Decoding the Paradox of Dividend Policy: An Empirical Study on Indian Banking Sector

Abstract

The dividend policy has often been treated as the most complicated and intriguing aspects in corporate finance. Profitability was always cited as the main source of confidence for dividend payments. Numerous articles written on the dividend policy explored several of its other determinants. The most popular Lintner’s Dividend Model has been assessed and applied by researchers in different sectors including the banking sector in India. The results from the banking sector also confirmed to a greater extent the accuracy of Lintner’s Dividend Model. Although Lintner’s Dividend Model had its firm footing in the Indian banking industry, the model has not much explored about liquidity constraints, ownership and managerial efficiency. The above-mentioned predictors are important in the present scenario where many public sector banks paid dividends while having high nonperforming assets. Recently the government has announced a dividend cut for 16 public sector banks due to high level of stressed assets. Hence, profitability and stressed assets are the paradoxical aspects in the dividend policy for the banking industry. Findings from this study have evidence of substantial influence of liquidity constraints, ownership and managerial efficiency and their influence on the dividend policy.

Keywords

Dividend policy nonperforming assets (NPA)managerial efficiency ownership Lintner model

Introduction

Dividend policy has always been treated with ambiguity by researchers and corporate. The confidence for dividend payment originates from profit but indecisiveness emerges due to conflict between shareholder wealth maximization and earnings retention. A complete payout creates appreciable increase in shareholder wealth creation whereas a pure retention can help in growth opportunities of enterprise.

Different articles have been written and models have been designed to solve the ambiguity. The Lintner’s model has become very popular and is applied by many researchers to understand the dividend policy. In India, many studies have been done on dividend determinants and Lintner’s model in various sectors. The banking sector being an important constituent for India’s economic growth has witnessed few studies on dividend policy. The sector has two very crucial parameters which can influence the confidence of dividend payment in this sector. The increase in nonperforming assets (NPAs) and government ownership can decide the fate of dividend payments.

The very evidence of the earlier statement can be elaborated by a snapshot of decisions taken over few years by two successive governments in India. During UPA regime, the finance ministry has directed the public sector banks (PSBs) to comply with dividend payments, the minimum rate being 20 per cent (maximum of capital or net profit) for financial year 2012–2013. This came as a shock to public sector enterprises due to two reasons:

Higher NPAs and weak credit demand had left the PSBs shattered.

Increase in provisioning for bad loans has made banks to face losses in previous fiscal years.

The alternative was capital infusion and borrowings from government for dividend outflow. This had a negative impact on capital adequacy ratio of banks and violated the RBI Guidelines of dividend allocation from previous year surplus.

In year 2014, the regime of NDA government at center witnessed announcements for major reforms in the economy and encouraged infrastructure projects for the India’s growth story. The irony was that, until December 2016, the earnings were far from satisfactory of public banks. The capital loading was aiding for dividend payments but higher provisioning and weak credit growth has already eroded the system.

Taking a cue from private banks (PBs), the public banks also had lesser exposure to infrastructure and power projects while increasing exposure to retail loans. The direct consequence was large projects were left with fund shortage. The gross NPAs of public banks increased from 5.43 per cent in 2014–2015 to 9.32 per cent in 2015–2016. Mounting the losses of public banks necessitated the need to signal for reducing or omitting dividend payments.

The present research focuses on investigating the dividend confidence model of listed banks in India while tracing the positioning of NPA provisions in the model. The article attempts to create a dividend model with NPA provisioning embedded in it beneficial for policy implications.

Literature Review

Lintner’s (1956) model was applied by many researchers to understand the pattern of dividend payments. The major focus of Lintner was on earnings per share and previous year’s dividend payment which gives confidence for present dividend payment.

Subsequent studies on dividend policy by researchers explored many other variables that can influence a dividend policy. Profitability is always considered the motive behind dividend payments whereas a research by Guerard, Bean, and Andrews (1987) designed a model which proved interaction of investment and debt for dividend decisions of the manufacturing sector.

Dividend receipt increases shareholder wealth but the funds are sacrificed from the growth perspective of the firm. The ownership dilemma in this context has been explored by Desai, Foley, and Hines (2007) in their research paper on dividend policy inside the multinational firm. They found that the parent company needs funds in the form of dividend from subsidiaries for US multinational firms. Shareholder inclination to dividend receipts and deciding managerial performance accordingly makes dividend policy a paradox. Brealey and Myers (2002)identified dividend policy among top 10 puzzles in finance. Anand (2002) applied factorial analysis on response of 81 Chief Financial Officers for analyzing dividend decisions. His findings reveal that investor urge and client reaction decide dividend policy design.

