## IMF Working Paper — “India’s Banks: Lending to Productive Firms?” — Introduction & Conclusion (wpiea2022073-print-pdf)

## Source details

**Canonical URL:** [IMF Working Paper — “India’s Banks: Lending to Productive Firms?” — Introduction & Conclusion (wpiea2022073-print-pdf)](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022073-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022073-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022073-print-pdf.pdf.json)

---

### Background and motivation
- Capital misallocation is widely thought to underpin productivity and income gaps between advanced and emerging economies.
- Prior studies suggest large capital misallocation in India, reflected in unusually high dispersion of productivity across firms.
- Credit market misallocation can channel finance to less productive firms and sustain zombie (unviable) firms, impeding entry and growth of other firms.
- Banks—particularly public sector banks (PSBs), which are majority government owned—play an outsized role in India’s credit markets, making PSB allocation behavior critical for understanding credit misallocation.
- Prior evidence on PSBs in India:
  - PSBs lend more to agriculture, rural areas, and government and less to trade, transport, and finance.
  - PSB lending allocation is sensitive to election cycles (e.g., increases in agricultural credit around elections not associated with higher output).
  - Branch deregulation can induce PSBs to reduce lending to poorly performing firms, suggesting competition improves credit allocation.
- This paper studies how well Indian banks allocate capital across firms of varying productivity and whether PSB exposure matters.

### Data and measurement
- Firm-level data: Centre for Monitoring Indian Economy (CMIE) Prowess (non-financial firms).
- Bank-level data: Database of Indian Economy (DBIE) from the Reserve Bank of India (RBI), including nonperforming assets and restructured loans and selected performance and financial soundness ratios for scheduled commercial banks.
- Matching procedure:
  - Match Prowess “Banker” names to RBI DBIE bank names by year using fuzzy matching and manual confirmation.
  - Matches over 99 percent of bank–firm links reported by Prowess in the sample.
- Main firm variables:
  - Stock of bank credit; measures of capital productivity: sales and physical capital (net plants, property, and equipment).
  - Controls: size (log of assets), age (years), sector (5-digit National Industrial Classification codes), location (district), sales growth (log growth in sales), leverage (debt to assets), interest coverage (earnings before taxes to interest expenses), ownership type (including government ownership).
- Bank-level variables constructed from RBI:
  - Ownership (public, new private, other), capitalization (capital to asset ratio), funding mix (share of non-deposit liabilities), asset quality (nonperforming plus restructured loans to assets ratio).
  - Firm-year bank variables are loan-weighted averages across banks listed by the firm.
- Sample and coverage:
  - Period: 2005–2020.
  - Sectors retained: manufacturing, electricity and gas supply, construction, wholesale and retail trade, transportation, accommodation and food services, information and communication, professional and scientific activities, administrative and support service activities, and other service activities.
  - Restrictions: firms with at least five years of non-missing data on total assets, net plant property and equipment, sales, total borrowing and bank borrowing; at least 5 firms in each 5-digit industry-year; lenders not in RBI database dropped; at least one matched bank per firm-year.

### Stylized facts
- PSBs account for a sizeable share of bank credit:
  - Since 2005, PSBs have accounted for 60–80 percent of outstanding loans.
  - Total loans grew to about 100 trillion rupees, more than 50 percent of GDP, by 2020.
  - Most remaining credit is disbursed by “new private banks”.
  - The share of new credit from PSBs varies substantially year-to-year, from less than 20 percent in some years (e.g., 2016 and 2017) to over 90 percent in other years (e.g., 2009, 2010).
- PSBs differ on observable metrics:
  - PSBs have higher nonperforming asset ratios (including restructured loans), lower Tier 1 capital ratios, and a lower share of non-deposit liabilities than new private banks.
  - PSBs and new private banks operate at broadly similar scales; both are larger than other banks.
- A significant fraction of new bank credit is extended to unproductive firms:
  - In an average year, around 45 percent of new bank credit is allocated to unproductive firms (firms with sales to physical capital ratios in the bottom tercile within industry-year).
  - Nearly all new credit goes to large firms, so credit misallocation occurs primarily within large firms.
- Substantial year-to-year variation:
  - In 2015, less than 10 percent of new credit went to unproductive firms.
  - In 2018, over 80 percent of new credit went to unproductive firms.
  - The share of new credit going to unproductive firms is correlated with the share of new credit disbursed by PSBs in the year.

