Abstract
Uruguay: Selected Issues
Bank Lending and Competition in the Banking Sector1
The previous paper showed the benefits of relaxing financial frictions in terms of growth and reduction of inequality. This paper aims at investigating the sources of those financial constraints in more detail, especially the intermediation costs and low depth of the credit market. To do so, the first section documents some stylized facts about bank lending in Uruguay. The analysis then relates these findings to the competitive structure of the Uruguayan banking system (second section), and presents the results of a structural model of supply and demand for credit in Uruguay (third section). The fourth and last section concludes.
A. Stylized Facts
1. Credit to the private sector in Uruguay is low relative to regional and income group peers, as well as other benchmarks, suggesting a significant scope for financial deepening. At just 25 percent of GDP in 2014, Uruguay’s private credit intermediation ratio is among the lowest in the Latin American region, and falls well short of other countries at the same income level as well as a statistical benchmark estimated by the World Bank based on fundamentals.
Credit to Private Sector
(In percent of GDP)
Citation: IMF Staff Country Reports 2016, 063; 10.5089/9781498384223.002.A002
Source: World Bank, FinStats; and Fund staff calculations.2. The legacy of Uruguay’s 2002 crisis may help explain certain financial system dynamics that have contributed to low credit. Private sector credit peaked just before the crisis and then collapsed and has yet to recover. The massive deposit runs during the crisis, together with the failure of certain large foreign banks, and weak prudential regulations at that time, have certainly played a role in shaping the current policy and supervisory framework. But the crisis may have also had a direct and lingering impact on lender behavior in Uruguay, particularly through its impact on the structure of the banking system.
Credit to the Private Sector
(In percent of GDP)
Citation: IMF Staff Country Reports 2016, 063; 10.5089/9781498384223.002.A002
Sources: IMF, World Economic Outlook; International Financial Statistics; and Fund staff calculations.3. Furthermore, the high degree of financial dollarization in Uruguay may affect the supply and demand side dynamics in the credit market. As of mid-2015, approximately 80 percent of total deposits and 60 percent of total loans in Uruguay were denominated in U.S. dollars. The country’s history of high inflation and currency devaluations has led people to have a preference for holding their savings in U.S. dollars. Since the bulk of banks’ liabilities are in dollars, they limit their peso lending (the majority of which is to households) to avoid currency mismatches on their balance sheets. Dollar credit (mostly to firms) is also restrained, by high reserve requirements on foreign currency deposits that lead to higher lending rates and a preference for liquidity on the supply side (banks prefer investing in liquid dollar-denominated securities rather than lending domestically), and the existence of (potentially less costly) alternatives to bank lending, such as direct lending, bond markets and FDI, on the demand side.
Dollarization
(In percent of total deposits)
Citation: IMF Staff Country Reports 2016, 063; 10.5089/9781498384223.002.A002
Sources: Banco Central del Uruguay; and Fund staff calculations.4. The Uruguayan credit market is highly segmented (Figure 1). In 2014, 60 percent of total credit went to firms, 35 percent to households, and 5 percent to the public sector. Of the credit extended to firms, 87 percent was denominated in U.S. dollars, whereas for households, only 4 percent of credit was U.S. dollar denominated, and 35 percent in the case of the public sector. Consequently, 55 percent of total credit in 2014 was U.S. dollar denominated. Looking at the sectoral composition of credit by currency, half of U.S. dollar credit in 2014 went to the agricultural and manufacturing sectors, with another one-fifth conferred to the commercial sector. On the peso side, three-quarters of credit went to the household sector, with consumer loans accounting for almost half, and mortgages accounting for about 40 percent; the remainder comprised loans extended by Administradoras de Creditos, non-deposit taking lenders, specializing in loans to lower income households.