Kapoor, Misra, and Kanwal (2010) found the factor for dividend policy determinants in the Indian service sector. From a total of 29 companies observed for study, nine were banks. Since the banking sector is regulated by the Banking Regulation Act, the authors used dummy variables for banking companies. Their findings indicate that higher profit fluctuations lead to higher dividend payout. The study affirms that agency dispute is of minor importance.

Since the research focuses on banking sector, a few journals related to dividend policies in banking industry is reviewed. Tahir, Ullah, and Mahmood (2015) concluded from the empirical analysis on Pakistan’s banks that investment and financing decisions had a close association but real investment and dividend decisions do not alter the financing decision.

In Indian banking industry, the application of Lintner’s model was done by researchers Pal and Goyal (2007). They confirmed the applicability of model and remarked that low adjustment speed implies less enthusiasm from management to achieve target payout ratio. Similar application of Lintner’s model on the Indian banking industry was done by Bodla, Pal, and Sura (2007). The researchers have identified the applicability of model in public and private sector banks distinctively. The study also explored that dividend decisions of PSBs have more stability than that of the private sector banks.

Research Gap

The existing literature review has been an evidence of not much studies done on dividend decision model of Indian banking industry. Moreover, the studies on banking sector have less focus on the liquidity pattern which generates confidence for dividend payments. Linear models are mostly used to describe the dividend determinants but decision criteria for dividend payouts has not been explored yet.

The parameters for liquidity and profitability have to be explored in a critical way so as to reflect the NPA situation which has become a larger issue for banks. The theoretical framework needs to be configured using a different set of criteria for the Indian banking industry.

Theoretical Framework

The theoretical framework needs to be constructed accordingly taking into account the liquidity, efficiency, and ownership issues.

The predictors and factors used in this framework are discussed as follows:

Cash earnings per share (CEPS)

This ratio takes into account cash profit instead of book profit in earnings per share.

CEPS = \frac{C a s h profit}{Number of share}

The ratio can highlight the performance of company in terms of liquidity.

Net interest margin (NIM)

This ratio calculates how much investment return is in excess of interest expense over average earning assets.

NIM = \frac{Return on investment - Expense on interest}{Average earning assets}

The performance efficiency can be determined in banks by this ratio. This ratio suggests how effectively the funds are utilized by a bank.

Interest income on total assets (IITA)

This ratio finds out the amount of operating income over total assets.

IITA = \frac{Interest income}{Total assets}

The efficiency in managing operations can be determined from this ratio.

Return on assets (ROA)

Here net income is expressed as a percentage of assets. This ratio indicates managerial efficiency in asset utilization.

ROA = \frac{Net profit}{Average total assets}

Return on equity (ROE)

The net income is expressed as a percentage of assets. This ratio indicates profit on shareholder equity.

ROE = \frac{Net income}{shareholders' equity}

Financing activities (FIN)

These activities can be in the form of capital infusion and debt financing. The increase in financing activities can influence the confidence of a firm to pay dividends.

Provisions during the year (PROV)

These include mainly provision for NPAs and other provisions such as for diminution in the value of investments, tax payment, etc. This parameter is considered to see the effect of accounting provisions on dividend decisions.

Profit after tax (PAT)

It is the net earnings after deducting all expenses, provisions and taxes from revenue. The net earnings indicate the surplus or deficit for the enterprise.

Ownership issue (TYPE)

The Ownership issue for the model is considered by segregating all the listed banks into PSBs and PBs. The factorial notation used is Type 1 for PSBs and Type 2 for PBs.

Dividend per share (DPS)

It is the total amount of dividend receipt for each share of stockholders.

Dividends per share = \frac{Total dividends paid}{Total shares outstanding}

This is the outcome variable in the model which is expressed in percentage and indicates the level of confidence by firm for dividend payment.

The dividend confidence model is constructed with the dynamics of ownership, liquidity, and efficiency, and the pattern had to be investigated accordingly.

The linear mixed model creates this flexibility through factorial interactions wherein the deviations in data and time trend can be studied for large number of firm-specific observations.

