Determinants of behaviour of working capital requirements of BSE listed companies: An empirical study using co-integration techniques and generalised method of moments

Abstract The paper investigates the determinants of working capital to forecast the future requirement of working capital of BSE-listed top 150 companies in India. The study is conducted by collecting the data of 150 top-performing BSE listed companies for the time period of 2009–2017. The ten firm-level explanatory variables and one dummy variable to characterize the nature of business i.e. manufacturing and service sector companies are used to discover the significant determinants of gross and net working capital of selected companies. The four alternative methods are used to verify and validate the results obtained from each other viz. Ordinary Least Square (OLS), fully modified OLS (FMOLS), dynamic OLS (DOLS) and generalized methods of moments (GMM). The empirical findings of four different methods indicate that tangibility, leverage, nature of business and board size are observed as significant factors to forecast the future requirements of net working capital. Return on common equity (ROCE), board size; indicate the positive association with gross working capital. On the other hand, tangibility, nature of business and size of the firm show a negative relationship with gross working capital requirements. The findings from different methods have resulted in similar significant determinants to predict the future requirement of the net and gross working capital. The outcome of the study will be useful for management authorities for maintaining the optimum level of working capital and to forecast future requirements of working capital.


PUBLIC INTEREST STATEMENT
Efficiently managing working capital is a great challenge for every organization. Every organization intends to maintain the optimum level of working capital. Since the excessive working capital creates an idle fund and inadequate working capital interrupts the day to day operation of the business. For this purpose, it becomes much essential to investigate certain determinants to forecast determinants of working capital to forecast the future requirement of working capital. The study has been conducted after taking into consideration the recent historical economic reforms in the eminent Asian's emerging economy. The study investigates many authentic determinants of working capital after employing four alternative econometric techniques to forecast the future requirement of working capital. The study will be mainly lucrative to policymakers to design optimum level of working capital and academician as it adds significantly to the existing body of knowledge association with gross working capital. On the other hand, tangibility, nature of business and size of the firm show a negative relationship with gross working capital requirements. The findings from different methods have resulted in similar significant determinants to predict the future requirement of the net and gross working capital. The outcome of the study will be useful for management authorities for maintaining the optimum level of working capital and to forecast future requirements of working capital.

Introduction
In recent years Indian economy has experienced tremendous growth. During 2014-2017, gross domestic product (GDP) has remained quite high and there was a huge jump in the BSE and NSE Sensex values. It was due to combined performance of the manufacturing and service sectors. After 2014, Indian economy witnessed historical economic reforms. These reforms affected the different sectors of the Indian economy in a significant manner. Demonetization in 2016 affected the manufacturing and service sectors. In 2017, the implementation of Goods and Services Tax Act also affected both these sectors. The present government has brought favorable policy to boost the growth of various sectors. Recently in 2019, the government slashed the corporate tax rates. These developments have encouraged researchers to explore different areas. Investigating the determinants of working capital is one such area. The working capital influences the day to day operation of the business. To maintain an optimum level of working capital to enhance the performance of the business is certainly the prime concern of every company's decision-makers. Since we know that inadequate or excessive working capital are both dangerous for the company's financial health. Inadequate levels of working capital hinder day to day operations of the business and excessive working capital creates idle funds. Every organization wants to avoid these two situations. Every company wants to keep an adequate level of current assets to meet the expenses of day to day operations. It is thus essential to investigate the key determinants of working capital requirements. The present study tries to answer the following research questions: (1) what are the key determinants of gross and networking capital after the historical economic reforms in India.
(2) How these determinants are helpful to management authorities for maintaining an optimum level of working capital. To find the answer to these questions, the study is carried out using four alternative techniques to identify the key determinants of working capital.
Liquidity always remains the prime concern for all the undertakings whether dealing with the manufacturing or service sector. All the firms would like to maintain that level of liquidity which will enable them to carry on day to day transactions smoothly. Usually, manufacturing companies need more liquid assets as these companies have to incur various expenses in day to day basis. Efficient management of liquid assets or current assets helps the firm to increase profitability and sustain itself in the business world for a longer period of time. The problem of liquidity may put obstacles for doing different types of business activities which results in a decline in productivity as well as the performance of the business. Similarly, there should not be excessive working capital as it remains idle and puts the burden on business which ultimately affects profitability. A firm's net working capital position influences its ability to obtain debt financing as many loan agreements with financial institutions require a firm to maintain a minimum net working capital position (McGuigan, Moyer, Rao, & Kretlow, 2012). The management of current assets is similar to that of fixed assets in the sense that both affect return and risk. The management of fixed assets and current assets differs in significant ways: First for managing fixed assets time plays a vital role, so discounting and compounding techniques play a very important role in capital budgeting and a minor one in the management of current assets. The second: high level of current assets, specifically cash strengthen the liquidity position of the company but also reduces the profitability as excessive working capital result in idle capital. Third, the level of current assets and fixed assets depends upon future sales, but only current assets can be adjusted due to fluctuation in sales in the short run. (Pandey, 2007).
Working capital management should be more effective as it prevents the company from liquidity crises and also affects profitability, solvency position, and survival of the company (Mukhopadhyay, 2004). Effective working capital management depends upon the planning and controlling of current assets and current liabilities. As it is the most important factor responsible for the elimination of the risk of failure to meet short term obligations and additionally to avoid excessive investment in the assets (Eljelly, 2004). Working capital mainly manages the optimal difference between its components i.e. cash, receivables, payables and inventory, etc which is the cause of the success of any business (Filbeck & Krueger, 2005).
The developing economies generally face the problem of underutilization of resources. While fixed capital and working capital are both contributors to the total capital of the developing country, working capital makes utilization of the production capacity generated by the fixed assets possible (Vijayakumar & Venkatachalam, 1996).