### Empirical specifications
- Primary dependent variable:
  - C_it: bounded annual growth in the stock of bank credit received by firm i at time t (bounded growth can take values between -200 and 200 percentage points).
- Productivity measure:
  - Productivity_it-1: lagged ratio of sales to physical capital (sales to net plants, property, and equipment).
- Two firm-level PSB measures:
  - PSB dependence_it-1: weighted share of PSBs among a firm’s banking relationships, weights based on each bank’s share of total lending in that year.
  - PSB share of credit_t-1: proportion of new loans granted by PSBs in each year (time-varying across years but not across firms).
- Core regression (conceptual):
  - Equation (1) relates credit growth to Productivity_it-1, PSB dependence_it-1, PSB share of credit_t-1, their interactions, firm controls (size, sales growth, leverage, interest coverage, ownership), sector-time fixed effects (δ_st), and district-time fixed effects (μ_dt).
  - Key coefficient of interest: β6 on Productivity_it-1 * PSB share of credit_t-1 * PSB dependence_it-1. A negative β6 implies that when PSBs supply a large share of new credit, productive firms dependent on PSBs receive lower credit growth relative to others.
- Heterogeneity by firm size:
  - Equation (2) introduces interactions with log of assets to assess whether effects differ for large versus small firms.
  - Interest is in comparing β4 (triple interaction without size) to β8 (quadruple interaction including size).
- Controlling for other bank characteristics:
  - Equation (3) further adds bank characteristics (capitalization, funding mix, asset quality) and their interactions to test whether PSB ownership effects persist after accounting for these observable bank attributes.

### Key empirical results and findings
- When and where PSBs are more important sources of bank credit, productivity has a weaker association with bank credit growth at the firm level:
  - Firms with higher PSB dependence see stronger credit growth in years where PSBs account for a larger share of overall credit (positive interaction between PSB share of credit (t-1) and PSB dependence (i,t-1)).
  - However, this credit growth is not concentrated in productive firms: the interaction between PSB share of credit (t-1), Productivity (i,t-1), and PSB dependence (i,t-1) is negative and statistically significant, indicating a weaker productivity–credit link for PSB-dependent firms when PSBs supply more new credit.
- Effects concentrated among large firms:
  - Large firms receive the bulk of corporate credit, and the misallocation result is driven by large firms.
  - Large firms with higher PSB dependence see stronger credit growth in years when PSBs account for a larger share of overall credit.
  - Yet for large PSB-dependent firms, the link between productivity and credit growth is weaker in those years (negative and statistically significant quadruple interaction between Log assets, PSB share of credit, Productivity, and PSB dependence).
  - For small firms, credit growth is strongly and positively linked to productivity regardless of PSB dependence.
- Illustration of combined coefficient implications (described):
  - In years when PSBs account for a large share of overall credit:
    - Firms with low PSB dependence: credit growth responds strongly to productivity—low-productivity firms see very low credit growth; high-productivity firms see strong credit growth.
    - Firms with high PSB dependence: link between productivity and credit growth is much weaker; low-productivity firms with high PSB dependence obtain credit growth similar to medium-productivity firms with low PSB dependence.
  - In years when PSBs account for a small share of credit: little difference in the importance of productivity for credit growth between high and low PSB-dependent firms.
- Role of observable bank characteristics:
  - PSBs differ from other banks on capitalization, asset quality, and funding mix.
  - Adding these bank characteristics and their interactions (equation (3)) does not eliminate the role of PSB dependence: the negative and statistically significant quadruple interaction for large firms remains with similar magnitude.
  - This suggests that the effects are most likely related to PSBs’ governance and supervision or ownership per se rather than just observable performance differences.

### Counterfactual and quantitative magnitudes
- Counterfactual exercise for 2010–2014 (period with large PSB footprint and excluding global financial crisis effects):
  - More productive large firms saw credit growth of 9 percentage points per year on average.
  - Shifting half of the credit allocated by banks to large unproductive firms to more productive large firms could have raised credit growth for more productive firms from 9 to more than 13 percentage points per year (to 13.3 percentage points per year in the described counterfactual).
  - Shifting all credit allocated to unproductive firms to more productive firms could have raised credit growth for more productive firms to over 17 percentage points per year (to 17.3 percentage points per year in the described counterfactual).
  - These reallocation magnitudes imply substantial missed opportunities to relax credit constraints for productive firms and to boost aggregate productivity and growth.

