The Sectoral Impact of Monetary Policy Transmission in India: A Panel VAR Approach

Abstract

A structural panel vector autoregression (VAR) analysis is done to analyze the impact of monetary policy shock on people associated with various occupations. To understand the efficacy of bank lending channel, it is important to capture the differential occupation-wise effect of interest rate. The article finds that due to a monetary policy shock, the impulse responses capture the movement of loans in theoretically expected direction in most cases. The Granger causality tests successfully establish the long-run relationship between loans and interest rate. Also, the empirical results of good-performing states support the direct link between greater financial penetration and higher economic activity. Monetary policy shock significantly affects the lending behavior in all the sectors, except in agriculture and personal loans sector. The weak link of transmission in these sectors is mainly attributed either to lack of access to formal credit or a preference to informal credit sources over banks.

Keywords

Monetary policy bank lending channel panel VAR priority sector lending financial inclusion

Introduction

After the onset of recession in 2008, the standard New Keynesian dynamic stochastic general equilibrium (DSGE) model was highly criticized. In such a model, interest rates emerged as the main policy tool. According to this theory, during recession (boom) the central bank should adopt a loose (tight) monetary policy identified by a reduction (increase) in interest rates. The fall (increase) in interest rates is expected to shift the aggregate demand up (down) by encouraging more (less) consumption and more (less) investment which in turn, will increase (reduce) the total output of the economy. Based on this theory post 2008 many central banks have adopted low interest rate policies with some economies resorting to a target of zero nominal interest rate to even negative interest rates.¹

For example, the last decade witnessed the Federal Reserve Bank adopting a zero interest rate policy and the European Central Bank and Bank of Japan adopting negative interest rate policy.

The main objective of such a policy was the dire need to boost the aggregate demand by reducing the interest rates as much as possible and so much so to an extent that the lower bound on nominal interest rate was also violated. However, none of the economies exhibited higher employment or rise in inflation or improvement in their economic activity. The interest rate policy’s role in achieving full employment and higher growth was deemed ineffective.

A standard New Keynesian DSGE model does not have a banking sector. This is due to the fact that inclusion of banking sector makes the model very complex and intractable. So the typical model that a central bank uses to set its monetary policy is not able to capture the actual role of its banking sector in the economy. In other words, how does a change in policy rates affects the banking sector’s ability to lend to different sectors of the economy was not fully rationalized by these countries. People started to ask the following questions. Is credit really flowing to the sectors that have the ability to generate maximum employment and demand? Are these sectors sensitive to interest rate changes?

Hence the efficacy of monetary policy is established by how significantly and to what extent it affects the credit flows to various sectors of the economy. The reduced sensitivity of consumers and businesses to interest rate witnessed in the USA, UK, and Japan has imposed important considerations for policymakers. To understand the situation in the Indian context, in this article we examine how a change in Reserve Bank of India’s (RBI) monetary policy stance affect lending in various key sectors of the Indian economy such as total agriculture, agri-direct finance, agri-indirect finance, industry, professional, and other services and personal loans. These sectors have the ability to generate huge employment and many of them are even categorized as priority sectors by the RBI. A change in interest rate is expected to affect their scale of operation. In addition to the usual sectors, loans in agriculture, personal loans such as loans for housing and loans for purchase of consumer’s durables etc. are extremely crucial in determining the aggregate demand of the economy.

We apply panel VAR estimations to assess the role of monetary policy shocks on bank lending to people associated with various occupations. The dataset is obtained from Basic Statistical Returns, RBI on ‘amount outstanding’ that is lending in agriculture, industry, professional, and other services and personal loans categories. A panel VAR analysis is suitable in this article to study how shocks are transmitted across different states in India in regards to lending made to people associated with different occupations. Panel VARs are standard VARs added with a cross-sectional dimension (Canova & Ciccarelli, 2013). They have evolved as a very powerful tool as they capture both static and dynamic dependencies among countries, sectors, and regions. In order to draw comparisons we run the estimations on two samples: whole sample (all the states and union territories in India) and best performing states. The best performing states are chosen based on their value of net state domestic product (NSDP) in the last decade. The best performing states category included Maharashtra, Gujarat, Haryana, and Tamil Nadu. Particularly, post 2000 there have been numerous steps taken by the center and the RBI especially to develop and expand the banking provisions in the country. A comprehensive analysis on the lending activity and how it is affected by the monetary policy will provide meaningful insights and guidance for future policy decisions.

