vietnamese Accounting Standards and U.S. Accounting Standards to predict future operating cash flows of Viet- namese listed firms. It also adds empirical evidence to researches on the same topic in the context of Vietnam - a developing country in Asia with a newly opened stock market, where the forecast of cash flows has not been fully recognized; the determinants to forecast operating cash flows have not been fully assessed; experiences of chief ac- countants are the major sources to forecasts; and forecast methods applied to cash flows are simple and have not been seriously considered. The remainder of the paper is structured as follows: In Section 2, Literature Review; the next section demonstrates Research Hypotheses; Section 4, presents the Research Methodology; Results and Discussion about the relationship between working capital management and firm’s financial performance are presented in Section 5. Finally, the con- clusion and recommendations are explained in Section 6. 2. Literature Review Studies on the forecast of future operating cash flows of listed companies started with researches by Bowen et al. (1986), Greenberg et al. (1986) in the United States, Habib (2010) in Australia, Austin and Andrew (1989) in New Zealand. Since then, the topic has been conducted in other countries including Vietnam (Anh, 2013). These studies are diverse in terms of forecasting factors, data and research methodology as follows: 2.1 Factors in cash flow predictability The factors employed in cash flow prediction are categorized into groups, including (i) historical earnings; (ii) histor- ical operating cash flows; (iii) combination of historical cash flows and aggregated accruals; (iv) combination of cash flows and disaggregated accruals; and (v) cash flow ratios. (i) Historical earnings as a predictor of future cash flows The Financial Accounting Standard Board (FASB) has mentioned forecast operating cash flows since 1978 in which also emphasized was on the stronger interest of investors in historical earnings than historical cash flows in order to predict future cash flows. Nonetheless, FASB did not present empirical evidences to confirm this statement. Since then, researchers have continuously searched for evidences. Historical earnings is considered as a prominent factor used in forecasting future cash flows by various authors such as McBeth (1993), Ali (1994), Finger (1994), Supriyadi (1998), Quirin et al. (1999), Barth et al. (2001), Stammerjohan & Nasiripour (2001), Jordan and Waldron (2001), Al- Attar and Hussain (2004), Mooi (2007), Zhao et al. (2007) and Anh (2013). The value included in a forecasting model can be the normal earnings with one-year lag, two-year lag, thee-year lag or four-year lag (Stammerjohan et al., 2000/2001). Some authors used transformed earnings instead of normal earnings. The transformed earnings are cal- culated by dividing the original earnings by total assets (Stammerjohan et al., 2000/2001; Farshadfar, 2008; Ebaid, 2011) to minimize the firm-size effects on research results; or dividing the normal earnings by average share value (Murdoch et al., 1989/1990); or dividing the normal earnings by outstanding share numbers (Barth et al., 2001; Zhao, 2007). The authors confirmed the predictability of historical earnings to future cash flows, using either normal or transformed earnings. However, the authors have not agreed whether the earnings model is better than other forecast models to predict cash flows (Stammerjohan et al., 2000/2001; Seng, 2006; Ebaid, 2011; Farshadfar et al., 2008). (ii) Historical operating cash flow as a predictor of future cash flows Different research scholars acknowledge the predictability of historical cash flows to future cash flows. The historical H. A. Nguyen and T. H. Nguyen / Management Science Letters 10 (2020) 685 operating cash flows used as a predictor can be the cash flow lagging one year (Barth et al., 2001; Ebaid, 2011) or lagging several years (Barth et al., 2001; Mooi, 2007). Based on empirical evidences, the authors agreed on the pre- dictability of historical operating cash flows to future cash flows. Moreover, some researchers compared the predict- ability of historical cash flows or other forecasting factors, but they did not agree on the best forecasting model. (iii) Combination of past operating cash flows and aggregated accruals The empirical evidences indicate the benefits of combined cash flows and aggregated accruals of earnings in the pre- diction of future cash flows (Barth et al., 2001; Al-Altar, 2003; Ebaid, 2011). However, they did not have consensus on the best forecasting model. (iv) Combination of historical operating cash flows and disaggregated accruals Disaggregated accruals consist of depreciation, change of accounts receivable, change of prepaid expenses, change of inventory, and change of accounts payable. There is a common agreement among most of studies that the forecasting model, using disaggregated accruals as predictors is useful to cash flow prediction, and this model is more powerful than the models, which use solely past earnings or past cash flows (Barth et al., 2001; Al-Altar, 2003; Ebaid, 2011). However, disagreed with this statement, some researchers stated that the inclusion of disaggregated accruals into the cash flow model did not raise the predictability power and the model, using historical cash flows to predict future cash flows, was the best choice. (v) Cash flow ratios as a predictor of future cash flows Up to date, cash flow ratios have been used as predictors in some studies in countries such as Thailand. However, it came to the conclusion that the cash flow ratios are not useful to predict cash flows in listed Thai companies during the research period (1994 - 2002). The research admitted the results were affected by the economic crisis in Thailand during this time. Therefore, other researches