### Policy implications and recommendations
- The results point to governance and supervision of PSBs, and public bank ownership, as likely drivers of credit misallocation:
  - Improving governance and supervision of PSBs or reducing public bank ownership (via privatizations) may be necessary to address misallocation of credit and to foster economic growth.
- Competition and branch deregulation may improve PSB lending allocation (evidence that branching expansion reduces PSB lending to poorly performing firms).
- Reallocating credit away from large unproductive firms toward more productive large firms could meaningfully increase credit available to productive firms and support growth.
- From the Conclusion:
  - Strengthening PSBs remains important; suggested measures include further recapitalization of PSBs, and the establishment of the National Asset Reconstruction Co (NARCL) to resolve bad PSB assets.
  - Plans announced by the government and RBI to privatize additional PSBs and to improve the quality of governance at PSBs more broadly could have an important role to play (IMF 2021).
  - Acharya and Rajan (2020) review options including independent and representative boards, both partial and full privatization, and more market-based implementation of policy mandates for the banking sector.
  - D’Souza and Surti (2021) suggest improving RBI’s supervisory powers vis a vis PSBs, including the ability to replace management and board members from PSBs, withdraw PSB bank licenses, and commence the resolution of failing PSBs.

### Robustness and scope notes
- The core empirical result—a weaker productivity–credit growth link for PSB-dependent firms when PSBs supply a large share of new credit—is robust to controlling for other important bank characteristics.
- The result is particularly driven by large firms, which account for most credit in India.

*Source: IMF Working Paper — “India’s Banks: Lending to Productive Firms?” — Introduction and Conclusion (pdf: wpiea2022073-print-pdf).*

### Introduction

### Introduction

### Background and motivation
- Capital misallocation is widely thought to underpin productivity and income gaps between advanced and emerging economies.
- In India, prior studies suggest large capital misallocation, reflected in unusually high dispersion of productivity across firms.
- Credit market misallocation can channel finance to less productive firms and sustain zombie (unviable) firms, impeding entry and growth of other firms.
- Banks—particularly public sector banks (PSBs), which are majority government owned—play an outsized role in India’s credit markets, making the allocation behavior of PSBs critical for understanding credit misallocation.
- Prior evidence on PSBs in India:
  - PSBs lend more to agriculture, rural areas, and government and less to trade, transport, and finance.
  - PSB lending allocation is sensitive to election cycles (e.g., increases in agricultural credit around elections not associated with higher output).
  - Branch deregulation can induce PSBs to reduce lending to poorly performing firms, suggesting competition improves credit allocation.
- The link between bank lending and firm-level productivity in India is not well understood; this paper studies how well Indian banks allocate capital across firms of varying productivity and whether PSB exposure matters.

### Data and measurement
- Firm-level data: Centre for Monitoring Indian Economy (CMIE) Prowess (non-financial firms).
- Bank-level data: Database of Indian Economy (DBIE) from the Reserve Bank of India (RBI), including nonperforming assets and restructured loans and selected performance and financial soundness ratios for scheduled commercial banks.
- Matching procedure:
  - Match Prowess “Banker” names to RBI DBIE bank names by year using fuzzy matching and manual confirmation.
  - Matches over 99 percent of bank–firm links reported by Prowess in the sample.
- Main firm variables:
  - Stock of bank credit; measures of capital productivity: sales and physical capital (net plants, property, and equipment).
  - Controls: size (log of assets), age (years), sector (5-digit National Industrial Classification codes), location (district), sales growth (log growth in sales), leverage (debt to assets), interest coverage (earnings before taxes to interest expenses), ownership type (including government ownership).
- Bank-level variables constructed from RBI:
  - Ownership (public, new private, other), capitalization (capital to asset ratio), funding mix (share of non-deposit liabilities), asset quality (nonperforming plus restructured loans to assets ratio).
  - Firm-year bank variables are loan-weighted averages across banks listed by the firm.
- Sample and coverage:
  - Period: 2005–2020.
  - Sectors retained: manufacturing, electricity and gas supply, construction, wholesale and retail trade, transportation, accommodation and food services, information and communication, professional and scientific activities, administrative and support service activities, and other service activities.
  - Restrictions: firms with at least five years of non-missing data on total assets, net plant property and equipment, sales, total borrowing and bank borrowing; at least 5 firms in each 5-digit industry-year; lenders not in RBI database dropped; at least one matched bank per firm-year.