Background and hypothesis development
An organization is required to maintain a certain level of working capital. This is also required for taking loans from financial institutions. Organizations focus on inventory management for improving their working capital performance. This helps them in reducing the working capital investment. There are several factors that impact working capital management (Wasiuzzaman & Arumugam, 2013). Working capital management is done by monitoring the current assets and current liabilities of the organization (Atseye, Ugwu, & Takon, 2015). This is required for the proper functioning of daily operations and maintaining sufficient cash flow. Working capital investments should be kept minimum otherwise they decrease the profitability of the organization (Manoori & Muhammad, 2012). It is the responsibility of managers to maintain an optimum level of working capital in order to minimize risks . The literature review has been arranged as per the different determinates used in the present study.

Leverage
Working capital has a negative relationship with leverage. An organization opts for external financing when it does not have sufficient internal financing. An organization with more debt has less working capital Salawu & Alao, 2014). The organization has less money for day-to-day operations when it has to pay debts. In such a scenario, the organization is expected to utilize the working capital in an efficient manner. Thus, organizations with high debt ratios are cautious about working capital (Wasiuzzaman & Arumugam, 2013).
In the present study, leverage has been measured through the debt to equity (D/E) ratio. It is expected that leverage (LEV) has a negative relationship with both types of working capital used in the current study i.e., gross working capital and net working capital. Alternative Hypothesis (H 1a ): Leverage has a significant negative influence on the level of working capital.

Growth opportunities
There are contradictory arguments regarding the relationship between working capital and the growth of an organization. There are studies that demonstrate a positive relationship between sales growth and working capital (Manoori & Muhammad, 2012;Wasiuzzaman & Arumugam, 2013). An organization that expects growth in sales will increase its level of inventory and short-term investments. This will lead to an increase in working capital. Thus, growth in sales leads to increased working capital (Wasiuzzaman & Arumugam, 2013). This positive relationship between working capital and sales growth was also observed in a study related to Nigerian manufacturing firms (Salawu & Alao, 2014). But, several studies have also established a negative relationship between the revenue growth rate of an organization and working capital (Wasiuzzaman & Arumugam, 2013).
Revenue growth has been used as the proxy of immediate investment or growth opportunities for the selected companies used in the study. The same measure was also used by (D'Mello, Krishnaswami, & Larkin, 2008). Revenue for the current period or t period has been divided by revenue of the previous period i.e., t-1 period. Then, the natural logarithm of the calculated value has been taken to find the revenue growth rate of BSE listed 150 companies including BSE Sensex 30 companies. There is an absolute consensus regarding the type of relationship of revenue growth rate with two different types of working capital measures used in the present research. An alternative hypothesis has been built in the following way: Hypothesis (H 2a ): Growth opportunity has a significant impact on the level of Working capital.

Asymmetric information
It is difficult for less transparent organizations to obtain external financing. Information asymmetry is lower for organizations with more tangible assets. Organizations with higher information asymmetry have to focus more on internal resources. There is a negative relationship between information asymmetry and working capital. Thus, organizations with a high level of asymmetric information have lower working capital (Wasiuzzaman & Arumugam, 2013). Hill, Kelly, and Highfield (2010) measure asymmetric information as the book-to-market ratio (MTB), especially in connection with the long-term investment opportunities of a firm. In the present study, the same measure has been used as a proxy of asymmetric information. In the present study, the market value of equity has been calculated by taking closing prices with BSE at the end of the respective financial year. On the other handbook, the value of equity has been calculated by taking all total assets except fictitious assets of the respective financial year and subtracting all long term debt and current liabilities. Then, market to book value (MTB) has been calculated by dividing the market value of equity by book value of equity. From the above discussion and study, it is expected that asymmetric information may have a negative impact on the working capital of the firm. Accordingly, an alternative hypothesis has been formulated: Hypothesis (H 3a ): The level of asymmetric information has a negative impact on the future working capital requirement of the firm.

Nature of business
The nature of business is an important determinate of working capital. The future requirement of working capital in firms may differ as per the nature of the business. In the BSE Sensex, there is a combination of two types of companies that companies belonging to the manufacturing sector and service sector. In the present study, for quantifying the nature of business dummy variables have been used as 1 has been assigned to manufacturing companies and 0 to service sector companies. The working capital management differs from industry to industry. In this way, the nature of the business may be addressed as one of the determinants of working capital. A simple comparison of the service industry and manufacturing industry can clarify the point. In the service industry, there is no inventory and therefore, one big component of working capital is already avoided. So, the nature of the industry is a factor in determining the working capital requirement (Efinance Management, 2018) It is expected that the nature of business has a negative impact on the future requirements of BSE listed companies (Banerjee, 2017).
Hypothesis (H 4a ): The nature of business has a negative influence on the future requirements of working capital in BSE listed companies.

Size or capital market access
In comparison to small organizations, large organizations have better access to capital. Large organizations have better growth opportunities and require more money for their daily operations. But they are better at cash management. Thus, a negative relationship has been observed between working capital and the size of the organization. Larger organizations can afford to invest less money in working capital (Manoori & Muhammad, 2012;Wasiuzzaman & Arumugam, 2013). Small organizations require higher working capital as they are more vulnerable to risk (Bereznicka, 2014). But, a study conducted in Nigeria established a positive relationship between working capital and the size of the organization. This is because large organizations can make working capital investments due to plenty of financing alternatives (Salawu & Alao, 2014).
Hypothesis (H 5a ): Size or Capital Market Access has a positive impact on the future working capital requirement.