are needed in new context to verify this statement. Data and research methods of forecasting cash flows The researches mainly use data extracted from the financial statements of non-financial listed firms rather than finan- cial listed companies due to the absence of aggregated accruals data such as inventory and receivables in financial statement (Ebaid, 2011). In addition, several studies (Barth et al., 2001; Ebaid, 2011; Farshadfar et al., 2008; Jordan & Wallace, 2007) classified non-financial listed companies into industry groups. For example, Jordan and Wallace (2007) only studied cash flow forecasts in the U.S. oil industry. Besides, a number of authors (Stammerjohan & Na- siripour, 2001; Mooi, 2007; Ebaid, 2011; Anh, 2013) used pooled data. In terms of time series, the first study of cash flow forecasts by Greenberg et al. (1986) employed a 20-year data set (1963 - 1982) in the United States. Afterwards, Al-Altar (2004) used a 10-year data set (1991 - 2000) in the UK; etc. Regarding to the methodology, after formulating the cash flow forecasting equations (the dependent variable is operating cash flow and the independent variables are past earnings, past cash flows, disaggregated accruals, cash flow ratios, etc.), most of the authors give regressions results using Ordinary Least Squares (OLS) (Barth et al., 2001; Mooi, 2007; Ebaid, 2011), while others used Fixed Effect Model (FEM), Random Effect Model (REM) and Step-wise Regression to improve the confidence of research results. In addition, the majority of authors relied on the coefficient of determination (denoted by R2) obtained from the regression analysis to find the best predictive model. Accordingly, the model with higher R2 provides more accurate prediction (Barth et al., 2001; Mooi, 2007; Ebaid, 2011) 3. Research Hypotheses 3.1. Predictability of historical earnings on forecasting operating cash flows The relationship between cash flow and earning has been confirmed in both Accounting Standards and empirical studies of various scholars. According to the Vietnamese Accounting Standard No. 24, Cash Flow Statements prepared by the indirect method are accounting before tax earnings adjusted for non-cash items to generate operating cash flows (such as fixed asset depreciation, provisions, etc.), change of inventory, receivables, payables and other items, of which cash flows from investment activities generate cash effects. A large number of scholars support the importance of earnings and the relationship between earnings and cash flows. Kim and Kross (2005) found a weakening relationship between earnings and stock prices over time, and a significantly increasing relationship between current earnings and future operating cash flows over time and investors tend to be interested in both earnings and cash flows because a profitable company needs to have enough cash to pay dividends. Therefore, earnings and cash flows are two important inputs to decisions of business managers. Other scholars, including Murdoch and Krause (1990), Finger (1994), 686 Dechow et al. (1998), Barth et al. (2001) and Ebaid (2011) investigated the predictability power of earnings in cash flow forecasting. They all agreed that earnings could be used to predict future operating cash flows (Barth & et al., 2001; Ebaid, 2011; Anh, 2013). Nonetheless, the predictive power differed across studies due to the diversity in re- search data as well as time series. Therefore, the following hypothesis is employed in this research: H1: Past earnings are significant in predicting future operating cash flows of non-financial companies listed on the Ho Chi Minh Stock Exchange. 3.2. The possibility to use past operating cash flows in the forecast of future operating cash flows The International Accounting Standard Board (IASB) IAS 7 “Cash Flow Statement” identifies that past cash flows are often used as a predictor of amounts, timing and uncertainty of future cash flows. Malaysian Accounting Standard (MASB, FRS 107, 2010) states cash flow statements are useful to accounting information users to assess the predict- ability on cash flows as well as the timing and uncertainty of predicted cash flows in companies. The empirical studies also explored the predictive ability of historical cash flows (Greenberg et al., 1986; Bowen et al., 1986; Finger, 1994; McBeth, 1993; Quirin et al., 1999; Barth et al., 2001; Anh, 2013), which all suggested that historical cash flows possessed significant explanatory power in predicting future cash flows. Similar to earnings, the predictive power of past cash flows varied among studies. Therefore, the following hypothesis is employed in this research: H2: Historical operating cash flows are significant in predicting future operating cash flows of non-financial companies listed on the Ho Chi Minh Stock Exchange. 3.3. Aggregated accruals in the forecast of future operating cash flows Various researchers (Dechow, 1994; Dechow & Dichev, 2002; Dechow et al., 2008) have disaggregated earnings into two components, which are the cash component in earnings (i.e., revenue and expenses actually collected and paid in cash) and the accrual component in earnings (i.e., revenue and expenses pending to receipt and payment). Thus, the accruals and cash flows are interrelated and the use of accruals to predict future cash flows is valid. Barth et al. (2001), Al-Attar and Hussain (2004), and Ebaid (2011), studied simultaneously the effects of cash flow and accrual compo- nents in the forecast of future cash flows, instead of solely either cash flows or earnings as previous studies. These studies confirmed the predictive power of accruals on future cash flows, though the predictive ability of historical cash flows, in conjunction with accruals, differed across studies. Therefore, the following hypothesis is employed in this research: H3: Historical operating cash flows combined with aggregated accruals are significant in predicting future operating cash flows of non-financial companies listed on the Ho Chi Minh City Stock Exchange. 