New Banks, credit, and misallocation

In the late 1990s, India opened its banking sector to new private banks – both domestic and foreign. This article shows that this resulted in greater flow of credit – but only for big firms. Higher loan amounts reduced capital market distortions for big firms, improving within-firm productivity with no change in between-firm resource allocation. Overall, there was a gain of 3-5% in manufacturing outputs attributable to the new banking relationships.

Developing countries typically have inefficient public-sector banks, resulting in high borrowing costs and limited access to finance for many firms.¹ Opening up the banking sector to competition is often a proposed way of removing these supply-side constraints. Under the 1998 World Trade Organization (WTO) services trade agreement – General Agreement on Trade in Services (GATS) – India opened up its banking sector to both domestic private and foreign banks. Figure 1 presents the districts in India with new domestic private and foreign bank branches at two different points in time.

The difference is striking. The period between 2001 and 2007 saw exponential growth in the number of new bank branches in the country. In particular, during 1995-2000 the average number of bank branches was around 150. Between 2001 and 2007, it increased to more than 1,000. Twelve new private banks and 1,100 new branches of those banks were added; on the other hand, 17 new foreign banks and 89 new foreign bank branches were opened.

Figure 1. Domestic private and foreign bank branches, district level, 2000 and 2007

Note: The left panel shows the cumulative concentration of domestic private/foreign bank branches across Indian districts for the year 2000. The right panel shows the same for 2007.

Did the entry of new banks impact the amount of credit received by firms?

To examine whether this increase in the number of branches also led to a subsequent increase in the amount of borrowing or loans received by firms, we exploit data from the Centre for Monitoring Indian Economy’s PROWESS database, which gives detailed information on the banking or credit relationships of each firm over time. We use this information on firm-bank relations to construct a binary variable at the firm level that takes value 1 if a firm starts a credit relationship with at least one of the new domestic private and/or foreign banks given that the firm had no previous relationship with any such bank. Therefore, our ‘control’ or comparison group is the sample of firms who never had any relation with any of the new banks.

Controlling for prior public-sector banking relationships, we find that relationships with new domestic private and foreign banks result in firms experiencing about 5-10% increase in the amount of new credit received, which is around 2% of their assets.² This finding is completely driven by firms larger than the 75^th percentile of the size distribution.³ Figure 2 plots the coefficients, relative to the year when a new banking relation started, on credit received by fourth quartile firms (top 25% firms by size) for five years preceding and five years after the formation of a new credit relationship with a bank. The plot clearly shows that while there were no obvious differences in the amount of loans received by these firms before the beginning of a new relationship, it starts to get significantly different from the second year onward when a new relationship is formed with at least one of the new banks.⁴

Figure 2. Loans received by top 25% largest firms with new banking relationships, 1996-2007

Notes: (i) The coefficient plot represents the differences in the amount of loan received by fourth quartile (or big firms) when forming new banking relationships with a new bank compared to firms, which did not have any relation with a new bank throughout the years of our study. In effect, we plot a ‘staggered differences-in-differences’ graph as different firms formed new credit relationships with new bank branches at different points in time. (ii) Our coefficient estimates are controlled for firm-level covariates (such as, age and age-squared of a firm), firm fixed effects, industry-year fixed effects, and public-sector bank branch trends. (iii) Standard errors clustered at the firm and year level.

Our results imply that the introduction of the new domestic private and/or foreign banks led to what the literature terms as ‘cherry picking’ or ‘cream skimming’: a lending strategy that involves extending credit only to the wealthy and transparent segments of the credit market which are primarily the big firms (Detragiache et al. 2008, Beck and Peria 2010), while excluding segments that comprise of less wealthy and/or marginal borrowers (Berger and Udell 1998). In particular, Detragiache et al. (2008) in the context of developing countries point out that “countries with larger foreign bank presence have shallower credit markets.” Other studies have also found that presence of new private banks does not necessarily enhance overall credit availability and may aggravate the conditions of credit constraints rather than alleviating such constraints, mainly for smaller firms (see Khwaja and Mian (2008) for Pakistan, Beck and Peria (2010) for Mexico, Gormley (2010) for India, and Lin (2011) for China).

What happened to allocation of resources?

One question immediately crops up: given that the credit flows from the new banks went only to the big or top quartile of firms, how did they affect the allocation of resources? To check, we follow Bartelsman et al. (2013) and create a ‘misallocation measure’ at the industry level (a covariance term between total revenue productivity (TFPR) and sales)⁵ and run a simple correlation (unconditional) with the intensity of the new banking relationships (number of banking relations/total sales at the industry level) across all 2-digit industries⁶in Figure 3. If the new bank relationships are associated with a reduction in misallocation, then more resources should have flowed to more productive industries and therefore, the covariance term would have been positively correlated with the intensity of new bank relationships. We find no such evidence – the correlation is almost zero.

Figure 3. Covariance of revenue productivity and new bank relations intensity, 1995-2007

Note: The figure plots the unconditional correlation between the covariance term of the TFPR measure (indicating misallocation) and the intensity of new bank relations (measured by the number of new banking relationships (scaled by sales) at the 2-digit industry level) across the period 1995 to 2007.

To find the cause of this absence of correlation between misallocation and new bank relationships, we use the Hsieh and Klenow (2009) ‘misallocation accounting framework’ to disaggregate the overall misallocation measure into capital market and product market distortions⁷. We then estimate these two distortions separately for each firm within their corresponding industry (at the 2-digit level), as a function of the wedges relative to a frictionless economy and examine how these new bank relations affected them. We find that the new bank relations led to a decline in capital market distortions by 9-11%, but for only the big firms. Correspondingly, we find the opposite effect on product market distortions, neutralizing the positive effect on the capital market side.