The conceptual framework for this model is depicted further:

Figure 1 describes the DPS as repeated observations in different periods of time (year wise). The observations are firm specific wherein the factors are PSBs and PBs. The fixed effects interaction model is used because the ownership type has already been specified and the interactions are restricted to only these two types.

This article attempts to have evidence of decisions influencing the confidence for dividend payments. For decisions on dividend payments, the multidimensional logit model is used. The advantage of using this model is that it specifies the predictors responsible for dividend decisions.

Data and Methodology

To explore the suitability of the model, the sample extracted consists of all 47 listed banks in India. The final sample consists of a total of 38 listed banks that have been almost regular in paying dividends. The data are extracted from the BSE website database. For model estimation, we have used linear mixed models and multinomial logit model which has been already discussed in the theoretical framework.

Figure 1.

Source:Authors’ own computation.

In this research, the multinomial logit model for dividend decision has covered two distinct time frames.

1. The period from 2007 to 2013 to represent earlier governance dynamics. The confidence for dividend payment represented by DPS is divided into four categories.

(a) DPS = 1 ranging from 100 per cent and above

(b) DPS = 2 ranging from 50 to 100 per cent

(d) DPS = 4 ranging from zero to 10 per cent

The two constructs for the four categories are represented as follows:

1 (D P S = a) / (D P S = b) = A o + B 1 C E P S + B 2 N I M + B 3 I I T A

(1)

where a represents the first three categories of DPS and b represents the reference fourth category.

(D P S = a) / P (D P S = b) = Y o + X 1 P A T + X 2 F I N + X 3 P R O V

(2)

2. The period from 2014 to 2016 is to represent the present governance dynamics. Moreover, this period is very crucial as PSBs have been forced to reduce or omit dividends due to higher NPAs. The confidence for dividend payment represented by DPS is divided into three categories.

(a) DPS = 1 is zero.

(b) DPS = 2 ranging from 100 per cent

The third construct for the three categories is represented as follows:

(D P S = k) / P (D P S = \int) = C o + D 1 P A T + D 2 F I N + D 3 P R O V

(3)

Where a represents the first two categories of DPS and b represents the reference third category. The models have been estimated taking private and public ownership into embedded factors. SPSS 20 was used to extract the results and analyze the output.

Results and Discussion

The results of linear mixed model for the conceptual framework are presented in Tables 1–4.

The model dimension in Table 1 speaks about the number of banks (=38), fixed effects interactions and factors with covariates. The factor is type here represented in two levels for PSBs and PBs. The repeated frequency talks about the years in which the dividend payment is made. The dependent variable being DPS predicted with independent variables CEPS, IITA, ROA, and ROE.

Table 1.

Model Dimension

		Number of Levels	Covariance Structure	Number of Parameters	Subject Variables	Number of Subjects
Fixed	Intercept	1	–	1	–	–
Effects	TYPE	2	–	1	–	–
	CEPS	1	–	1	–	–
	IITA	1	–	1	–	–
	ROA	1	–	1	–	–
	ROE	1	–	1	–	–
Repeated	YEAR	10	Diagonal	10	ID	38
Effects	–	–	–	–	–	–
Total	–	17	–	16	–	–

Source: Authors’ own computation using SPSS.

Note: Dependent variable: DPS.

Table 2 represents the F-tests for the fixed effects factors and covariates embedded in the model. The factor TYPE and covariates CEPS, IITA are significant at P < 0.05 level specifying the influence of these variables on DPS. The managerial efficiency ratios (ROA, ROE) accept the null hypotheses which implies the effect does not contribute to dividend confidence.

Table 2.

Type III Tests of Fixed Effects

Source	Numerator df	Denominator df	F	Significance
Intercept	1	274.935	13.905	0.000
TYPE	1	280.470	7.022	0.009
CEPS	1	70.216	4,857.986	0.000
IITA	1	265.959	8.702	0.003
ROA	1	221.691	0.013	0.910
ROE	1	278.259	0.111	0.740

Source: Authors’ own computation using SPSS.

Note: Dependent variable: DPS.

Table 3 represents the fixed effect estimate and significance tests for the covariates in interaction with five different factors. The table suggests that the PSB effect is significantly negative in the model as compared to PB for respective covariates at P < 0.05 level.

Table 3.