Uruguay: Segmentation of the Credit Market
Citation: IMF Staff Country Reports 2016, 063; 10.5089/9781498384223.002.A002
Sources: Banco Central del Uruguay and Fund staff calculations.5. The segmentation permeates the banking sector, which exhibits a high degree of concentration, particularly with regard to the peso deposit and credit markets. There are two public banks in Uruguay and ten (soon to be nine) private banks.2 The large public bank, Banco de la Republica Oriental de Uruguay (BROU), holds 40 percent of total banking assets, and the top four banks hold three-quarters of total assets. There is a high degree of heterogeneity and segmentation between BROU and the private banks. BROU enjoys a monopoly on public accounts by law, and until recently held most public employees’ checking accounts. The bank has a 48 percent market share in the peso deposit market. Before the 2014 Financial Inclusion Law, BROU was also the only bank to provide payroll loans, which contributed to a 46 percent market share for BROU in the high spread peso lending market.3 The foreign banks, on the other hand, have highly dollarized deposit bases and mostly engage in U.S. dollar lending to commercial and higher income segments, with a higher degree of competition and more compressed spreads than seen in the peso market. Indeed, the Herfindahl-Hirschman index (HHI) for the peso credit market yields a concentration level of 0.26, while the U.S. dollar credit market is slightly less concentrated at 0.19.4
Bank Share of Credit in Uruguay Lending Markets
(In percent)
Citation: IMF Staff Country Reports 2016, 063; 10.5089/9781498384223.002.A002
Sources: Banco Central del Uruguay and Fund staff calculations.6. Partly owing to the high degree of concentration, bank credit in Uruguay is expensive. In 2014, the spread between average lending and deposit rates in pesos was 15¼ percent, at least 7 percentage points higher than the median spread in the LA5. On the other hand, the spread between average lending and deposit rates in the U.S. dollar market was only 4¼ percent. The high average lending rates in the peso market could be a factor that crowds out some households and firms from the market. The next section further investigates the link between concentration in the banking sector and credit provision.
Lending–Deposit Spreads
(In percent)
Citation: IMF Staff Country Reports 2016, 063; 10.5089/9781498384223.002.A002
Sources: Banco Central del Uruguay; and Fund staff calculations.B. The Effect of Banking Concentration on Lending
7. In theory, banking competition can have two opposite effects on bank lending (Claessens and Laeven, 2005). According to the market power hypothesis, competition reduces the cost of finance (by reducing banks’ monopolistic profit margins) and increases the availability of credit. In the presence of information asymmetries and agency costs (i.e., costly monitoring of borrowers), competition may however reduce credit provision by making it more difficult for banks to internalize the returns from investing in lending.
8. In practice, we find a significant negative relationship across countries between the degree of concentration of the banking sector and the ratio of private sector credit to GDP (Table 1). In order to gauge the relationship between private sector credit and banking sector concentration, a model is estimated for a panel of 137 countries over the period 1986–2011. The ratio of top 5 banks’ assets to total banking sector assets for each country is used as a proxy for banking sector concentration. In addition, the model controls for the level of economic development (proxied by the log of GDP per capita in PPP dollars), the quality of the legal and institutional framework (measured by the KKM Governance Index Rule of Law indicator),5 and the level of government borrowing (using the log of public debt-to-GDP). Each specification regresses the change in private credit to GDP on bank concentration, the control variables, and country fixed effects. The results show a statistically significant negative relationship between banking concentration and private credit growth, with a 1 percentage point increase in top 5 bank assets as a ratio of total bank assets associated with a drop in private credit-to-GDP of between 1 and 5 percent. These results are in line with other empirical studies that found a negative relationship between banking sector concentration and credit provision.6
Credit to the private sector and bank concentration
The OLS regressions are ordinary least squares regressions with standard errors adjusted for clustering at the country level for a panel of 137 countries from 1986-2011. Selected country and/or regional dummies are included.
The FE regression estimates country fixed effects for all countries in the sample, but only the LA6 results are reported in this table.
Top 5 banks assets as a share of total banking assets.
KKM Governance Index Rule of Law indicator.