Using a heterogeneous panel VAR, Mishra, Montiel, Pedroni, and Spilimbergo (2015) analyzed the strength of bank lending channel measured by pass-through from policy rate to lending rate for group of advanced, emerging, and low-income economies. They show that bank lending channel is very weak in poor countries, which are usually characterized by poor institutional environment and uncompetitive banking sector. In case of India, many studies have established the significant role of interest-rate channel, with Pandit, Mittal, Roy, and Ghosh (2006) have also established the existence of bank lending channel. A contractionary monetary policy shock decreases the amount of credit available in the economy through a reduction in supply of bank loans (known as the bank lending channel) or through a fall in borrower’s demand (known as the interest-rate channel). Nachane, Ray, and Ghosh (2002) using structural VAR approach tried to find out the reasons for differential responses of different states in India to a monetary policy shock. They concluded that the strength of interest-rate channel directly depends on the share of manufacturing in NSDP and the level of financial deepening of each state. They also provided support for the fact that states with more number of small firms respond more to monetary policy shock. However, how a monetary policy shock affect the credit availability to people from different occupations and how this occupation-wise lending differ across different samples is unknown for India and is the topic of interest of the current article.

We perform Granger causality tests, impulse response analysis, and variance decomposition analysis to evaluate the impact of interest rates on occupation-wise lending patterns. The results of the study are summarized as: (a) Granger causality tests successfully establish the two-way long-run relationship between loans and interest rate for majority of the samples, (b) the impulse responses capture the movement of loans in theoretically expected direction due to a monetary policy shock providing evidence of correct identification of monetary policy, (c) the better performing states show greater variability in loans in different occupation for variability in interest rates establishing the direct link between well-functioning banking sector and greater economic activity, (d) loans in sectors like industry and professional and other services are affected significantly by the change in interest rates for both the samples, (e) agricultural sector is largely immune to changes in interest rates which can be attributed to the limited direct access to bank credit for the Indian farmers or a blind preference for informal credit sources, and (f) personal loans including housing loans and loans in consumer durables are not affected by changes in interest rates providing evidence for the weak link of the bank lending channel.

The invulnerability to interest rates in the agricultural sector and in personal loans including housing loans and loans in consumer durables could be explained by both supply and demand side. The argument behind low demand for financial services can be low income, financial illiteracy, lack of awareness, social exclusion and so on; on the other hand, supply side factors include dearth of bank branches in vicinity, absence of financial packages which are suitable for poor people, low profitability, language barrier, high operation cost and so on. Ease of availability of credit from informal sources such as money lenders and landlords compared with requirement of proper documentation in case of formal sources is an important reason why Indian farmers prefer informal credit. Moreover, schemes introduced by the Government of India like loan default waiver and interest subvention schemes to support the farmers may have reduced the sensitivity of loans in the agricultural sector to monetary policy.

A very important channel through which interest rate affects aggregate demand is by affecting the consumer spending on durables and affecting the housing loans which are supposed to be interest rate sensitive. Both the sectors have inter-linkages across many industries and hold key to the strength of monetary policy transmission. However we find weak link between these variables (which are part of personal loans) and interest rate. Various reasons can be associated with such impaired relation including the fact that a change in policy rates can affect aggregate demand only when the change is passed on to the end consumers. Moreover housing finance in India is yet to evolve and even if financial sector reforms have increased the institutional credit flow into the sector, majority of housing finance is still happening with black money across the country. In absence of any regulatory authority in the sector, non-transparent financing practice in housing market is rampant.