### Stylized facts
- PSBs account for a sizeable share of bank credit:
  - Since 2005, PSBs have accounted for 60–80 percent of outstanding loans.
  - Total loans grew to about 100 trillion rupees, more than 50 percent of GDP, by 2020.
  - Most remaining credit is disbursed by “new private banks”.
  - The share of new credit from PSBs varies substantially year-to-year, from less than 20 percent in some years (e.g., 2016 and 2017) to over 90 percent in other years (e.g., 2009, 2010).
- PSBs differ on observable metrics:
  - PSBs have higher nonperforming asset ratios (including restructured loans), lower Tier 1 capital ratios, and a lower share of non-deposit liabilities than new private banks.
  - PSBs and new private banks operate at broadly similar scales; both are larger than other banks.
- A significant fraction of new bank credit is extended to unproductive firms:
  - In an average year, around 45 percent of new bank credit is allocated to unproductive firms (firms with sales to physical capital ratios in the bottom tercile within industry-year).
  - Nearly all new credit goes to large firms, so credit misallocation occurs primarily within large firms.
- Substantial year-to-year variation:
  - In 2015, less than 10 percent of new credit went to unproductive firms.
  - In 2018, over 80 percent of new credit went to unproductive firms.
  - The share of new credit going to unproductive firms is correlated with the share of new credit disbursed by PSBs in the year.

### Empirical specifications
- Primary dependent variable:
  - C_it: bounded annual growth in the stock of bank credit received by firm i at time t (bounded growth can take values between -200 and 200 percentage points).
- Productivity measure:
  - Productivity_it-1: lagged ratio of sales to physical capital (sales to net plants, property, and equipment).
- Two firm-level PSB measures:
  - PSB dependence_it-1: weighted share of PSBs among a firm’s banking relationships, weights based on each bank’s share of total lending in that year.
  - PSB share of credit_t-1: proportion of new loans granted by PSBs in each year (time-varying across years but not across firms).
- Core regression (conceptual):
  - Equation (1) relates credit growth to Productivity_it-1, PSB dependence_it-1, PSB share of credit_t-1, their interactions, firm controls (size, sales growth, leverage, interest coverage, ownership), sector-time fixed effects (δ_st), and district-time fixed effects (μ_dt).
  - Key coefficient of interest: β6 on Productivity_it-1 * PSB share of credit_t-1 * PSB dependence_it-1. A negative β6 implies that when PSBs supply a large share of new credit, productive firms dependent on PSBs receive lower credit growth relative to others.
- Heterogeneity by firm size:
  - Equation (2) introduces interactions with log of assets to assess whether effects differ for large versus small firms.
  - Interest is in comparing β4 (triple interaction without size) to β8 (quadruple interaction including size).
- Controlling for other bank characteristics:
  - Equation (3) further adds bank characteristics (capitalization, funding mix, asset quality) and their interactions to test whether PSB ownership effects persist after accounting for these observable bank attributes.

### Key empirical results and findings
- When and where PSBs are more important sources of bank credit, productivity has a weaker association with bank credit growth at the firm level:
  - Firms with higher PSB dependence see stronger credit growth in years where PSBs account for a larger share of overall credit (positive interaction between PSB share of credit (t-1) and PSB dependence (i,t-1)).
  - However, this credit growth is not concentrated in productive firms: the interaction between PSB share of credit (t-1), Productivity (i,t-1), and PSB dependence (i,t-1) is negative and statistically significant, indicating a weaker productivity–credit link for PSB-dependent firms when PSBs supply more new credit.
- Effects concentrated among large firms:
  - Large firms receive the bulk of corporate credit, and the misallocation result is driven by large firms.
  - Large firms with higher PSB dependence see stronger credit growth in years when PSBs account for a larger share of overall credit.
  - Yet for large PSB-dependent firms, the link between productivity and credit growth is weaker in those years (negative and statistically significant quadruple interaction between Log assets, PSB share of credit, Productivity, and PSB dependence).
  - For small firms, credit growth is strongly and positively linked to productivity regardless of PSB dependence.
- Figures (described in text) illustrate combined coefficient implications:
  - In years when PSBs account for a large share of overall credit:
    - Firms with low PSB dependence: credit growth responds strongly to productivity—low-productivity firms see very low credit growth; high-productivity firms see strong credit growth.
    - Firms with high PSB dependence: link between productivity and credit growth is much weaker; low-productivity firms with high PSB dependence obtain credit growth similar to medium-productivity firms with low PSB dependence.
  - In years when PSBs account for a small share of credit: little difference in the importance of productivity for credit growth between high and low PSB-dependent firms.
- Role of observable bank characteristics:
  - PSBs differ from other banks on capitalization, asset quality, and funding mix.
  - Adding these bank characteristics and their interactions (equation (3)) does not eliminate the role of PSB dependence: the negative and statistically significant quadruple interaction for large firms remains with similar magnitude.
  - This suggests that the effects are most likely related to PSBs’ governance and supervision or ownership per se rather than just observable performance differences.