Asset tangibility
A negative relationship is expected between working capital and asset tangibility. Limited financial resources are available which have to be invested wisely. Organizations choose to invest in areas that offer high returns. Organizations prefer making long term investments thus reducing working capital investment (Wasiuzzaman & Arumugam, 2013). Assets tangibility has been measured by the ratio of net fixed assets to the total value of assets during the study period of selected BSE listed companies. Banos -Caballero, Teruel, and Martínez (2010) used this variable to predict working capital but they observed that the effect of tangibility on working capital investment is not clear.
Hypothesis (H 6a ): Asset Tangibility has a negative impact on the future working capital requirement of the firm.

Revenue volatility
Revenue volatility is associated with sales volatility. It becomes difficult to forecast revenue generation due to sales volatility. This financial parameter is an indicator of operating risk. Organizations with greater revenue volatility tend to invest less money in working capital. Thus, a negative relationship is observed between revenue/sales volatility and working capital investment (Wasiuzzaman & Arumugam, 2013). Hill et al. (2010) find varied results for the relationship between working capital requirements and revenue volatility based on past literature. They argue that with higher revenue volatility, or high deviations in demand, coming up with the optimal level of inventory may be difficult and so to be on the safe side, firms could increase inventory levels.
Hypothesis (H 7a ): Revenue Volatility has a negative impact on the future working capital requirement of the firm.

Profitability
Profitability has a significant impact on working capital. Organizations with high profitability are likely to invest more in working capital due to the availability of cash. Thus, there is a positive relationship between working capital investment and profitability of the organization (Manoori & Muhammad, 2012;Wasiuzzaman & Arumugam, 2013). Whereas, a study conducted in Athens established a negative relationship between profitability and working capital . High profitability helps firms strongly to bargain with both vendors and customers, and firms can use these competitive advantages to improve their liquidity (Petersen & Rajan, 1997;Shin & Soenen, 1998). Profitability has a significant positive impact on working capital requirements. More investment in working capital means more sources engaged and make more opportunity cost for firms (Deloof, 2003).In the present study return on common equity (ROCE) has been used to measure the profitability of selected organizations.
Hypothesis (H 8a ): Profitability has a positive influence on the future requirements of working capital in BSE listed companies.

Operating cash flow
Operating cash flow has a significant impact on working capital. Organizations with better operating cash flow tend to increase their working capital investment Wasiuzzaman & Arumugam, 2013). On the contrary, some researchers believe that operating cash flow has a negative relationship with working capital. Better cash flow results in reduced working capital investment (Manoori & Muhammad, 2012). Appuhami (2008) and Chiou, Cheng, and Wu (2006) explored the significant negative relationship between operating cash flows (OCFA) and working capital requirements. But other researchers did not find any statistically significant relationship between working capital and operating cash flows (Al Taleb, AL-Zoued, & AL-Shubiri, 2010; Nazir & Afza, 2008).
Hypothesis (H 9a ): Operating cash flow has a significant (positive or negative) influence on the future requirements of working capital in BSE listed companies.

Board characteristics or board size
A study of 192 companies in Malaysia established that the board characteristics (board size and board independence) do not have a significant impact on working capital investment. This may be due to the fact that board members do not make decisions regarding working capital. Also, the decisions made by the board members on significant issues do not impact working capital investments (Wasiuzzaman & Arumugam, 2013).
Hypothesis (H 10a ): Board Characteristics or Board Size has a significant influence on the future requirements of working capital in BSE listed companies.

Economic conditions
A study conducted in Singapore found that GDP has a negative impact on working capital investment (Manoori & Muhammad, 2012). On the contrary, a study conducted in Nigeria established a positive relationship between working capital and GDP (Salawu & Alao, 2014). During the recession, the organizations may choose to invest more in working capital for the smooth functioning of day-today operations. But some researchers have a different point of view. They argue that economic conditions have a negative impact on working capital investments. The organizations reduce working capital investments due to financial constraints during the recession. They believe that working capital investments increase during economic growth (Wasiuzzaman & Arumugam, 2013).
Hypothesis (H 11a ): Economic conditions significantly affect the future requirements of working capital in BSE listed companies.

Objectives of the study
The following are the main objectives of the study: O1: To identify the determinants of gross working capital requirements of selected BSE listed companies.
O2: To identify the determinants of net working capital requirements of selected BSE listed companies.
O3: To provide useful suggestions to the board of directors companies for forecasting working capital requirements and to arrive at an optimum level of working capital.