3.4. Disaggregated accruals in the forecast of future operating cash flows Wilson (1986) and Lorek and Willinger (1996) showed the outperformance of the research models using additional variables from the company's financial statements rather than earnings or cash flows only. Jordan et al. (2007) indi- cated that the best cash flow predictor was not solely cash flow or accruals, but probably the combination of them. Some empirical studies worldwide adopted disaggregated accruals as predictors (Barth et al., 2001; Jordan et al., 2007). They confirmed the disaggregated accruals possessed a predictive power on future cash flows. However, the forecasting ability of historical cash flow models with disaggregated accruals varied among studies. Therefore, the following hypothesis is employed in this research: H4: Historical operating cash flows combined with disaggregated accruals are significant in predicting future operating cash flows of non-financial companies listed on the Ho Chi Minh Stock Exchange. 3.5. Cash flow ratios in the forecast of future operating cash flows Cash flow ratios are better measurement of the operating performance than other financial ratios derived from Income Statement and Balance Sheet, because the effects of other non-cash items such as depreciation of tangible assets could be eliminated from cash flow ratios (Kelly & O'connor, 1997; Plewa & Friedlob, 1995). Therefore, the financial per- formance of a business is correctly assessed. These studies shed a light on new research on the use of cash flow ratios to predict future operating cash flows. Researches on the predictability of cash flow ratios on future cash flows have not been found in Vietnam, therefore the study of predictive power of cash flow ratios in Vietnam will better inform decisions of investors. Therefore, the following hypothesis is employed in this research: H. A. Nguyen and T. H. Nguyen / Management Science Letters 10 (2020) 687 H5: Historical cash flow ratios are significant in predicting future operating cash flows of non-financial companies listed on the Ho Chi Minh Stock Exchange. 3.6. Comparing predictability in forecasting models It has been observed that researchers around the world have tried to find empirical evidence to prove the predictability of accounting information on future operating cash flows. Their conclusions, however, are not consistent when it comes to the selection of the best forecasting model, probably due to divergence of research context, methods, among others. Following the idea of previous researchers, the sixth hypothesis is included in the study with the expectation to fill the knowledge gap of operating cash flow prediction in non-financial listed companies given the research context of Vietnam. H6: The model that combines operating cash flows and disaggregated accruals best predicts operating cash flow of non-financial companies listed on Ho Chi Minh Stock Exchange. 4. Research Methodology 4.1. Data Collection The research uses accounting data from financial reports of companies listed on Ho Chi Minh Stock Exchange (HOSE). There are 384 companies listed on HOSE in 2018, including 90 financial companies and 294 non-financial firms. Since financial companies are subject to a special accounting regime, which is different from that in non-finan- cial businesses, only non-financial companies are included in the sample to achieve data consistency. The sampling method was adopted by a large number of scholars (Ebaid, 2011; Stammerjohan & Nassiripour, 2001, Farshadfar et al., 2008; Mooi, 2007). In addition, the non-financial companies selected for the research sample should be listed on HOSE in at least five consecutive years between 2009 and 2018, financial year ending by December 31 with available operating cash flows and earnings. Consequently, the data sample was composed of 242 non-financial companies listed on HOSE. 4.2. Research Models Learning the models of cash flow prediction employed by other authors, the predictability on future operating cash flows, using accounting information, is assessed for non-financial companies listed on the Hochiminh City Stock Exchange of Vietnam by the specific models as follows: (i) Prediction of cash flows from operating using historical earnings (Earnings Model) CFOt = β0 + β1 EARNt-1 + ε, (1) CFOt = β0 + β1EARNt-1 + β2EARNt-2 + ε, (2) CFOt = β0 + β1EARNt-1 + β2EARNt-2 + β3EARNt-3 + ε. (3) (ii) Prediction of cash flows from operating using historical Operating Cash Flows (Operating Cash Flows Model) CFOt = α0 + α1 CFOt-1 + μ, (4) CFOt = α0 + α1 CFOt-1 + α2 CFOt-2 + μ, (5) CFOt = α0 + α1CFOt-1 + α2 CFOt-2 + α3 CFOt-3 + μ. (6) (iii) Prediction of cash flows from operating using historical Operating cash flows combined with aggregated accruals CFOt = λ0 + λ1 CFOt-1 + λ2 ACRt-1 + ε, (7) CFOt = λ0 + λ1 CFOt-1 + λ2 CFOt-2+ λ3 ACRt-1 + λ4 ACRt-2+ ε, (8) CFOt = λ0 + λ1 CFOt-1 + λ2 CFOt-2+ λ3 CFOt-3+ λ4 ACRt-1 + λ5 ACRt-2+ λ6 ACRt-3 + ε. (9) (iv) Prediction of cash flows from operating using historical operating cash flows combined with disaggregated ac- cruals CFOt = e 0 + e1 CFO t-1 + e2 ΔARt-1 + e3ΔAP t-1 + e4ΔINV t-1 + e5 ΔOTH t-1 + e6DPRM t-1 + ρ, (10) CFOt = e 0 + e1 CFOt-1 + e2 CFOt-2 + e3 ΔARt-1 + e4 ΔAR t-2 + e5ΔAP t-1+ e6ΔAPt-2 e7ΔINVt-1 + e8ΔINV t-2 + e9ΔOTH t-1 + e10 ΔOTHt-2 + e11DPRM t-1 + e12DPRM t-2 + ρ , (11) CFOt = e 0 + e1CFOt-1 + e2CFOt-2 + e3 CFO t-3 + e4ΔAR t-1 + e5 ΔARt-2 +e6 ΔAR t-3 + e7ΔAP t-1 + e8ΔAP t-2 + e9ΔAP t-3 + e10ΔINV t-1 + e11ΔINV t-2 + e12ΔINVt-3 + e13 DPRM t-1 + e14DPRM t-2 + e15DPRM t-3 +e16ΔOTH t-1 + e17 ΔOTH t-2 + e18 ΔOTH t-3 +ρ. (12) 688 (v) Prediction of cash flows from operating using historical cash flows ratios (Cash flows ratios models) CFOt = β0 + βi CFR1t-1 + βi CFR2t-1 + βi CFR3t-1 + βi CFR4t-1 + βi CFR5t-1 + βi CFR6t-1 + βi CFR7t-1 + βi CFR8t-1 + βi CFR9t-1 + ε, (13) CFOt = β0 + β1 CFR1t-1 + β2 CFR2t-1 + β3 CFR3t-1 + β4 CFR4t-1 + β5CFR5t-1 + β6 CFR6t-1 + β7 CFR7t-1 + β8 CFR8t-1 + β9 CFR9t-1 + β10CFR1t-2 + β11 CFR2t-2 + β12 CFR3t-2 + β13 CFR4t-2 + β14 CFR5t-2 + β15 CFR6t-2 + β16 CFR7t-2 + β17 CFR8t-2+ β18 CFR9t-2 + ε. (14) Variable Measurements Research data collected directly from financial statements are presented in Table 1. Table 1 Measurement of Dependent and Independent Variables Code Variables Measurements CFOt CFOt-1 CFOt-2 CFOt-3 Operating cash flows corresponding to year t (forecast year) and year t-1 (one year prior to the forecast year), year t-2 (two years prior to the forecast year), year t-3 (three years prior to the forecast year) Value of "Net cash flow from operating activities" is taken from item coded 20 on Cash Flow Statement. EARNt-1 EARNt-2 EARNt-3 Earnings of year t-1 (one year prior to the forecast year), year t-2 (two years prior to the forecast year), year t-3 (three years prior to the forecast year) Value of "Net earning from business activities" is taken from item coded 30 on Income Statement. ACRt-1 ACRt-2 ACRt-3 Total value of aggregated accruals of year t-1 (one year prior to the forecast year), year t-2 (two years prior to the forecast year), year t-3 (three years prior to the fore- cast year) Differences between value of “Earnings from business activities" is taken from item coded 30 on Income Statement and the value of "Net cash flow from operating activities" taken from item coded 20 on Cash Flow Statement. ΔARt-1 ΔARt-2 ΔARt-3 Change of accounts receivable in year t-1 (one year prior to the forecast year), year t-2 (two years prior to the forecast year), year t-3 (three years prior to the fore- cast year) The value is taken directly from item coded 09 "Change of receiv- ables" on Cash Flow Statement. ΔAPt-1 ΔAPt-2 ΔAPt-3 Change of accounts payables in year t-1 (one year prior to the forecast year), year t-2 (two years prior to the forecast year), year t-3 (three years prior to the forecast year) The value is taken directly from item coded 11 "Change of paya- bles" on Cash Flow Statement. ΔINVt-1 ΔINVt-2 ΔINVt-3 Change of inventory in year t-1 (one year prior to the forecast year), year t-2 (two years prior to the forecast year), year t-3 (three years prior to the forecast year) The value is taken directly from item coded 10 "Change of inven- tory" on Cash Flow Statement. DPRMt-1 DPRMt-2 DPRMt-3 Depreciation of tangible fixed assets and intangible fixed assets in year t-1 (one year prior to the forecast year), year t-2 (two years prior to the forecast year), year t-3 (three years prior to the forecast year) The value is taken directly from item coded 02 "Depreciation ex- penses" on Cash Flow Statement. ΔOTHt-1 ΔOTHt-2 ΔOTHt-3 Change of other prepaid expenses in year t-1 (one year prior to the forecast year), year t-2 (two years prior to the forecast year), year t-3 (three years prior to the fore- cast year) The value is taken directly from item coded 12 "Change of pre- paid expenses" on Cash Flow Statement. CFR1 Cash readiness for main business activities Net cash flow from operating activitiesPayment for fixed assets and other long term assets + Dividend payment + Principal re- payment where: Principal repayment is taken from code 34 "Principal repayment" on the Cash Flow Statement. Payment for fixed assets and other long-term assets: taken from item coded 21 "Payment for procurement and construction of fixed assets and other long-term assets" on Cash Flow Statement. Dividend payment: taken from item coded 36 "Dividend and profit payment to shareholders" on Cash Flow Statement. CFR2 Cash readiness for loan repayment Principal repaymentNet cash flow from operating activities CFR3 Cash readiness for dividend payment Dividend paymentNet cash flow from operating activities CFR4 Cash readiness for investment Payment for fixed assets and other long-term assetsNet cash flow from operating activities CFR5 Ratio of Liabilities to Cash flow Average LiabilitiesNet cash flow from operating activities CFR6 Ratio of Depreciation to Cash flow DepreciationNet cash flow from operating activities CFR7 Ratio of Cash flow to Net revenue Net cash flow from operating activitiesNet revenue where: The value of Net revenue is taken from item coded 03 "Net reve- nue from sales of goods and services" on Income Statement. CFR8 Ratio of Cash flow to Earning Net cash flow from operating activitiesNet earnings CFR9 Ratio of Cash flow to Total assets Net cash flow from operating activitiesTotal assets where: Total assets balance is calculated by dividing the sum of total as- sets (code 270 in Balance Sheet) at the beginning and at the end of the period by 2. H. A. Nguyen and T. H. Nguyen / Management Science Letters 10 (2020) 689 4.3 Research Design Three common estimation methods applied to mixed data, including Ordinary Least Squares (OLS), Random Effects Model (REM) and Fixed Effects Model (FEM). These approaches are currently adopted by researchers in Vietnam such as Chuong and Quynh (2015) and Duong and Hung (2014). These estimations use OLS, FEM, REM to examine sensitivity and variation of variables structure in the models. Therefore, OLS, FEM, REM regressions are performed by using STATA 12 software in this study to estimate the models. Having the OLS, FEM, and REM regression results, it is followed by test to select the suitable model among OLS, FEM and REM. In the Hausman test, the null hypothesis is that REM is suitable against the alternative hypothesis in favour of FEM. If the p-value in Hausman test is less than 0.1 (10%), the null hypothesis is rejected and FEM is selected to discuss the results. Otherwise, REM is opted. Given the chosen model, additional tests are made to examine the model suitability and check the