Estimates of Fixed Effects

Parameter	Estimate	Standard Error	df	t	Significance	95% Confidence Interval
Parameter	Estimate	Standard Error	df	t	Significance	Lower Bound	Upper Bound
Intercept	6.886710	1.759226	278.940	3.915	0.000	3.423666	10.349755
(TYPE = 1)	–1.207050	0.455505	280.470	–2.650	0.009	–	–0.310407
		–	–	–	–	2.103693	–
(TYPE = 2)	0b	0	–	–	–	–	–
CEPS	0.129019	0.001851	70.216	69.699	0.000	0.125328	0.132711
IITA	–0.586097	0.198681	265.959	–2.950	0.003	–0.977284	–0.194911
ROA	0.066055	0.582232	221.691	0.113	0.910	–	1.213473
	–	–	–	–	–	1.081363	–
ROE	–0.013686	0.041133	278.259	–0.333	0.740	–0.094659	0.067286

Source: Authors’ own computation using SPSS.

Note: “b” is set to zero because it is redundant. Dependent variable: DPS.

Table 4 specifies the variability for repeated observations over different years. The variability is more pronounced in the year 2008.

Table 4.

Estimates of Covariance Parameters

Parameter		Estimate	Standard Error	Wald Z	Significance	95% Confidence Interval
Parameter		Estimate	Standard Error	Wald Z	Significance	Lower Bound	Upper Bound
RepeatedMeasures	Var: (YEAR = 2007)	7.899709	1.921702	4.111	0.000	4.903947	12.725546
	Var:(YEAR = 2008)	41.283841	10.601578	3.894	0.000	24.957085	68.291450
	Var:(YEAR = 2009)	6.928565	1.637962	4.230	0.000	4.359263	11.012183
	Var:(YEAR = 2010)	8.639926	2.023440	4.270	0.000	5.459610	13.672831
	Var:(YEAR = 2011)	12.040569	2.834684	4.248	0.000	7.590153	19.100445
	Var:(YEAR = 2012)	9.751544	2.304120	4.232	0.000	6.136904	15.495212
	Var:(YEAR = 2013)	20.352135	4.806752	4.234	0.000	12.810729	32.333010
	Var:(YEAR = 2014)	15.435986	3.608759	4.277	0.000	9.761871	24.408194
	Var:(YEAR = 2015)	4.297729	1.048224	4.100	0.000	2.664584	6.931842
	Var:(YEAR = 2016)	6.626907	1.691470	3.918	0.000	4.018350	10.928837

Source: Authors’ own computation using SPSS.

Note: Dependent variable: DPS.

The results of multinomial logit model for the first construct from 2007 to 2013 is depicted further:

Table 5 gives evidence of better model prediction at P < 0.05 level.

Table 5.

Model Fitting Information

Model	Model Fitting Criteria	Likelihood Ratio Tests
Model	–2 Log Likelihood	Chi-square	df	Significance
Intercept only	617.052	–	–	–
Final	324.000	293.052	12	0.000

Source: Authors’ own computation using SPSS.

Table 6 explains the variance from 6.87 to 7.52 per cent in DPS for the said model.

Table 6.

Pseudo R-square

Model	Variance
Cox and Snell	0.687
Nagelkerke	0.752
McFadden	0.473

Source: Authors’ own computation using SPSS.

Table 7 explains the important criteria of model fitting through different predictors. CEPS, NIM, and TYPE contribute to the model for DPS decision whereas IITA has no contribution for the model at P < 0.05.

Table 7.

Likelihood Ratio Tests

Effect	Model FittingCriteria	Likelihood Ratio Tests
Effect	–2 Log Likelihoodof ReducedModel	Chi-square	df	Significance
Intercept	324.000^a	0.000	0
CEPS	577.864	253.864	3	0.000
NIM	336.539	12.539	3	0.006
IITA	325.848	1.848	3	0.605
TYPE	344.173	20.173	3	0.000

Source: Authors’ own computation using SPSS.

Note: “a” indicates that the reduced model is equivalent to the final model because omitting the effect does not increase the degrees of freedom.

Table 8 expresses the influence of different predictors for different categories of dividend decisions. From the table, it can be seen that DPS for different decisions is influenced by CEPS and NIM.

Table 8.