Panel VAR Estimation

Panel VARs assume that all variables are endogenous and interdependent. VAR models also make the same assumption. However, panel VARs are added with a cross-sectional dimension to it. As quoted from Canova and Ciccarelli (2013), panel VARs have emerged as an important tool as they are able to ‘(a) capture both static and dynamic interdependencies, (b) treat the links across units in an unrestricted fashion, (c) easily incorporate time variations in the coefficients and in the variance of the shocks, and (d) account for cross-sectional dynamic heterogeneities’.

A panel VAR can be represented as follows:

y_{i t} = B_{0} + B (l) y_{i t - 1} + f_{i} + u_{i t}, i = 1, \dots . N; t = 1, \dots . T_{i}

(1)

where y_it is a vector of endogenous variables, B₀ is vector of constants, B(l) is a polynomial in the lag operator and f_i is state specific fixed effects. u_it is random disturbances vector with u_t ∼ iid (0, ∼ _u ) that is independently and identically distributed.

In our analysis, i indicates different Indian states. y_i is a two variable vector {Interest Rate, Loans}. Interest rate is the call-money rate (the short-term interest rate) taken from Organisation for Economic Co-operation and Development (OECD) database capturing the monetary policy of RBI. The data on loans are the ‘amount outstanding’ obtained from Basic Statistical Returns, RBI for lending in agriculture, agri-direct, agri-indirect, industry, professional, and other services and personal loans categories. It is an annual dataset and covers the period from 2002 to 2014. The logarithmic value of loan is used in the VAR.

In the panel structure, fixed effects control for individual heterogeneity. Conventionally the mean-differencing approach is employed to remove them. However, applying this approach may lead to biased coefficients as the fixed effects are assumed to be correlated with the regressors. The solution is to adopt forward-mean differencing or the ‘Helmert procedure’ (see Arellano & Bover, 1995). This procedure removes only the mean of all the future observations available for each state-year and preserves the orthogonality between transformed variables and lagged independent variables. Hence lagged regressors can be used as instruments and the system is estimated by generalized method of moments (GMM) (Love & Zicchino, 2006).

The variance–covariance matrix of errors in Equation (1) may not be diagonal and it is difficult to isolate the shocks to any one of the variables. Hence, in order to recover the structural parameters from the reduced form model given by Equation (1), we use the Cholesky decomposition of reduced from innovations such that they become orthogonal as suggested by Sims (1980). This imposes a recursive structure defined by Equation (2) to identify the model.

(\begin{matrix} \in_{t}^{I n t e r e s t R a t e} \\ \in_{t}^{L o a n s} \end{matrix}) = (\begin{matrix} . & 0 \\ . & . \end{matrix}) (\begin{matrix} u_{t}^{I n t e r e s t R a t e} \\ u_{t}^{L o a n s} \end{matrix})

(2)

Where u_t is reduced form innovations as given in Equation (1) and ∈ _t is the structural form innovations that we derive using Cholesky ordering as in Equation (2). The identifying assumption used in Equation (2) is that the monetary policy captured by the interest rate is not contemporaneously affected by the amount of loans made. That is the central bank when setting its monetary policy takes the credit flow in the previous periods in to account but not the current amount of loans made. The rationale behind this assumption is that the central bank may not have immediate access to the data on lending activity for the current period when it is setting the monetary policy. However, credit flowing in to the economy is contemporaneously affected by the monetary policy shocks. Both the variables are allowed to affect each other with a lag.

Several panel VAR estimations are performed to assess the impact of monetary policy shocks on the total amount of credit lent to people from different occupations. Two different samples are used to assess and compare the results: whole sample consisting of all the Indian states and Union Territories and good-performing states.

In the following section we offer a detailed panel VAR analysis comparing the performance of models of different samples for each selected occupation. GMM style estimations are done by implementing panel VAR package developed by Abrigo and Love (2015). Lag 1 is selected for all cases based on the minimized values of the consistent moment and model Akaike information criteria (MAIC), the consistent moment and model Bayesian information criteria (MBIC) and the consistent moment and model Hannan-Quinn information criteria (MQIC). There are few possible cases where MBIC, MAIC, and MQIC are not minimized for lag 1. However, we continue our analysis with lag 1 for two reasons: first, the shocks are correctly identified given by the impulse response functions moving in the theoretically expected direction and second, uniform lag allows better comparisons across sectors. Please refer to Table A1 in the appendix for the lag selection criterion. Table A2 in the appendix presents the eigenvalue stability condition for the panel VAR models. Stability requires the eigenvalues should be inside the unit circle. Our model satisfies this condition.