Research methods
The present study is an empirical analysis of determinants of working capital of the top 150 BSE listed companies including BSE Sensex 30 companies using the most recent available data. There are around 5000 companies listed with the Bombay Stock Exchange at the time of the study. A sample of 150 companies has been taken as these companies are top-performing companies and these companies represent the entire economy as well as different industries. These companies engaged in diverse business viz., banking and finance, FMCG, power and infrastructure, telecommunication, refinery, pharmaceuticals, Information technology, etc. For arriving at good analysis and an adequate number of observations for the study more recent data have been taken off the last 9 years. Secondary data related to selected companies have been taken from the financial statements available at the prowess database of the Centre for Monitoring Indian Economy (CMIE). Data have been collected for the last nine years i.e., from 2009 to 2017. Selected companies have age either 15 years or more when data was retrieved from the abovementioned sources. In the present paper eleven exogenous and two endogenous variables (NWC and GWC) are used. These exogenous and endogenous variables are stated in Table A1. In the present study, four different methods have been used to develop models to forecast the future requirement of working capital of selected companies. The first method is the ordinary least square method (OLS). The paper employs two cointegration techniques (FMOLS and DOLS) and generalized method of moment (GMM). We consider alternative methods of panel data estimation to check whether the findings from the methods are consistent and justify the relevance of the findings from the earlier studies. Before applying cointegration techniques (FMOLS and DOLS), it is mandatory to check that data series of every endogenous and exogenous variables should be stationary. There is also a second essential condition that there should be cointegration in the model which is supposed to be developed using FMOLS and DOLS. To check stationary of data series, three different unit root tests have been applied viz., ADF (Dickey & Fuller, 1979), PP (Philips & Perron, 1988) and KPSS (Kwiatkowski, Phillips, Schmidt, & Shin, 1992). These tests have been applied without constant (none) and trend (T), with constant and constant and trend (C and T). Johansen cointegration unrestricted rank test is performed by taking the criteria of trace value and maximum Eigenvalue. After meeting these two conditions, FMOLS and DOLS have been applied to develop models to forecast future requirements of net working capital and gross working capital. The determinants observed in this study can be utilized to know the future working capital requirement selected BSE listed companies. But the interpretation can be made for the whole set of Indian companies listed with BSE. The measures of working capital have been identified from the studies conducted across the globe. At the initial stage, the various determinants of working capital are pulled from the previous studies conducted around the world. Consequently, OLS, cointegration tests (FMOLS, DOLS) and GMM are employed to determine whether there exists a relationship between the multiple explanatory variables. Regression equations are formed to develop the models using firm-specific determinants and Eviews10 software has been used to obtain the results of four different methods.

Ordinary least square vs cointegration tests
Dynamic OLS and fully modified OLS are better than ordinary least square methods due to a number of reasons. First OLS estimates are super-consistent, but the t-statistic obtained without stationary are only approximately normal. Even though, OLS is super-consistent, in the presence of a large finite sample bias' the convergence of OLS can be low in finite samples.
Second OLS estimates may suffer from heteroskedasticity autocorrelation since the omitted dynamics are captured by the residual so that inference using the normal tables will not be valideven asymptotically. Therefore, "t" statistics for the estimates OLS estimates are useless.
Dynamic and fully modified OLS take care endogeneity by adding the leads and lags (DOLS). In addition, white heteroskedastic standard errors are used. FMOLS does the same using a nonparametric approach (Arize, John, & Ghosh, 2015) and (Arize, Thomas, & Slottje, 2000).
Both the cointegration tests DOLS and FMOLS are proffered over OLS estimator as they take into consideration small sample bias and endogeneity bias by using leads and lags of the first-differenced regressors. However, the parametric DOLS is prioritized over nonparametric FMOLS. As the later imposes additional conditions that all the variables should be integrated of the same order [i.e., I (1)] and that the regressors themselves should not be cointegrated (Masih & Masih, 1996).

Equations for OLS, FMOLS and DOLS
(a) NWC = α+ β1 (Nature of Business) + β2(Growth) + β3(Profitability) + β4(Size)+ β5(Assets Tangibility) + β6(Leverage) + β7(Revenue Volatility)+ β8(Operating Cash Flows)+ β9(Board Size)+ β10(Asymmetric Information)+ β11(Economic Condition)+e In the above equations α is the intercept, β1 to β16 are the slope of different exogenous variable and e is the error term in the models 4.2. Generalised method of moments Hansen (1982) first introduced the generalized method of moments (GMM). GMM is one of the main statistical tools for the analysis of economic and financial data. GMM can be applied crosssectional data, time-series data, and panel data analysis Generalized Method of Moments (GMM) is an estimation procedure that allows economic models to be specified while avoiding often unwanted or unnecessary assumptions, such as specifying a particular distribution for the errors.
This method requires that a certain number of moment conditions are to be specified for developing a model using GMM. These moment conditions are functions of the model parameters and the data, such that their expectation is zero at the parameters' true values. The GMM method then minimizes a certain norm of the sample averages of the moment conditions. The GMM estimators are known to be consistent, asymptotically normal, and efficient in the class of all estimators that do not use any extra information aside from that contained in the moment conditions. The specification we use for this purpose for the two models takes the following general form: Leverage it , Growth it , Revenue Volatility it , operating cash Flow it , Tangibility, it )+ n i +n t +ε it GWCit = α 0 + f (GWC it−1, it , Age it , BOD Size it , Price Earning Patio it , Size it , Profitability(ROCE) it , Leverage it , Growth it , Revenue Volatility it , operating cash Flow it , Tangibility, it )+ n i +n t +ε it Where NWCit and GWCit are the net working capital and gross working capital ratios of the firm I in year t, n i is the unobserved firm-specific effects, n t is the time-specific effects and εit is the error term. The remaining variables in the above equations are already used as explanatory variables in the previous two methods of estimation except for the nature of the business. Due to adding the lagged value of the dividend payout ratio as an exogenous variable in the model, there was problem of multicollinearity due to the nature of the business which was taken as a dummy predictor variable in the previous two models. Now, this variable has been removed as the results of the GMM model were not coming due to the high level of multicollinearity. Unobserved factors of the firm which may have a significant impact on a firm's dividend payout ratio are expressed as n i . n t expresses the effects of macroeconomic factors such as interest rate, inflation, etc.
OLS estimation technique remained biased in the presence of firm-specific effects, by presuming that n i is unobservable and there is covariance between predictors variables and unobserved characteristics (n i ) (Hsiao, 1985).In this situation, an instrumental variable estimation method provides consistent results if the error term (ε it ) are serially uncorrelated. Taking the assumption that error term serially uncorrelated, the generalized method of the moment is proposed by Arellano and Bond (1991). GMM is the most efficient method in the class of instrumental variable estimators (Honore & Hu, 2003). In the present study, the two-step GMM estimation has been used. It is observed by many previous studies that two-step GMM is more efficient than the onestep estimation when the disturbances are expected to show heteroscedasticity in large sample data with a relatively long time period (Arellano & Bond, 1991;Blundell & Bond, 1998).
However, the effectiveness of GMM estimation depends upon certain assumptions that lagged value of the dependent variable and other predictors' variables should be valid instruments and error terms of model do not exhibit autocorrelation. For this purpose, Arellano and Bond (1991) proposed three tests. The first is to test that there is not first-order serial correlation or autocorrelation of the error term of the proposed model. Under the null hypothesis of no serial correlation, the test statistic is distributed as a standard normal. Second is to check that there should not be second-order autocorrelation of the error term of the GMM model, which is distributed as a standard normal under the null hypothesis of no serial correlation. The third is the Sargan (1958) test or J statistics of over-identifying restrictions. This tests the validity of the instruments and is asymptotically distributed as Chi-square under the null of instrument validity.