Parameter Estimates

DPS^a	B	Standard Error	Wald	df	Significance	Exp (B)	95% ConfidenceInterval forExp (B)
DPS^a	B	Standard Error	Wald	df	Significance	Exp (B)	Lower Bound	Upper Bound
1 Intercept	–5.070	4.603	1.213	1	0.271	–	–	–
CEPS	0.225	0.035	40.589	1	0.000	1.252	1.168	1.341
NIM	–1.226	0.604	4.119	1	0.042	0.293	0.090	0.959
IITA	0.517	0.581	0.794	1	0.373	1.677	0.538	5.235
(TYPE = 1)	–1.274	1.095	1.353	1	0.245	0.280	0.033	2.394
(TYPE = 2)	0^b			0
2 Intercept	–3.274	3.815	0.736	1	0.391	–	–	–
CEPS	0.140	0.032	18.890	1	0.000	1.150	1.080	1.224
NIM	–1.118	0.500	5.004	1	0.025	0.327	0.123	0.871
IITA	0.601	0.486	1.533	1	0.216	1.825	0.704	4.727
(TYPE = 1)	1.341	0.876	2.344	1	0.126	3.823	0.687	21.275
(TYPE = 2)	0^b	–	–	0	–	–	–	–
3 Intercept	2.344	3.499	0.449	1	0.503	–	–	–
CEPS	0.076	0.030	6.353	1	0.012	1.079	1.017	1.145
NIM	–1.542	0.490	9.887	1	0.002	0.214	0.082	0.559
IITA	0.302	0.447	0.458	1	0.499	1.353	0.564	3.246
(TYPE = 1)	1.498	0.808	3.443	1	0.064	4.474	0.919	21.780
(TYPE = 2)	0^b	–	–	0	–	–	–	–

Source: Authors’ own computation using SPSS.

Note: “a” indicates that the reference category is 4 and “b” indicates that this parameter is set to zero because it is redundant.

Table 9 represents the classification table which states that 77 per cent of the model is predicted correctly.

The results of multinomial logit model for the second construct from 2007 to 2013 is depicted further:

Table 9.

Classification

Observed	Predicted
Observed	1	2	3	4	Percentage Correct
1	74	5	3	0	90.2
2	9	23	21	1	42.6
3	2	5	93	1	92.1
4	0	0	11	4	26.7
Overall	33.7	13.1	50.8	2.4	77.0
Percentage

Source: Authors’ own computation using SPSS.

Table 10 gives evidence of better model prediction at P < 0.05.

Table 10.

Model Fitting Information

Model	Model Fitting Criteria		Likelihood Ratio Tests
Model	−2 Log Likelihood	Chi-square	df	Significance
Intercept	619.824
only
Final	476.115	143.709	12	0.000

Source: Authors’ own computation using SPSS.

Table 11 explains the variance from 4.35 to 4.75 per cent in DPS for the said model.

Table 11.

Pseudo R-square

Model	Variance
Cox and Snell Model	Variance 0.435
Nagelkerke	0.475
McFadden	0.232

Source: Authors’ own computation using SPSS.

Table 12 explains the important criteria of model fitting through different predictors.

Table 12.

Likelihood Ratio Tests

Effect	Model Fitting Criteria	Likelihood Ratio Tests
Effect	−2 Log Likelihood of Reduced Model	Chi-square	df	Significance
Intercept	476.115^a	0.000	0
PAT	526.015	49.900	3	0.000
FIN	487.871	11.756	3	0.008
PROV	483.948	7.833	3	0.050
TYPE	489.702	13.587	3	0.004

Source: Authors’ own computation using SPSS.

PAT, FIN, PROV, and TYPE contribute to the model for DPS decision at P < 0.05.

Table 13 expresses the influence of different predictors for different categories of dividend decisions. From the table, it can be seen that DPS for different decisions are positively influenced by PAT, FIN, and PROV. TYPE contributes to model for DPS decisions of 100 per cent or above.

Table 13.