Panel VAR–Granger Causality Wald Test

In this section, Tables 1 and 2 provide the results (p-value) from the Granger causality tests. A variable x is said to Granger-cause variable y if, given the past values of y and past values of x, y can be predicted better compared with the prediction of y done using only the past values of y. In order to perform the Granger causality test, y is regressed on its own lagged values and on lagged values of x. Then we test the null hypothesis that the estimated coefficients on all lagged values of x are jointly zero. A p-value of less than 0.05 fails to reject the null hypothesis, and the alternative hypothesis of existence of Granger causality is accepted.

Table 1 (panel A) reports the Granger Causality tests for the null hypothesis that monetary policy does not Granger-cause lending in the different sectors and Table 1 (panel B) reports the Granger causality tests for the null hypothesis that lending in the different sectors does not Granger-cause monetary policy. For all occupations in both samples comprising all states and good-performing states, we reject the null hypothesis. That is, monetary policy Granger-cause the lending activity as reported by p-value of 0.00 in total agriculture, agri-direct, agri-indirect, professional, and other services, industry and personal loans in both the samples.

Table 1.

Granger Causality Tests

Null	Panel A: Monetary Policy Does Not Granger-cause Occupation Loan Variable p-values		Panel B: Occupation Loan Variable Does Not Granger-cause Monetary Policy p-values
Occupation	All States	Good Performing States	All States	Good Performing States
Agriculture	0.00	0.00	0.18	0.07
Agri-direct finance	0.00	0.00	0.02	0.00
Agri-indirect finance	0.00	0.00	0.18	0.99
Industry	0.00	0.00	0.00	0.00
Professional & other services	0.00	0.00	0.01	0.01
Personal loans	0.00	0.00	0.12	0.20
Total	0.00	0.00	0.00	0.00

Source: The author.

The null is rejected for total occupation-wise lending (see last row of Table 1 [panel B]). That is total lending in the economy Granger-cause the interest rates. Similarly, null is rejected for lending in agri-direct finance, industry, and professional and other services. Lending in agri-direct finance, industry, and professional and other services Granger-cause the monetary policy. On the contrary, lending in agriculture, agri-indirect finance, and personal loans does not Granger-cause interest rates. All the aforementioned results presented remain robust to both the sample of all states and the sample of good states.

Panel VAR–Impulse Response Analysis

Figure 1 shows the impulse response functions from the panel VAR estimation for the model consisting of total loans and interest rate. Panels A and B represent the impulse response functions for estimated model for the all states and good-performing states. The top graphs presented in the top row of Figure 1 give the impulse responses of total loans and interest rate, given a shock to total loans for sample of all states and good-performing states. Similarly, the bottom graphs presented in the bottom row of Figure 1 give the impulse responses of total loans and interest rate, given a shock to monetary policy (interest rate) for both the samples.

A positive shock to interest rate (contractionary monetary policy shock) causes a significant decline in total lending in the economy (see bottom row of Figure 1). A monetary policy shock of 0.8 percent causes total loans to significantly decline by almost 0.15 percent and around 0.13 percent for all states and for good-performing states, respectively. The peak decline in total loans in the samples happens after the 5th period.

On the contrary, a positive shock to total lending will prompt the RBI to respond by increasing the policy rates significantly captured by the top graphs in panels A and B. Around 0.17 percent increase in total loans in panel A causes the interest rate to rise significantly by 0.18 percent. Around 0.13 percent increase in total loans in panel B of good-performing states causes the interest rate to rise up significantly to 0.12 percent. The impulse responses are shown to move in the theoretically expected direction for both the samples. Hence, our estimated two variable panel VAR model consisting of lending and interest rate (monetary policy) identifies the lending shock and the monetary policy shock correctly.