### Counterfactual and quantitative magnitudes
- Counterfactual exercise for 2010–2014 (period with large PSB footprint and excluding global financial crisis effects):
  - More productive large firms saw credit growth of 9 percentage points per year on average.
  - Shifting half of the credit allocated by banks to large unproductive firms to more productive large firms could have raised credit growth for more productive firms from 9 to more than 13 percentage points per year (to 13.3 percentage points per year in the described counterfactual).
  - Shifting all credit allocated to unproductive firms to more productive firms could have raised credit growth for more productive firms to over 17 percentage points per year (to 17.3 percentage points per year in the described counterfactual).
  - These reallocation magnitudes imply substantial missed opportunities to relax credit constraints for productive firms and to boost aggregate productivity and growth.

### Policy implications and recommendations (as presented)
- The results point to governance and supervision of PSBs, and public bank ownership, as likely drivers of credit misallocation:
  - Improving governance and supervision of PSBs or reducing public bank ownership (via privatizations) may be necessary to address misallocation of credit and to foster economic growth.
- Competition and branch deregulation may improve PSB lending allocation (evidence that branching expansion reduces PSB lending to poorly performing firms).
- Reallocating credit away from large unproductive firms toward more productive large firms could meaningfully increase credit available to productive firms and support growth.

*Source: IMF Working Paper — “India’s Banks: Lending to Productive Firms?” — Introduction (pdf: wpiea2022073-print-pdf - Introduction).*

### Conclusion

### Conclusion

### Key findings on credit allocation and productivity
- The Indian banking sector is characterized by a large presence of PSBs.
- The paper examines the link between firm productivity and credit growth, and how that link depends on:
  - the extent to which firms maintain banking relationships with PSBs, and
  - the share of new credit in the economy provided by PSBs.
- The estimations show that the link between productivity and credit growth is weaker for PSB dependent firms in years when PSBs represent a large share of new credit in the economy.
- This weakening is driven by large firms, which account for most of the credit in India.
- The main result is robust to controlling for other important bank characteristics.

### Policy implications and recommendations
- Strengthening PSBs remains important; suggested measures discussed in the conclusion include:
  - further recapitalization of PSBs, and
  - the establishment of the National Asset Reconstruction Co (NARCL) to resolve bad PSB assets.
- Plans announced by the government and RBI to privatize additional PSBs and to improve the quality of governance at PSBs more broadly could have an important role to play (IMF 2021).
- Acharya and Rajan (2020) review a broad range of options to improve governance of PSBs, including:
  - independent and representative boards,
  - both partial and full privatization, and
  - more market-based implementation of policy mandates for the banking sector.
- D’Souza and Surti (2021) suggest improving RBI’s supervisory powers vis a vis PSBs to help reduce credit misallocation. In particular, the paper suggests that, as with private banks, RBI should have the ability to:
  - replace management and board members from PSBs,
  - withdraw PSB bank licenses, and
  - commence the resolution of failing PSBs.

### Robustness and scope notes
- The core empirical result—weaker productivity–credit growth link for PSB-dependent firms when PSBs supply a large share of new credit—is robust to controlling for other important bank characteristics.
- The result is particularly driven by large firms, which account for most credit in India.

*Source: Conclusion, wpiea2022073-print-pdf*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022073-print-pdf.pdf_