regression assumptions such as constant errors variance, independent residuals and autocorrelation. The F-statistic value obtained from model estimation is used to test suitability (i.e., the model is suitable if Prob (F-statistic) value is less than 0.1). The hypothesis of homoscedasticity is checked by Breusch-Pagan-Godfrey test. Breusch-Pagan-Godfrey test sets the null hypothesis of constant variance against the alternative hypothesis that variances are not constant. In case the F-statistic value resulted from the test is less than 0.1, it is concluded variances change in the model. To detect the presence of autocorrelation, the Durbin - Watson (DW) statistic is used. If 1<DW <3, the model has no autocorrelation; if 0<DW<1, the model has a positive autocorrelation; and if 3<DW <4, the model has a negative autocorrelation (Trong & Ngoc, 2013). The model defects discovered from test results are corrected followed by discussions on research results. Then, conclusions are drawn from the discussion of assumptions. Like previous studies, the preditability value of each model is measured by the R2 coefficient. Accordingly, a higher R2 value implies higher predictive power in the model. 5. Results and Discussion 5.1. Descriptive Statistics Table 2 Descriptive Statistics of research variables (Unit: VND 1,000,000) Variables Mean Median Max Min Standard Errors EARNt-1 175,000 42,900 12,300,000 (412,000) 694,000 EARNt-2 162,000 47,200 7,350,000 (246,000) 507,000 EARNt-3 129,000 39,800 5,340,000 (481,000) 379,000 CFOt 143,000 24,200 14,600,000 (3,050,000) 924,000 CFOt-1 96,200 18,800 11,500,000 (3,050,000) 724,000 CFOt-2 74,300 15,200 9,800,000 (2,790,000) 568,000 CFOt-3 61,800 13,300 7,230,000 (2,200,000) 410,000 ACRt-1 92,600 21,600 5,710,000 (2,110,000) 425,000 ACRt-2 101,000 27,000 3,180,000 (2,110,000) 396,000 ACRt-3 77,100 20,600 3,180,000 (1,840,000) 349,000 ΔAPt-1 52,600 0 8,680,000 (2,550,000) 422,000 ΔAPt-2 53,600 0 8,680,000 (845,000) 400,000 ΔAPt-3 33,600 0 3,840,000 (845,000) 253,000 ΔARt-1 43,700 0 4,410,000 (1,480,000) 256,000 ΔARt-2 50,800 721 4,410,000 (1,310,000) 250,000 ΔARt-3 48,300 0 4,410,000 (632,000) 239,000 ΔINVt-1 64,400 0 7,010,000 (1,160,000) 347,000 ΔINVt-2 61,500 21 4,360,000 (1,180,000) 256,000 ΔINVt-3 42,900 0 1,740,000 (1,180,000) 174,000 ΔOTHt-1 1,500 0 477,000 (1,080,000) 50,900 ΔOTHt-2 5,300 0 676,000 (455,000) 42,800 ΔOTHt-3 4,710 0 676,000 (225,000) 38,500 DPRMt-1 49,900 11,800 3,090,000 0 182,000 DPRMt-2 41,500 9,680 2,800,000 0 147,000 DPRMt-3 33,900 8,150 1,730,000 0 114,000 Table 2 shows the positive values of earnings (EARN), operating cash flows (CFO), aggregate accruals (ACR), indi- cating positive earnings and net operating cash flows in most of non-financial companies listed on HOSE in the period from 2009 to 2018. The observed positive values of disaggregated accruals (ΔAP, ΔAR, ΔINV, ΔOTH and DPRM) mean that non-financial companies listed on HOSE have higher year-end receivables, payables, inventory, prepaid expenses, and fixed asset depreciation against the year-beginning values. 5.2 Discussions The Hausman's test results show that the FEM regression is the most suitable for all cash flow forecasting models for non-financial companies listed on HOSE. Specifically, the FEM regression results of cash flow forecasting models are presented in the following tables: 690 Regression results of earnings model The relationship between future operating cash flows and historical earnings (with one-year lag, two-year lag and three-year lag) is shown in Table 3: Table 3 Summary of FEM regression results of earnings model Variables One-year lag Two-year lag Three-year lag EARNt-1 0.708268*** 1.037564*** 0.841508*** EARNt-2 0.832427*** 1.072893*** EARNt-3 0.699031*** Prob (F- Statistic) 0.00000 0.00000 0.00000 DW 2.03 1.83 1.88 Adjusted R2 0.773219 0.792889 0.813363 *Significant at 0.1 level. ** Significant at 0.05 level. Significant at 0.01 level. The Prob (F-statistic) value from the FEM regression indicates that earnings models have predictive ability on firm's future cash flows. The Adjusted - R2 coefficients in these earnings models are 0.77, 0.79 and 0.81, respectively, further show that earnings with one-year lag, two-year lag and three-year lag explain 77%, 79% and 81%, respectively, of the variation in future operating cash flows. FEM regression results for three detailed earnings models (with the independ- ent variables lagging one year, two years and three years) indicate that these models have predictive ability on future cash flows This result is shared by various authors who studied the cash flow prediction by earnings such as Barth et al. (2001), Ebaid (2011), Chong (2012) and Anh (2013). In addition, the longer lag in predictors of these models results in stronger predictive power (from 77% to 81%), meaning that the model with a three-year lag has higher predictive ability than the models with one-year lag or two-year lag. As a result, the first research hypothesis (H1) is accepted. Regression results of the cash flow model The relationship between future operating cash flows and historical operating cash flows (with one-year lag, two-year lag and three-year lag) is shown in Table 4: Table 4 Summary of FEM regression results of the cash flow model Variables One-year lag Two-year lag Three-year lag CFOt-1 0.497463*** 0.493368*** 0.515405*** CFOt-2 0.411615*** 0.337359*** CFOt-3 0.260078*** Prob (F- Statistic) 0.00000 0.00000 0.00000 DW 2.613703 2.695733 2.648768 Adjusted R2 0.764312 0.789254 0.796054 *Significant at 0.1 level. ** Significant at 0.05 level. Significant at 0.01 level. FEM regression results show all cash flow models (with one-year lag, two-year lag and three-year lag) are capable of predicting future cash flows. Predictability of historical cash flow models (with one-year lag, two-year lag and three- year lag) on future cash flows are 76%, 78% and 