Parameter Estimates

DPS^a		B	Standard Error	Wald	df	Significance	Exp (B)	95% Confidence Interval for Exp (B)
		B	Standard Error	Wald	df	Significance	Exp (B)	Lower Bound	Upper Bound
1	1 Intercept	−2.456	0.663	13.744	1	0.000	–	–	–
	PAT	0.008	0.003	7.945	1	0.005	1.008	1.003	1.014
	FIN	−0.001	0.000	8.131	1	0.004	0.999	0.998	1.000
	PROV	0.007	0.003	3.862	1	0.049	1.007	1.000	1.014
	(TYPE = 1)	−1.976	0.919	4.618	1	0.032	0.139	0.023	0.840
	(TYPE = 2)	0^b		–	0		–	–	–
2	Intercept	−2.306	0.663	12.114	1	0.001		–	–
	PAT	0.008	0.003	6.805	1	0.009	1.008	1.002	1.014
	FIN	−0.001	0.000	8.323	1	0.004	0.999	0.998	1.000
	PROV	0.007	0.003	4.081	1	0.043	1.007	1.000	1.014
	(TYPE = 1)	−1.560	0.916	2.901	1	0.089	0.210	0.035	1.265
	(TYPE = 2)	0^b		–	0		–	–	–
3	Intercept	−1.165	0.622	3.510	1	0.061		–	–
	PAT	0.006	0.003	3.952	1	0.047	1.006	1.000	1.012
	FIN	−0.001	0.000	5.025	1	0.025	0.999	0.998	1.000
	PROV	0.007	0.003	4.430	1	0.035	1.007	1.001	1.014
	(TYPE = 1)	−0.542	0.865	0.393	1	0.531	0.581	0.107	3.170
	(TYPE = 2)	0^b		–	0	–	–	–	–
		−	−	−	−	−	−	−	−

Source: Authors’ own computation using SPSS.

Table 14 represents the classification table which states that 55.60 per cent of the model is predicted correctly.

Table 14.

Classification

Observed	Predicted
Observed	1	2	3	4	Percentage Correct
1	47	0	35	0	57.3
2	21	3	29	1	5.6
3	10	2	85	4	84.2
4	1	0	9	5	33.3
Overall Percentage	31.3	2.0	62.7	4.0	55.6

Source: Authors’ own computation using SPSS.

The results of multinomial logit model for the third construct from 2014 to 2016 are depicted as follows:

Table 15 gives evidence of better model prediction at P < 0.05.

Table 15.

Model Fitting Information

Model	Model FittingCriteria	Likelihood Ratio Tests
Model	−2 LogLikelihood	Chi-square	df	Significance
Intercept	225.430	–	–	–
only	–	–	–	–
Final	94.498	130.932	8	0.000

Source: Authors’ own computation using SPSS.

Table 16 explains the variance from 6.83 to 7.93 per cent in DPS for the said model.

Table 16.

Pseudo R-square

Model	Variance
Cox and Snell	0.683
Nagelkerke	0.793
McFadden	0.581

Source: Authors’ own computation using SPSS.

Table 17 explains the important criteria of model fitting through different predictors.

Table 17.

Likelihood Ratio Tests

Effect	Model Fitting Criteria	Likelihood Ratio Tests
Effect	−2 LogLikelihood ofReduced Model	Chi-square	df	Significance
Intercept	94.498^a	0.000	0	–
PAT	157.952	63.453	2	0.000
FIN	96.634	2.136	2	0.344
PROV	94.827	0.329	2	0.848
TYPE	102.038	7.539	2	0.023

Source: Authors’ own computation using SPSS.

Note: “a” indicates that this reduced model is equivalent to final model because omitting the effect does not increase the degrees of freedom.

The results show evidence of only PAT and TYPE for model contribution at P < 0.05.

Table 18 expresses the influence of PAT and TYPE over different dividend decisions.

Table 18.

Parameter Estimates

DPS^a		B	Standard Error	Wald	df	Significance	Exp (B)	95% Confidence Interval for Exp (B)
		B	Standard Error	Wald	df	Significance	Exp (B)	Lower Bound	Upper Bound
1	Intercept	−19.008	1.863	104.063	1	0.000
	PAT	−0.009	0.005	3.712	1	0.054	0.991	0.982	1.000
	FIN	0.000	0.000	0.701	1	0.402	1.000	0.999	1.000
	PROV	0.000	0.001	0.205	1	0.651	1.000	0.999	1.002
	(TYPE = 1)	19.519	0.000		1		299,888,821.68	299,888,821.68	299,888,821.68
	(TYPE = 2)	0^b			0		–	–	–
2	Intercept	−1.399	0.472	8.794	1	0.003	–	–	–
	PAT	0.001	0.000	4.793	1	0.029	1.001	1.000	1.002
	FIN	0.000	0.000	1.702	1	0.192	1.000	1.000	1.000
	PROV	0.000	0.000	0.137	1	0.712	1.000	1.000	1.001
	(TYPE = 1)	−1.727	0.850	4.124	1	0.042	0.178	0.034	0.942
	(TYPE = 2)	0^b	–	–	0	–	–	–	–
	–	–	–	–	–	–	–	–	–

Source: Authors’ own computation using SPSS.