Figure 2 provides the impulse responses for the model comprising total agricultural loans and interest rate. The top graphs presented in the top row of Figure 2 give the impulse responses of agricultural loans and interest rate, given a shock to agricultural loans for sample of all states and good-performing states. The bottom graphs presented in the bottom row of Figure 2 give the impulse responses of agricultural loans and interest rate due to a shock to monetary policy for both the samples. In a similar setting, Figure 3 provides the impulse responses for the model with loans in agri-direct finances and interest rate where the top row captures the impact of shocks to loans in agri-direct finances and the bottom row captures the impact of shocks to monetary policy.

Figure 1.

Source: The author.

Due to a monetary policy shock of 0.75 percent, agricultural loan fall by 0.04 percent in the whole sample and fall by 0.12 percent in the sample of good-performing state. However, the impulse responses are insignificant for both the samples in the analysis for agricultural loans. Shocks to monetary policy show higher responses of loans in agri-direct finance than loans in total agriculture (Figure 3). For the whole sample, lending reduces by 0.1 percent and for good-performing state, lending reduces to 0.14 percent. Although the responsiveness of lending is higher in the agri-direct sector compared with total agricultural lending, however, the impulse responses of loans remain insignificant due to a monetary policy shock. A similar comparison holds for loans in agri-indirect sector (see Figure A1 in appendix) where loans are far less sensitive to interest rates compared with loans in total agricultural and agri-direct finance. These results hold robust to sample of both all states and good-performing states.

Figures 4–6 in the appendix provide the impulse responses for the models along with interest rate appearing with agri-indirect loans, industrial loans, professional and other services loans, respectively. Due to a monetary policy shock of approximately 0.8 percent, industrial loans (Figure 4) falls to 0.05 percent in the first year for the sample of all states vis-à-vis a fall of 0.1 percent in the first year for the good-performing states. By the fifth period, the fall in the industrial loans was 0.18 percent for sample of all states whereas the fall was 0.2 percent in the sample of good states. The fall in industrial loans is greater and more significant for the good-performing states than the sample of all states.

Figure 2.

Source: The author.

Figure 3.

Source: The author.

Similarly, due to a monetary policy shock of approximately 0.8 percent, loans in professional and other services (Figure 5) falls to 0.08 percent in the first year for the sample of all states vis-à-vis a fall of 0.12 percent in the first year for the good-performing states. By the fifth period, the fall in loans was 0.2 percent for sample of all states whereas the fall was 0.28 percent in the sample of good states. The fall in professional services loans is greater and more significant for the good-performing states than the sample of all states. Due to a monetary policy shock, we get insignificant responses of personal loans (Figure 6). The lending in personal loans sector comprises of lending for purchase of consumer durables and housing loans. Both are expected to be interest sensitive and are believed to be the major influencer of the aggregate demand. The insensitivity of loans to interest rate in the personal loans category reflects the inability of the Indian monetary policy to push the aggregate demand in the targeted direction. It is suggested that policy makers should be cautious when using interest rate as the only monetary policy instrument and should use additional measures in conjunction with this to achieve the desired objectives.

The variance decomposition of loans in various sectors to monetary policy shock is evaluated in Table 2 (panel A) and variance decomposition of monetary policy to loan shocks is evaluated in Table 2 (panel B). The model results with both the samples of states are presented. For the model with total loans for sample of all states (Table 2 [panel A]), monetary policy shock explains 2 percent of variation in total loans in the 2nd period, increasing to 10 percent in the fourth period. The contribution of monetary policy shocks gradually increases to 24 percent and 29 percent in explaining the variation in the total loans in 8th period and 10th period ahead. As expected, for the sample of good-performing states, the variations in total loan shocks in response to variations in interest rate shocks are much greater. Monetary policy shocks explain 29 percent of variation in total loan shocks in the 4th year rising to around 48 percent in the 10th year.

Figure 4.

Source: The author.

Figure 5.

Source: The author.

Figure 6.

Source: The author.

Table 2.