79%, respectively. This conclusion concurs with the authors of pre- vious studies in the same topic such as Greenberg et al. (1986), Bowen et al. (1986), Mc Beth (1993), Barth et al. (2001), Chotkunakitti (2005) and Anh (2013). Predictability of FEM regression models increases from 76% to 79% (shown by the Adjusted R2 coefficient). In addition, the longer lag in predictors in these models results in stronger predictive values (from 76% to 79%), meaning that the model with a three-year lag has higher predictive ability than the models with one-year lag or two-year lag. Thus, the second research hypothesis (H2) is accepted. FEM regression results of the cash flow model combined with aggregated accruals The relationship between future operating cash flows and historical operating cash flows combined with aggregated accruals (with one-year lag, two-year lag and three-year lag) is shown in Table 5. The table shows FEM regression results of three prediction models (with one-year lag, two-year lag and three-year lag in the predictors) are capable of predicting future operating cash flows. This conclusion is similar to previous studies such as Barth et al. (2001), Ebaid (2011), among others. The Adjusted - R2 in these models are 0.82, 0.82 and 0.87, respectively, supporting the increasing explanatory value of future cash flow variations from 82% to 87% in cases of longer lag in predictors. The results in this study coincide with what was found by Barth et al. (2001) in terms of both the significance of predictors and the effect direction of predictors on future operating cash flows. Thus, the third research hypothesis is accepted. H. A. Nguyen and T. H. Nguyen / Management Science Letters 10 (2020) 691 Table 5 Summary of FEM regression results in the model of cash flows combined with aggregated accruals Variables One-year lag Two-year lag Three-year lag CFOt-1 0.187462** 0.192402** 0.080905 CFOt-2 0.029462 0.415047*** CFOt-3 0.763306*** ACRt-1 0.903857*** 0.879525*** 1.001574*** ACRt-2 0.095209 0.16145** ACRt-3 1.106636*** Prob (F- Statistic) 0.0000 0.000000 0.000000 DW 2.418038 2.434564 2.737023 Adjusted R2 0.828346 0.828309 0.875557 *Significant at 0.1 level. ** Significant at 0.05 level. Significant at 0.01 level. FEM regression results of the cash flow model combined with disaggregated accruals The relationship between future operating cash flows and historical operating cash flows combined with the disaggre- gated accruals (with one-year lag, two-year lag and three-year lag) is shown in Table 6: Table 6 Summary of FEM regression results of the cash flow model combined with the disaggregated accruals Variables One-year lag Two-year lag Three-year lag CFOt-1 0.06368 0.067608 -0.105004 CFOt-2 -0.278544*** -0.206533*** CFOt-3 -0.106958 ΔAPt-1 -0.724623*** -0.763389*** -0.706308*** ΔAPt-2 0.079242 -0.113543 ΔAPt-3 0.018948 ΔAR t-1 0.877728*** 0.751735*** 0.803101*** ΔAR t-2 -0.190873*** 0.095524 ΔAR t-3 0.244711 ΔINV t-1 0.969675*** 0.856248*** 0.871865*** ΔINV t-2 0.581657*** 0.889791*** ΔINV t-3 0.474898*** ΔOTH t-1 -0.449346 -0.926885** -1.148572** ΔOTH t-2 0.440919 0.353376 ΔOTH t-3 -0.32586 DPRM t-1 2.492599*** 1.552724*** 0.321975 DPRM t-2 1.097229*** 0.780995** DPRM t-3 1.647527*** Prob (F- Statistic) 0.0000 0.0000 0.0000 DW 2.484538 2.548625 2.501873 Adjusted R2 0.883925 0.922246 0.93063 *Significant at 0.1 level. ** Significant at 0.05 level. Significant at 0.01 level. Table 6 shows FEM regression results of three detailed prediction models (with one-year lag, two-year lag and three- year lag in the predictors). Particularly, disaggregated accruals are capable of predicting future operating cash flows. This conclusion is similar to prior studies such as Barth et al. (2001) and Ebaid (2011), among others. Moreover, the model predictability is stronger if the lagging period of forecast variables is longer (the model predictive value increases from 88% to 93%). Thus, the fourth research hypothesis (H4) is accepted. FEM regression results of the cash flow ratio model The relationship between future operating cash flows and cash flow ratios (with one-year lag and two-year lag) is shown in Table 7. FEM regression results show the predictivity of the cash flow ratio model. However, the predictive values lower than those in the previously presented earning models and cash flow models (the highest predictability is 51.8% in the cash flow ratio model with two-year lagging predictors). Thus, the fifth research hypothesis (H5) is accepted. 692 Table 7 Summary of FEM regression results of the cash flow ratio model Variables One-year lag Two-year lag CFR1t-1 0.000343 0.000225 CFR2t-1 -0.000016 -0.000098** CFR3t-1 -0.000389 0.000180 CFR4t-1 0.000601* 0.001132*** CFR5t-1 -0.000150* -0.000414*** CFR6t-1 0.002356* 0.005206*** CFR7t-1 -0.000567 -0.000538 CFR8t-1 -0.000072 -0.000023 CFR9t-1 -0.039756* -0.055162** CFR1t-2 0.000223 CFR2t-2 0.000060 CFR3t-2 0.001788 CFR4t-2 -0.008647*** CFR5t-2 -0.000131 CFR6t-2 -0.005951** CFR7t-2 -0.001585 CFR8t-2 -0.000340 CFR9t-2 -0.053952** Prob(F- Statistic) 0.000000 0.000000 Adjusted R2 0.333233 0.518494 *Significant at 0.1 level. ** Significant at 0.05 level. Significant at 0.01 level. Evaluation of predictability in cash flow forecasting models A summary of Adjusted - R2 coefficients in the models is presented in Table 8. Table 8 Adjusted R2 values in prediction models No. Forecasting Model Adjusted R2 value 1 Earnings model (with one-year lag) 0.773219 2 Earnings model (with two-year lag) 0.792889 3 Earnings model (with three-year lag) 0.813363 4 Cash flow model (with one-year lag) 0.764312 5 Cash flow model (with two-year lag) 0.789254 6 Cash flow model (with three-year lag) 0.796054 7 Aggregated accrual model (with one-year lag) 0.828346 8 Aggregated accrual model (with two-year lag) 0.828309 9 Aggregated accrual model (with three-year lag) 0.875557 10 Disaggregated accrual model (with one-year lag) 0.883925 11 Disaggregated accrual model (with two-year lag) 0.922246 12 Disaggregated accrual model (with three-year lag) 0.930630 13 Cash flow ratio model (with one-year lag) 0.333233 14 Cash flow ratio model (with two-year lag) 0.518494 As shown in Table 8, the models with longer-lagging independent variables are more predictive. The models with independent variables lagging three years indicate higher predictivity than those with independent variables lagging one or two years. This conclusion is consistent with the statement given by different researchers such as Barth et al. (2001), Chong (2012), etc. Barth et al. (2001) found the R2 coefficient increased from 27% to 35% in the regression model of cash flow and accruals when the lag of independent variables is longer. Therefore, the sixth research hypoth- esis (H6) is accepted. 