Analysis and interpretation
The descriptive statistics of all eleven explanatory and two endogenous variables (GWC and NWC) are presented in Table A2. The mean value of the proportion of gross and networking capital (GWC NWC) to total assets has remained 0.57 and 0.07 respectively during the study period. The standard deviation of these two measures of working capital is quite high which reflects high volatility in the gross and networking capital of selected companies. Specifically the case of NWC, volatility is quite high.
The average board of directors (BOD size) of selected BSE listed companies has remained 11 with a standard deviation of 2.90. The average GDP of the country has remained 0.07 or 7% with a standard deviation of 0.01 or 1%. It shows much consistency in terms of the economic growth of the Indian economy. Growth could be an important indicator to predict future working capital requirements of selected organizations. Growth, as measured through the natural logarithm of revenue of selected BSE companies, has remained quite high i.e., 0.12 or 12% with very high standard deviation. Leverage, as measured through total debt/total assets, is 0.33 but it also shows high dispersion as revealed from standard deviation. For studying the nature of the business, dummy variables have been used (Where 1 for manufacturing and 0 for service Sector Company) average score of nature of the business is 1 and the standard of deviation is 0.48. Operating Cash Flow (OCFA) has mean value 0.20 and the standard deviation is quite high (1.01). Market to book value (MTB) has an average score of 5.64 with a high standard deviation. Revenue volatility (REVVOL) as measured through the coefficient of variation of net sales of the rolling period of 4 years has an average value of 0.21 with a standard deviation of 0.11. The average return on common equity (ROCE) is quite high (22.29 %) of selected BSE listed companies with a high standard deviation. The size of the firm as measured through the natural log of total assets is 10.56 with a low standard deviation. Assets tangibility (TANG) as measured by net fixed assets to total assets (NFATA) has a mean value of 0.28 of selected BSE listed companies with high volatility as revealed from high standard deviation.
According to (Malhotra & Dash, 2009) to test the assumption of a normal distribution, Skewness should be within the range ±1 and Kurtosis value should be within range ±3. Some people use ±2 range of Kurtosis.
Further, the most widely used method, at least in econometrics, that has been suggested and used for testing whether the distribution underlying a sample is normal is the (Bowman & Shenton, 1975) statistic, which subsequently was derived by Bera and Jarque as the Lagrangian Multiplier (LM) test against the Pearson family distributions. For that reason, the JB test is also referred to as the Jarque-Bera test (Bera & Jarque, 1982;Bowman & Shenton, 1975;Jarque & Bera, 1987;Shenton & Bowman, 1977). The JB statistic has an asymptotic chi-square distribution with two degrees of freedom. The JB test is simple to compute and its power has proved comparable to other powerful tests. It turns out that this test statistic can be compared with a χ 2 (chi-square) distribution with 2 degrees of freedom. The null hypothesis of normality is rejected if the calculated test statistic exceeds a critical value from the χ 2 (2 degrees) distribution. JB test statistic is much greater than 5.99 of all dependent and independent variables (chi-square table value at 2 degrees of freedom), that the null hypothesis of normality is rejected and data series is not normal as per JB statistics. The Jarque-Bera test statistic of all variables exceeds the critical values for any reasonable significance level to lead to the conclusion that data series of different variables do not follow a normal distribution. Table A3 reports the correlation coefficients of variables and VIF coefficients. There are two alternative measures of working capital which have not been correlated. The correlation coefficient between these variables is 0.24. The variance inflation factor (VIF) coefficient is the indicator of the existence of multicollinearity between explanatory variables. Among the explanatory variables, SIZE is highly correlated with leverage (correlation coefficient is 0.72), similarly, MTB is also showing a high correlation with ROCE i.e., 0.68. On the other hand nature of business (NOB) is showing a high negative correlation with leverage (−0.57) and size demonstrate a high inverse correlation with MTB.Size also depicts a high inverse correlation with tangibility (correlation coefficient is −0.51). One measure of working capital (GWC) is highly correlated with ROCE (correlation coefficient is 0.43. Whereas, another measure of working capital (NWC) is showing a high negative correlation with TANG. Other explanatory measures Viz., BOD Size, GDP, OCFA, and REEVOL are minimum correlated with other explanatory variables as well as with two measures of working capital. VIF tests reveal that the value corresponding to each explanatory variable is less than 10, which indicates that multicollinearity is not a serious problem here. Multicollinearity is a serious problem if the value of the variance inflation factor (VIF) is greater than 10 (Nachane, 2006).