Next, we use these measures of misallocation to define TFPR and total physical factor productivity (TFPQ). The former is a combination of all the marginal revenue productivities of inputs (capital and labour in our case) to production and can be termed as allocative efficiency or the between-firm productivity effect.⁸ On the other hand, we define TFPQ as the within-firm productivity.

We find a net zero effect on between-firm productivity or allocation of resources. This is because although the marginal productivity of capital declined for those firms that had credit relationships with the new banks, the marginal productivity of labour, on the other hand, increased. This resulted in the TFPR distribution across firms remaining unchanged. In other words, the re-weighting of the capital and product market wedges led to a null effect in the allocation of resources.

This increase in the marginal productivity of labour (and other inputs) for the big firms is a result of the increase in tangible and intangible investments, managerial inputs, and capital employed in production. New bank relations influenced big firms to resort to higher capital- and skill-intensive techniques of production which increased the physical productivity of those firms. This led to significant improvements in TFPQ for the big firms.⁹ Firms in the smaller quartiles were not impacted.

This increase in investments, and hence within-firm productivity, with no change in between-firm allocation points toward the fact that credit flow from new entrant banks may have been concentrated only towards certain firms and not across the firm size distribution. Petersen and Rajan (1995) show that in the case of the US, banks cherry-pick their clients because of lack of ‘soft information’¹⁰, among others.

Did the entry of new banks have an effect on overall output?

Lastly, we use our model to compute aggregate productivity, a combination of both physical productivity and allocative efficiency, which in turn is used to compute the aggregate potential gains in output. To find the contribution of entry of the new banks, we follow Bau and Matray (2023) and run a counterfactual exercise where we compute aggregate productivity without the new banking relationships. We plot our counterfactual estimates given by the black dashed line in Figure 4. We find that the new banks are responsible for at least 3-5% of the overall gain in manufacturing output, and these gains are realised when most of the new banking relationships are formed.

Figure 4. Aggregate potential gains in manufacturing output over time

Notes: (i) The blue (solid) line is the potential gains in output estimated using our model. The dashed black line is the potential gains without the effect of new banking relationships. (ii) The difference between the lines captures the effect of new banks.

Conclusion

Our study finds that the entry of private and foreign banks had no discernible impact on misallocation in the Indian manufacturing sector. These new banks cherry-picked larger firms and offered them easier access to credit. For these firms, cheaper credit translated into higher managerial compensation and capital deepening, contributing to within-firm gains in physical productivity. While capital market distortions appeared to decline, these gains were offset by persistent or rising distortions in product markets. Rather than scaling up, the beneficiary firms chose to enhance efficiency, leaving overall misallocation unchanged.

The persistence of product market distortions may be attributed to legacy policies, such as product reservations and rigid labour laws, which continue to impede firm expansion, even when credit is accessible. Although identifying the precise mechanisms is beyond the scope of this study, our findings underscore the need for coordinated reforms across sectors. Without addressing inter-sectoral synergies, piecemeal or unilateral reforms may risk undermining the broader gains from sector-specific liberalisation.

Notes:

A large proportion of firms in developing countries frequently report access to finance as one of the major impediments to their growth (Bloom et al. 2010).
Following Gormley (2010), we also check our findings at the district level. For the district level, we use a dataset compiled by the Reserve Bank of India (RBI) to track the opening of all the new branches in a district and compare the amount of borrowing by firms located in these districts to those where no new branches (of new private and/or foreign bank) were opened. Similar to our firm-level finding, we find the effect only for firms above the 75^th percentile of the size distribution.
Quartiles are defined according to the total assets of a firm. A firm whose total assets, during 1995-2007, are below the 25^th percentile of the total assets of the corresponding industry, belongs to the first quartile, and so on.
Our results are robust to several ‘selection’ issues related to the firm characteristics (for example, end-use category), differential trends, industry (for example, government support for certain industries), and regional unobservables (for example, local political support) of the sample that may not be representative of the broader population, and therefore may confound our estimates.
We define industry-level TFPR as the sales weighted average of firm-level TFPR. To construct our measure of misallocation, we decompose it into an unweighted average of firm-level TFPR and a ‘residual’ term as in Olley and Pakes (1996). The residual is positive if more productive firms have higher sales. This is expected if misallocation of resources is low in an industry, otherwise it is negative or zero.
NIC two-digit manufacturing sectors are defined according to National Industrial Classification (NIC) codes. NIC codes are a statistical standard for developing and maintaining a comparable database for various economic activities, developed with an intent to ascertain and analyse as to how each economic activity is contributing towards national income.
Capital market distortions capture variation in credit access across firms within an industry. This can be because of lower competition on the supply side, as we show in our research paper, or imperfect monitoring, etc. Product market distortions can arise from size-dependent policies, including but not limited to reservation of certain products for MSMEs, etc. Both these wedges are measured relative to labour-market wedges, since all of them cannot be separately identified.
In a world without misallocation, we would expect TFPR to be equated across firms within the same industry (and in the process, firms with the highest TFPR to have more resources and more sales, as discussed above).
This may have further neutralised the possible gains from the reallocation of resources among other firms by increasing the marginal productivities of both capital and labour proportionally for those big firms.
Soft information refers to unobservables, like managerial ability, which the banks resort to in order to evaluate the credit worthiness of a firm. This is based on long-term relationships between firms and banks. With more competition, arm’s-length lending takes precedence where credit worthiness is typically based on ‘hard information’ like revenue, assets, or past profits.