Forecast Error Variance Decomposition

		Panel A: Loans to Monetary Policy Shocks		Panel B: Monetary Policy to Loan Shocks
Sector	Horizon	All States	Good States	All States	Good States
Agriculture	2	1	1	1	0
	4	0	8	6	1
	8	1	22	23	7
	10	1	27	31	10
Agri-direct finance	2	0	3	1	0
	4	1	15	5	3
	8	3	29	23	10
	10	4	33	32	13
Agri-indirect finance	2	1	1	1	0
	4	0	1	6	3
	8	1	1	23	11
	10	1	1	31	15
Industry	2	0	15	2	1
	4	2	37	15	4
	8	5	53	49	12
	10	7	56	60	15
Professional & other services	2	1	16	1	0
	4	3	31	10	3
	8	7	47	37	12
	10	9	51	47	15
Personal loans	2	0	1	0	0
	4	2	3	0	1
	8	5	6	1	3
	10	6	7	2	4
Total	2	2	10	0	0
	4	10	29	2	2
	8	24	45	9	7
	10	29	48	13	9

Source: The author.

Panel VAR–Variance Decomposition

For the same model with total loans in the sample of good states, even more variation in loans is explained by monetary policy shock. Due to an interest rate shock we get 10 percent of variations in total loans in the 2nd period, increasing to 29 percent, 45 percent, and 48 percent in the 4th, 8th, and 10th period, respectively. There is much bigger contribution of monetary policy shocks to the variations in lending in the sample of good states for all the occupations including sectors like agriculture, agri-direct finance, industry, professional, and other services. While variations sssin industrial loans and loans for professional services are similar and by the 10th period it is over 50 percent. The variation in agricultural loans is also substantial and goes up to 27 percent in the 10th period of analysis. However the role of monetary policy shocks is negligible for the variations in personal loan and agri-indirect finance for the good-performing states, and the same fact hold true for the sample of all the states. Overall, good-performing states show a greater variability in loans from different sectors to fluctuations in interest rates. This provides evidence and support for greater interest rate pass through and better developed banking sector in these states.

Comparing good-performing states with all states, monetary policy shocks cause much more fluctuations in total agricultural lending especially lending in agri-direct finance for the good-performing states. Although starting on a similar note, by the 8th year, 22 percent of agricultural lending explained by monetary policy shock for the good-performing states compared with 1 percent for all states and the trend continues on the 10th year, where 27 percent of lending explained by monetary policy shock for better-off states vis-a-vis 1 percent for the all India case. The comparative figures for agri-direct lending are 15 percent, 29 percent, and 33 percent for the better-off states and 3 percent, 29 percent, and 33 percent for the overall states for the 4th, 8th, and 10th year, respectively.

The variance decomposition of monetary policy to shocks to loans in various sectors is evaluated in Table 2 (panel B). A shock to total loans in both the samples of all states and good-performing states does not cause interest rates to fluctuate substantially. In the 4th year, interest rate only fluctuates by 2 percent due to a shock in total loans. It increases slightly to 9 percent and 7 percent, respectively for all the states and good-performing states in the 8th year and increasing further to 13 percent and 9 percent, respectively, in the 10th year. Interest rate fluctuates significantly for the all India sample with increase in lending in agriculture, agri-direct, agri-indirect, industry, professional, and other services vis-à-vis the good-performing states sample where the interest rate fluctuations are relatively less compared with all India sample. Shocks to personal loans, however, do not contribute to interest rate fluctuations in any of the sample.

Conclusion

Indian economy has travelled a long route from a stage with double digit inflation, low growth rate to present status with inflation below 4 percent, growth rate being more than 7 percent and almost stable value of rupee per dollar in 2015. There have been numerous steps taken by RBI to strengthen the financial sector of the economy. However as the census report of 2011 suggests only 58.7 percent of households in India avail of banking services, India has to go a long way to achieve universal financial inclusion. These facts are reflected in our Panel VAR estimations. While loans in some of the sectors like industry and professional and other services are sensitive to interest rate changes, loans in sectors like agriculture and personal loans paint a different picture. The immunity of such loans to monetary policy hints lack of financial outreach with regards to certain occupations and geographical regions and/or a preference on informal sources over banks for various reasons. The results hint at the possible ineffectiveness/weakness of the Indian monetary policy in moving the aggregate demand in the desired direction to affect output and employment.