Unit root test (ADF and PP)
The unit root test is carried out by using (Dickey and Fuller, 1979) or ADF, (Philips and Perron, 1988) or PP and (Kwiatkowski et al., 1992)

or KPSS techniques
The following hypotheses are set up for ADF and PP.
H 0 : There is a unit root for series.
H 1 : There is no unit root for the series. The series is stationary

Unit root test (KPSS)
The unit root test is done by KPSS by setting hypothesis as: H 0 : The Series is stationary.
H 1 : The Series is not stationary.
The unit root tests (ADF and PP) for data series for dependent and independent variables have been conducted in three different ways i.e., without intercept and linear trend (without C and T), with intercept or constant (with C) and with both trend and intercept/constant (with C and T). Table A4 reports the results of the unit root test of ADF, PP, and KPSS. According to ADF and PP tests, the data series of all explanatory variables and two dependent variables are found to be stationary at level. T statics of two endogenous variables and eleven exogenous variables are shown in Table A4 accompanied by p-value in parenthesis. Since p-value < 0.05, so the null hypothesis is rejected. It reports that data series are stationary for two dependent variables (NWC and GWC). Without constant and trend using ADF test, BOD, GDP, Growth, NOB and size are not stationary at a level as their associated p-value >0.05. Similarly, p values of t statistics of PP test of variables viz., BOD, GDP, and Size are greater than 0.05 at without constant and trend. But with constant as well as constant and a trend p-value of all the exogenous variables are less than 0.05. It means that data series of all these variables are stationary at level.
According to the KPSS test, data series are assumed to be stationary if the null hypothesis is accepted or LM statistics of KPSS test are less than the critical value. In this test, critical values at a 5 % significance level are shown in the parenthesis. Data series of dependent variables (NWC and GWC) are found to be stationary, as the associted critical values (0.46300 and 0.1460) are quite higher than the actual calculated value of LM statistics as depicted in Table A5. Similarly, critically values at a 5% significance level are also higher than the calculated value of LM statistics of all exogenous variables. It means that data series for all eleven exogenous variables are stationary at level (Table A4).

Cointegration test
Johansen's unrestricted cointegration rank test is performed for model 1 (NWC). According to trace value criteria as shown in Table A5 there are eight cointegration equations out of eleven equations. In all cointegration equations, the calculated value of trace statistics is higher than the critical value at a 5 % significance level and p-value < 0.05. The trace values of first seven equations are 742. 4614, 486.5950, 349.0122, 262.3885, 202.1680, 155.1471, 111.0497, 72.12307, 17.82775 and 8.542550 respectively. All these trace values are higher than the critical value at a 5% significance level. It denotes the rejection of the null hypothesis and acceptance of the alternative hypothesis. It means there is a cointegration in model 1 (NWC).
Johansen's unrestricted cointegration rank test is also performed using the criteria of maximum Eigenvalue for model-1 (NWC). According to Eigenvalue criteria as depicted in Table A5 there are four cointegration equations. As maximum Eigenvalue statistics is greater the critical value at 0.05 level and p-value < 0.05. The maximum Eigen statistics of the first three equations are 137.5828, 86.62373, and 60.22047 respectively which is greater than the critical value at a 5 % significance level. The equation 11 is also showing significant value as its p-value < 0.05. These equations denote the rejection of the null hypothesis and acceptance of the alternative hypothesis. It means there is cointegration in model 1 (NWC) as per maximum Eigenvalue criteria (Table A5).
Johansen cointegration rank test is also performed for model 2 (GWC) using criteria of trace value and maximum Eigenvalue. According to trace value criteria as depicted in Table A6, all eleven cointegration equations have calculated value greater than the critical value and p-value < 0.05. These cointegration equations have trace value 680. 5220, 423.7147, 305.1810, 225.5563, 167.0569, 121.3377, 83.10222, 52.98266, 32.23084 and 8.132066 respectively. These equations reflect the rejection of the null hypothesis at the 0.05 level and acceptance of the alternate hypothesis. It means there is a cointegration in model 2 (Table A6).
According to maximum Eigenvalue criteria (Table A6) for the model −2 (GWC), the four cointegration equations are found significant at a 5% level. As maximum Eigenvalue statistics is greater the critical value and p-value < 0.05. The maximum Eigen statistics of the first four equations are 256.8072, 118.5338, 79.62470 and 58.49943 respectively. These equations reflect the rejection of the null hypothesis and the acceptance of the alternative hypothesis. It means there is also cointegration in model 2 (Table A6). Table A7 reports the significant determinants of networking capital (NWC) obtained through three different methods i.e., ordinary least square, fully modified OLS (FMOLS) and dynamic OLS (DOLS). According to OLS, there are five determinants of networking capital Viz., board size (BOD), leverage (LEV), nature of business (NOB), market to book value and assets tangibility (TANG). All these variables have t statistics greater than 1.96 and p-value < 0.05. All these variables are predictors of future requirements of networking capital (NWC) and showing a negative association with net working capital. It means due to an increase or decrease in these variables there will opposite change in net working capital. T statistics of leverage (8.078) are highest followed by tangibility (−6.677) and nature of business (−3.717). The constant or intercept is also found significant in OLS as t statistics are 7.945 and p-value < 0.05.