6. Conclusion and Recommendations 6.1 Conclusion The regression results of five cash flow forecasting models (historical earnings, historical cash flows, aggregated ac- cruals, disaggregated accruals, and cash flow ratios) point out the highest Adjusted R2 coefficient in the model of cash flows and disaggregated accruals. Therefore, this model outperforms the remaining ones in terms of predictability. H. A. Nguyen and T. H. Nguyen / Management Science Letters 10 (2020) 693 The historical cash flows combined with disaggregated accruals explain 88% to 93% of changes in future net operating cash flow. The results support the conclusion by Barth et al. (2001) and Ebaid (2011). The earnings model generates improved predictive values rather than the cash flow model. This empirical result is in favour of Greenberg et al. (1986), Barth et al. (2001), Ebaid (2011), and Anh (2013), and inconsistent with the observations by other researchers such as Mooi (2007). 6.2. Recommendations From the research results, the research team proposed some recommendations to the Vietnam as follows: Firstly, to the business: Results of this empirical research highlight the significance of accounting information on future cash flow forecasts with the predictive ability range between 51% and 93%. Business managers are recom- mended, therefore, to leverage the cash flow forecasting method introduced in this study in addition to traditional forecasting methods to better inform the decision-making. To mandate the forecast as a prior action to decision making, firms are encouraged to create a forecast team, in which specialists should be collaborative, connective and interactive. In addition, companies should foster professional skills for the staff in charge of cash flow forecasting. Besides, they need to create a positive work environment, closely coordinate accountants and forecasters to share detailed accounting data in order to make the cash flow prediction. Secondly, to the Ministry of Finance: It is suggested that Ministry of Finance in Vietnam request businesses to supplement cash flows ratios to the “Financial ratios” section in the Note to Financial Statements to inform a broad range of accounting data users. It is shown in the study that the explanatory values of cash flow ratios such as Cash flow from operating activities/Total assets, Payment for fixed assets/Cash flow from operating activities; Depreciation of fixed assets/Cash flow from operating activities are as high as 51% to future cash flows. The disclosure of cash flow ratios is not only useful to the policy development of state management agen- cies, but also the understanding of investors with regards to financial performance of enterprises is necessary. Moreo- ver, accounting policies and regimes should be introduced to regulate earnings management techniques in companies in the preparation of financial statements in order to improve the accuracy of cash flow forecasts. The forecasts of operating cash flow will be accurate if they are sourced from accurate earning data. More efforts from the Ministry of Finance are encouraged to promulgate Accounting Standards and corporate accounting regime, which respect the nature rather than the form of business operations, to reduce the managerial discretion in processing complicated transactions. In addition, the Ministry of Finance is suggested to create opportunity for listed companies to make cash flow forecasts. In Vietnam, the forecast of cash flows is not fully backed by an enabling environment. In accordance with the Circular No. 52/2012/TT-BTC, guiding the disclosure of information on securities market, the Ministry of Finance does not require and encourage listed firms to disclose a pro forma cash flow statement. Therefore, businesses do not voluntarily create their pro forma cash flows, thus both businesses and investors are affected due to a lack of information. Companies with poor business performance in the U.S show the willingness to publish cash flow forecasts because their positive prediction of cash flows was perceived as good signals and investors were more confident in future business prospects and sustainability. Consequently, the forecasts of cash flows are critical to Vietnam to sup- port the efficient investments of investors attributed to better information capture, the improved transparency in the stock market, the strengthened global integration, the enforced international commitments and the realized stock mar- ket upgrading solutions. References Al-Attar, A. & Hussain, S. (2004). Corporate data and future cash flow. Journal of Business Finance and Accounting, 31(7- 8), 861-903. Ali, A. (1994). The incremental information content of earnings, working capital from operations, and cash flows. Journal of Accounting Research, 32, 61-74. Anh, N.H. (2010). Using some accounting information to predict company’s operating cashflows. Journal of Economics and Development, 162(2), 13-17 Anh, N.H. (2013). Forecasting cash flows from operating activities