Note: “a” indicates that the reference category is 3 and “b” indicates that this parameter is set to zero because it is redundant.

Table 19 represents the classification table which states that 78.10 per cent of model is predicted correctly.

Table 19.

Classification

Observed	Predicted
Observed	1	2	3	Percentage Correct
1	16	0	3	84.2
2	0	17	16	51.5
3	0	6	56	90.3
Overall Percentage	14.0	20.2	65.8	78.1

Source: Authors’ own computation using SPSS.

Findings and Recommendations

The crucial findings inferred from this study are mentioned as follows:

CEPS is a major deciding factor which generates confidence for dividend payments.

Managerial efficiency in fund utilization and operations can be a deciding factor but when estimated in terms of ROA and equity, the parameters have no considerable influence on dividend decisions.

The increase in financing activities was utilized for dividend decisions rather than supporting growth opportunities of banks.

Provision for NPAs which allocated as a prudent accounting practice must be treated diligently. It has been observed that higher provisioning is booked as noncash item increasing the cash flow from operations amidst profitability is lower. Hence, higher provisions have also initiated dividend payments.

Recently, banks have initiated the cleanup of balance sheets, and particularly, PSBs have stopped dividend payments on earnings-based decisions.

Proposed Dividend Confidence Model

The models discussed in the research gave insights into the behavior of dividend decisions through some important predictors in terms of liquidity, ownership, and efficiency. Based on the findings, it was observed that banks were initiating dividend payments which were not aligned with the growth story of the Indian Economy.

Hence, a different supervised learning approach is proposed which can be helpful for policy makers, economists, and bankers to observe the pattern and take suitable remedial measures accordingly. The model is simpler in approach and easily interpretable. The model uses neurons for simulation and the binary model will predict the dividend behavior (Figure 2).

Figure 2.

Source: Authors’ own computation.

A higher cash flow from operations but a lower provisioning will only activate the dividend decision. Otherwise, the bank will stop signaling dividend payment. The Bias is introduced to factor the management and ownership perspective.

Conclusion

The simpler approach of model can easily assist the policy makers and bankers for signaling of dividend payments. The approach has embedded provisioning of NPAs and cash flows from operations to ensure sufficient liquidity and lower risk for dividend payments. The simpler version can act as a premise for advanced modeling by researchers for risk management approach in handling the complexities of dividend payments. The research has been confined to dividend decisions and predictors surrounding the decision. Further research on capital structure and risk management in dividend decisions can be explored in this sector.

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

The authors received no financial support for the research, authorship, and/or publication of this article.

References

Anand

(2002). Corporate finance practices in India. ICFAI Journal of Applied Finance , 10(2), 5−16.

Bodla

B. S.

Karam

Sura

J. S.

(2007). Examining application of Lintner’s dividend model in Indian banking industry. Journal of Bank Management , 6(4), 40−59.

Brealey

R. A.

Myers

S. C.

(2002). Principles of corporate finance . Boston, MA: Irwin/McGraw-Hill.

Desai

M. A.

Foley

C. F.

Hines

J. R.

(2007). Dividend policy inside the multinational firm. Financial Management , 36(1), 5−26.

Guerard

J. B.

Bean

A. S.

Andrews

(1987). R&D management and corporate financial policy. Management Science , 33(11), 1419−1427.

Hagerman

R. L.

Huefner

R. J.

(1980). Earnings numbers and dividend prediction. Review of Business and Economic Research , 15(3), 39−48.

Kapoor

Mishra

Kanwal

(2010). Dividend policy determinants of Indian service sector: A factorial analysis. Paradigm , 15(1), 24−42.

Lintner

(1956). Distribution of incomes of corporations among dividends, retained earnings and taxes. American Economic Review , 46(2), 97−113.

Pal

Goyal

(2007). Leading determinants of dividend policy: A case study of the Indian banking industry. Decision , 34(2), 87−112.

10.

Tahir

S. H.

Ullah

M. R.

& Mahmood

(2015). Banks dividend policy and investment decision as determinants of financing decision: Evidence from Pakistan. American Journal of Industrial and Business Management , 5, 3113−3123.