Results of OLS, FMOLS, and DOLS
According to FMOLS, there are only three significant predictors of the requirement of networking capital which are also found significant in OLS viz., leverage (LEV), nature of business (NOB) and tangibility (TANG). The t statistics of these variables are −4.1292, −2.1108 and −3.8899 respectively. Just like the previous method, these variables show a negative relationship with net working capital. This method validates the results of OLS.
The dynamic ordinary least square (DOLS) method reflects only two significant variables in Table  A6, which are also observed significantly in the previous two methods (OLS and FMOLS). These variables are leverage (LEV) and tangibility (TANG). The t statistics of these variables are −3.3739 and −3.2429 respectively. Since p values < 0.05, so both these variables are significant as per DOLS.
R square reflects the percentage of variance as properly explained by all significant variables of net working capital. In the case of OLS, all five significant variables explain 30.8 % percent of the variance of net working capital. In fully modified OLS (FMOLS) and dynamic OLS (DOLS), R square values are 0.2930 and 0.3816 respectively. It means that FMOLS and DOLS explain 29.30% and 38.16 % of the variance of the dependent variable (net working capital). The minimum difference between R square and adjusted R square validates the results of all the methods used to develop models to forecast future requirements of net working capital. Durbin Watson ratio reflects that there is no problem of autocorrelation in the case of the OLS method as this value is greater than 1.50. Anova value (24.916) is also found significant in OLS technique as p-value < 0.05 (Table A6). Table A8 reports the significant exogenous variables of gross working capital (GWC) obtained from three different methods Viz., OLS, and FMOLS and DOLS. OLS method has found five major determinants of gross working capital of BSE listed Indian companies viz., board size (BOD), return on common equity (ROCE), nature of business (NOB), size and tangibility (TANG). All these variables have t statistics > 1.96 and p-value < 0.05. All these variables are significant as per the OLS method. Return on common equity (ROCE) and board size (BOD) have a positive coefficient value which reflects the positive association with gross working capital. The remaining three significant determinants (NOB, SIZE, and TANG) demonstrate a negative relationship with gross working capital. Constant is also found significant in the ordinary least square (OLS) method with t statistics 4.916 and p value< 0.05. Four significant variables are obtained in the dynamic OLS (DOLS) method. All these variables are also significant in the previous two methods. These variables are the nature of business (NOB), operating cash flows (OCFA), return on common equity (ROCE) and tangibility (TANG). Only ROCE is showing a positive impact on gross working capital and the remaining three significant variables are negatively related to GWC. R square represents the total variance explained of the dependent variable (GWC) by all significant explanatory variables together. In OLS and FMOLS, the total variance explained of gross working capital by all exogenous variables is 33.9% and 33.81% respectively. In the dynamic OLS (DOLS) method, the total variance explained (42.8%) is quite high as compare to the previous two methods. In all the models, the difference between R square and adjusted R square is less than 0.05 which further authenticates the validity of these models to predict the future requirement of gross working capital (GWC). Durbin Watson's ratio reflects that there is no problem with autocorrelation in the case of the OLS method as this value is 1.3465. Anova value (23.968) is also significant in the case of the OLS technique as a p-value < 0.05 (Table A8). Table A9 reports the results of the GMM estimation for two alternative models specified of Eq. 1 and 2 using the alternative measures of working capital. The lagged working capital ratio found significant at one percent level for both models along with other explanatory variables. So both models meet the first condition as proposed by (Arellano & Bond, 1991).

Results of generalized method of moments (GMM) estimation
In model one (NWC), other explanatory variables which are also found significant at 1 percent level are GDP, the price-earnings ratio (P/E) and tangibility (Tang). Age and leverage are also found significant at the 10 percent level. In model two (GWC) along with lagged gross working capital ratio, age, and board size, operating cash flow (OCFA), market to book value ratio (P/B), return on common equity (ROCE), leverage and size are also found significant at 1 percent level. The price-earnings ratio is significant at 5% level. In model one, lagged net working capital ratio, price-earnings ratio and leverage are showing a positive association with net working capital. Other significant variables have a negative association with the dependent variable (NWC). In model two related to gross working capital, five variables are showing positive relationship endogenous variable (GWC ratio). The remaining variables are showing a negative relationship with the dependent variable. In both the models on age are showing a similar relationship, the rest of the significant variable has an opposite association in both the models. It is because of that in both models dependent variables are not the same. The first model is related to networking capital ratio and second deals with the gross working capital ratio.
In both the models, values of the Surgan test or J statistics are 20.20069 and 19.81940 respectively. Null hypothesis means that instruments are valid and they are uncorrelated with residual So, we accept the absence of correlation between instruments and residuals and instruments are valid. The validity of instruments is proven by the Sargan test with respective p-values 0.2640 and 0.2042, which are higher than the significance level of 5%. Results of this indicate that instruments are valid and instruments and error terms are uncorrelated. Serial correlation as calculated for the first order is found significant at 10 % (P < 0.10) level for NWC, but not for model two (GWC). For both the models, the serial correlation calculated by Arellano and Bond (1991) method is showing a negative value. But second-order correlation accepting the null hypothesis and showing positive values in both the models.
The tests of serial correlation of Arellano and Bond (1991) should show that the first order statistic is statistically significant, whereas the second-order statistic is not, which is what we would expect if the model error terms are serial uncorrelated in levels. In our study, the first-order serial correlation is statically significant for the first model at a 5% level and second-order serial correlation is not statistically significant. But in model two, first and second-order serial correlation is not statistically significant. Table A10 reports the acceptance and rejection of a hypothesis using four alternative methods of estimation. Leverage (H 1 ) is a strong predictor of networking capital as revealed from the four alternative methods of estimation. Hypothesis (H 4 ) is also accepted all the alternative methods that the Nature of business has a negative influence on the future requirements of working capital in BSE listed companies. According to the hypothesis (H 8 ), profitability can also be strongly recommended as a predictor of the future requirement of gross working capital, as it is found to be significant in all the methods. Hypothesis (H 10 ) is accepted that Board characteristics or Board Size have a significant influence on the future requirements of working capital in BSE listed companies. Hypothesis (H 6 ) strongly recommended as an explanatory variable to predict the future requirement of the working capital of Indian companies, as it is found significant in four alternative techniques. Hypothesis (H 3 ) is partially accepted that the level of asymmetric information has a negative impact on the future working capital requirement of the firm. It is found to be significant, according to OLS and FMOLS. Similarly, hypothesis (H 5 ) is also partially accepted that size or Capital Market Access has a positive impact on the future working capital requirement. According to GMM technique, hypothesis (H12) is accepted the lagged value of the dependent variable (NWC and GWC) positively affect the future requirement of Working Capital.