The results from the sample of good-performing states empirically validate the positive correlation between availability of greater financial services and higher real activity. This presses the need for the RBI to take more steps to achieve greater financial inclusion like exploiting the extensive coverage of fair price shop, post offices in rural areas to provide basic banking functions, make financial services more mobile friendly as there exist a large user base of mobile phones in rural areas and otherwise, improving on infrastructure in isolated areas, enhancing financial literacy by creating awareness among masses, encouraging bank-self-help group linkages and microfinance institutions to expand their operations.

Declaration of Conflicting Interests

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

Funding

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

Appendix

Table A1.

Lag Selection Test

	Lags	All States			Good-performing States
	Lags	MAIC	MBIC	MQIC	MAIC	MBIC	MQIC
Agriculture	1	36.04	−8.37	18.26	−9.81	−28.81	−16.44
Agriculture	2	42.52	12.91	30.67	−4.44	−17.10	−8.86
Agri-direct finance	1	41.84	−2.57	24.07	−7.50	−26.50	−14.13
Agri-direct finance	2	44.38	14.78	32.53	−4.40	−17.07	−8.82
Agri-indirect finance	1	41.25	−2.95	23.55	−14.82	−33.82	−21.45
Agri-indirect finance	2	30.64	1.17	18.84	−9.97	−22.63	−14.39
Industry	1	40.64	−3.77	22.87	−4.52	−23.52	−11.15
Industry	2	35.92	6.32	24.08	−2.78	−15.45	−7.20
Professional & other services	1	57.51	13.10	39.73	−7.28	−26.28	−13.91
Professional & other services	2	21.84	−7.77	9.99	−10.47	−23.14	−14.89
Personal loans	1	86.63	42.23	68.86	−8.35	−27.35	−14.98
Personal loans	2	5.68	−23.92	−6.17	−6.32	−18.99	−10.74
Total	1	48.24	3.84	30.47	−7.65	−26.66	−14.29
Total	2	23.12	−6.48	11.28	−9.10	−21.77	−13.52

Source: The author.

Table A2.

Eigenvalue Stability Condition

	All Sample	Good-performing States
Agriculture	0.97	0.90
Agriculture	0.81	0.88
Agri-direct finance	0.99	0.94
Agri-direct finance	0.78	0.74
Agri-indirect finance	0.97	0.90
Agri-indirect finance	0.81	0.90
Industry	0.97	0.98
Industry	0.84	0.66
Professional & other services	0.92	0.86
Professional & other services	0.87	0.86
Personal loans	0.91	0.90
Personal loans	0.81	0.68
Total	0.95	0.95
Total	0.84	0.68

Source: The author.

Figure A1.

Source: The author.

References

Abrigo

M. R. M.

Love

(2015). Estimation of panel vector autoregression in Stata: A package of programs (University of Hawaii Working Paper). Honolulu: University of Hawaii.

Arellano

Bover

(1995). Another look at the instrumental variable estimation of error component models. Journal of Econometrics , 68(1), 29–51.

Canova

Ciccarelli

(2013). Panel vector autoregressive models: A survey (Working Paper Series No. 1507). London: Centre for Economic Policy Research.

Love

Zicchino

(2006). Financial development and dynamic investment behaviour: Evidence from panel VAR. The Quarterly Review of Economics and Finance , 46(2), 190–210.

Mishra

Montiel

Pedroni

Spilimbergo

(2015, November). Monetary policy and bank lending rates in low-income countries: Heterogeneous panel estimates. Journal of Development Economics , 111(C), 117–131.

Nachane

Ray

Ghosh

(2002 November 23). Does monetary policy have differential state level effects? An empirical evaluation. Economic & Political Weekly , 37(47), 4723–4728.

Pandit

B. L.

Mittal

Roy

Ghosh

(2006). Transmission of monetary policy and the bank lending channel: Analysis and evidence for India (Study No. 25). Mumbai: Department of Economic Analysis and Policy, RBI.