of listed companies on Hochiminh city Stock Exchange. Journal of Economics and Development, Special Issue, 20-27 Arnold, A.J., Clubb, C.D.B., Manson, S., & Wearing, R.T. (1991). The relationship between earnings, funds flows and cash flows: evidence for the UK. Accounting and Business Research, 22(85), 13-19. Barth, M. E., Cram, D. P., & Nelson, K. K. (2001). Accruals and the prediction of future cash flows. Accounting Review, 76(1), 27-58. Block, B. (1999). A study of financial analysts: Practice and theory. Financial Analysts Journal, 55, 86-95. Bowen, R. M., Burgstahler, D. & Daley, L. A. (1986). Evidence on the relationships between earnings and various measures of cash flow. The Accounting Review, 61(4), 713-25. Call, A., Chen, S., & Tong, Y. (2009). Are analysts’ earnings forecasts more accurate when accompanied by cash flow forecasts? Review of Accounting Studies, 14, 358-392. Chuong, O.N & Quynh, N.T (2015). Research on diversity of income sources of households: the case of middle southern area of Vietnam. Journal of Economics and Development, 217, 17 – 23 694 Dechow, P.M., Richardson, S.A., & Sloan, R.G., (2008). The persistence and pricing of the cash component of earnings. Journal of Accounting Research, 46(3), 537- 566. Dechow, P. M., Kothari, S. P., & L. Watts, R. (1998). The relation between earnings and cash flows. Journal of Accounting and Economics, 25(2), 133-168. Dechow, P.M. (1994). Accounting earning and cash flows as measures of firm performance: the role of accounting accruals. Journal of Accounting and Economics, 18(1), 3-42. Dechow, P.M., & Dichev, I.D. (2002). The quality of accruals and earnings: the role of accrual estimation errors. The Ac- counting Review, 77 (supplement), 35-59. Duong, D.T & Hung, D.T (2014). Capital structure and profitability of listed companies on Vietnam Stock Market. Journal of Economics and Development, Special Issue in December, 54-60 Ebaid, I. E. (2011). Accruals and the prediction of future cash flows: empirical evidence from an emerging market. Man- agement Research Review, 34(7), 838-853. Farshadfar, S., Ng, C. & Brimble, M. (2008). The relative ability of earnings and cash flow data in forecasting future cash flows. Pacific Accounting Review, 20(3), 251-268. Finger, C. A. (1994). The ability of earnings to predict future earnings and cash flow. Journal of Accounting Research, 32(2), 210-23. Fulmer, J.G., Gavin, T.A., & Bertin, W.J. (1991). What Factors Influence the Lending Decision: A Survey of Commercial Loan Officers. Commercial Lending Review, 7(1), 64-70. Greenberg, R. R., Johnson, G. L. & Ramesh, K. (1986). Earnings versus cash flows as a predictor of future cash flow measures. Journal of Accounting Auditing and Finance, 1(4), 266-77. Habib, A. (2010). Prediction of operating cash flows: further evidence from Australia. Australian Accounting Review, 20(2), 134. Jordan, C.E., Waldron, A., & Clark, S.J. (2007). An analysis of the comparative predictive abilities of operating cash flows, earnings and sales. The Journal of Applied Business Research, 23(3),53-60. Kelly, J., & O'Connor, J. (1997). Is profit more important than cashflow?. Management Accounting: Magazine for Chartered Management Accountants, 75(6), 28-29. Kim, M. & Kross, W. (2005). The ability of earnings to predict future operating cash flows has been increasing – not de- creasing. Journal of Accounting Research, 43(5), 753 -780. Largay, J., & Stickney C. (1980). Cash flows, ratio analysis and the W.T. Grant Company bankruptcy. Financial Analysts Journal, 36(4), 51-54. Lorek, K., & Willinger, G. (1996). A multivariate time-series prediction model for cash-flow earnings data. Accounting Review, 71, 81–101. McBeth, K. H. (1993). Forecasting operating cash flow: Evidence on the comparative predictive abilities of net income and operating cash flow from actual cash flow data. The Mid-Atlantic Journal of Business, 29(2), 173-87. Mooi, T. L. (2007). Predicting future cash flows: Does cash flow have incremental information over accrual earnings? Ma- laysian Accounting Review, 6(2), 63-80 Murdoch, B. & Krause, P. (1989). An empirical investigation of the predictive power of accrual and cash flow data in forecasting operating cash flow. Akron Business and Economic Review, 20(3), 100-13. 301. Murdoch, B. & Krause, P. (1990). Further evidence on the comparative ability of accounting data to predict operating cash flows. The Mid - Atlantic Journal of Business, 26(2), 1-14. Pae, J., & Yoon, S.S (2012). Determinants of Analysts’ Cash Flow Forecast Accuracy. Journal of Accounting, Auditing & Finance, 27(1), 123-144 Percy, M., & Stokes, D.J. (1992). Further evidence on empirical relationships between earnings and cash flows. Accounting and Finance, 32(1), 27-49. Pfeiffer, R. J., Elgers, P.T., Lo, M. H. & L. L. Rees (1998). Additional evidence on the incremental information content of cash flows and accruals: the impact of errors in measuring market expectations. Accounting Review, 73(3), 373 – 385. Plewa, F.J. & Friedlob, G.T. (1995). Understanding Cash Flow. Wiley, New York Quirin, J. J., O'Bryan, D., Wilcox, W. E. & Berry, K. T. (1999). Forecasting cash flow from operations: additional evidence. The Mid - Atlantic Journal of Business, 35(2-3),135-42 Stammerjohan, W. W. & Nassiripour, S. (2001). Predicting SFAS 95 cash flows: the relative importance of prior earnings, cash flows and accruals. Accounting Enquiries, 10(1), 87-146. Waddell, D., & Sohal, A.S. (1994). Forcasting: The key to Managerial Decision Making. Management Decision, 32(1), 41 – 49. Wilson, G. (1986). The relative incremental information content of accruals and cash flows: Combined evidence at the earnings announcement and annual report date. Journal of Accounting Research, 24, 163-200. © 2020 by the authors; licensee Growing Science, Canada. This is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC- BY) license (