Conclusion
The study is carried out to explore the determinants of the gross and networking of BSE listed companies. Four alternative methods (OLS, FMOLS DOLS, and GMM) are chosen to develop models and to identify the significant variables affecting the working capital of selected companies. It is intended to use four different methods so that results obtained from one method can be verified from other methods. The first method OLS identified five significant variables to forecast the future requirement of net working capital. These variables are board size (BOD), leverage (LEV), market price to book value (MTB), nature of business (NOB) and asset tangibility. All these variables are showing a negative relationship with net working capital.
It means that due to change (increase/decrease) in these variables, the net working capital (NWC) will move in the opposite direction. The second method FMOLS method has obtained three significant variables (nature of business, leverage and assets tangibility). These three predictor variables are common with the previous method and also showing a negative relationship with net working capital. So FMOLS, validate the results obtained from OLS. The third method DOLS has explored only two significant variables (leverage and assets tangibility) to forecast future requirements of net working capital (NWC). These variables are also found significant in the earlier two methods and showing a negative association with net working capital (NWC). All these methods have also been used to develop a model to forecast the gross working capital requirements of selected companies. The first method of ordinary least square (OLS) has identified five predictors variables of gross working capital Viz., return on common equity (ROCE), board size (BOD), size of the firm (SIZE, assets tangibility (TANG) and nature of business (NOB). The two variables (ROCE and BOD) are showing a positive association with working capital and the remaining three variables (NOB, Size, and TANG) have shown a negative relationship with gross working capital.
FMOLS has explored six significant determinants of gross working capital. Three variables (board size, return on common equity and market to book value) out of them are showing a positive impact on working capital. Three other variables (nature of the business, assets tangibility and operating cash flows are depicting a negative relationship with gross working capital. Out of these six variables, the four variables (nature of the business, return on common equity, the board size, and assets tangibility) are also found significant in the previous method (OLS). Dynamic OLS has explored four significant predictors' variables of gross working capital viz., nature of the business, operating cash flows, return on common equity and asset tangibility. These variables are also significant as per fully modified OLS.
The results generalized methods of moments (GMM) are also similar to the previous three methods. Determinants, as identified for networking capital by earlier three methods (OLS, FMOLS, and DOLS), are supported by GMM except for few exceptions. Similarly, determinants, as identified for gross capital requirement, are also observed the same in GMM. But there is an addition of new determinants of GWC by using GMM. Now we can say that the findings of four alternative methods of panel data estimation are consistent and justify the relevance of the findings from the earlier studies with some new developments.
The study will have numerous benefits to policymakers specifically board of directors/mangers of BSE listed companies. Board directors always want to maintain an optimum level of working capital and want to avoid the inadequate and excessive working capital situation. As an inadequate flow of working capital put an obstacle in the day to day business operation and excessive working capital creates idle funds. The different significant variables as discovered in this study shall be useful to concerned authorities of the company for arriving at optimal working capital of the company and providing maximum benefits to existing shareholders of the company. Consequently, concerned authority will be able to utilize significant variables of the company in such a way, so that they may be able to forecast future requirements of working capital of their companies. The study has made recognized that any increase in profitability board size in Indian companies will lead directly to increment in working capital in a firm while increase in leverage, Size assets tangibility and operating cash flows will lead to a decrease in working capital if a firm. The managers or policymakers will be benefitted by into account all the variables showing positive or negative association while arriving optimum the optimum level of working capital so as to avoid waste in firm operations. The study contributes significantly to existing body of knowledge. This study has been carried by taking into account the historical economic reforms so it provides complete new insight for forecasting the requirement of working to existing Indian firms. Although this study is carried out by taking into account only 150 BSE listed firms but interpretation from this can be made for whole set of Indian firms as well as outside the India. This study will provide new base to academician to explore further the determinants by taking into consideration the similar situation of particular economy.

S.No
Variable Measure Citation
* The coefficient is significant at 1 per cent level. ** The coefficient is significant at 5 per cent level. *** The coefficient is significant at 10 per cent level.  Source: Authors Calculations Note: P value is presented in parentheses.
* The coefficient is significant at 1 per cent level. ** The coefficient is significant at 5 per cent level. *** The coefficient is significant at 10 per cent level. Source: Author's Calculation with Eviews Note: P value is presented in parentheses.
* The coefficient is significant at 1 per cent level.
** The coefficient is significant at 5 per cent level.
*** The coefficient is significant at 10 per cent level.