Demeaned estimates: fixed effect estimates minus a sample average of fixed effects. Robust standard errors are in italics. *** p<0.01, ** p<0.05, * p<0.1
Credit to the private sector and bank concentration
| Log of private credit/GDP | |||
|---|---|---|---|
| OLS (1) 1/ | OLS (2) 1/ | FE 2/ | |
| Log of GDP per capita (PPP) | 0.20*** | 0.18*** | 0.67*** |
| 0.06 | 0.06 | 0.12 | |
| Bank concentration 3/ | −0.01*** | −0.01*** | −0.05*** |
| 0.00 | 0.00 | 0.00 | |
| KKM Rule of Law 4/ | 0.54*** | 0.56*** | 0.43*** |
| 0.06 | 0.07 | 0.10 | |
| Log of Government debt/GDP | 0.12** | 0.15** | 0.03 |
| 0.06 | 0.06 | 0.05 | |
| LA6 dummy 5/ | −0.18*** | ||
| 0.63 | |||
| Non-LA6 dummy 5/ | 0.01*** | 0.01*** | |
| 0.61 | 0.62 | ||
| Brazil dummy 5/ | 0.04*** | −0.02*** | |
| 0.64 | 1.12 | ||
| Chile dummy 5/ | 0.26*** | 0.02*** | |
| 0.58 | 1.13 | ||
| Colombia dummy 5/ | 0.15*** | 0.10*** | |
| 0.66 | 1.07 | ||
| Mexico dummy 5/ | −0.58*** | −0.82*** | |
| 0.67 | 1.13 | ||
| Peru dummy 5/ | 0.03*** | 0.05*** | |
| 0.65 | 1.05 | ||
| Uruguay dummy 5/ | −0.73*** | −0.70*** | |
| 0.61 | 1.13 | ||
| Observations | 1,271 | 1,271 | 1,271 |
| R-squared | 0.824 | 0.836 | 0.975 |
The OLS regressions are ordinary least squares regressions with standard errors adjusted for clustering at the country level for a panel of 137 countries from 1986-2011. Selected country and/or regional dummies are included.
The FE regression estimates country fixed effects for all countries in the sample, but only the LA6 results are reported in this table.
Top 5 banks assets as a share of total banking assets.
KKM Governance Index Rule of Law indicator.
Demeaned estimates: fixed effect estimates minus a sample average of fixed effects. Robust standard errors are in italics. *** p<0.01, ** p<0.05, * p<0.1
Credit to the private sector and bank concentration
| Log of private credit/GDP | |||
|---|---|---|---|
| OLS (1) 1/ | OLS (2) 1/ | FE 2/ | |
| Log of GDP per capita (PPP) | 0.20*** | 0.18*** | 0.67*** |
| 0.06 | 0.06 | 0.12 | |
| Bank concentration 3/ | −0.01*** | −0.01*** | −0.05*** |
| 0.00 | 0.00 | 0.00 | |
| KKM Rule of Law 4/ | 0.54*** | 0.56*** | 0.43*** |
| 0.06 | 0.07 | 0.10 | |
| Log of Government debt/GDP | 0.12** | 0.15** | 0.03 |
| 0.06 | 0.06 | 0.05 | |
| LA6 dummy 5/ | −0.18*** | ||
| 0.63 | |||
| Non-LA6 dummy 5/ | 0.01*** | 0.01*** | |
| 0.61 | 0.62 | ||
| Brazil dummy 5/ | 0.04*** | −0.02*** | |
| 0.64 | 1.12 | ||
| Chile dummy 5/ | 0.26*** | 0.02*** | |
| 0.58 | 1.13 | ||
| Colombia dummy 5/ | 0.15*** | 0.10*** | |
| 0.66 | 1.07 | ||
| Mexico dummy 5/ | −0.58*** | −0.82*** | |
| 0.67 | 1.13 | ||
| Peru dummy 5/ | 0.03*** | 0.05*** | |
| 0.65 | 1.05 | ||
| Uruguay dummy 5/ | −0.73*** | −0.70*** | |
| 0.61 | 1.13 | ||
| Observations | 1,271 | 1,271 | 1,271 |
| R-squared | 0.824 | 0.836 | 0.975 |
The OLS regressions are ordinary least squares regressions with standard errors adjusted for clustering at the country level for a panel of 137 countries from 1986-2011. Selected country and/or regional dummies are included.
The FE regression estimates country fixed effects for all countries in the sample, but only the LA6 results are reported in this table.
Top 5 banks assets as a share of total banking assets.
KKM Governance Index Rule of Law indicator.
Demeaned estimates: fixed effect estimates minus a sample average of fixed effects. Robust standard errors are in italics. *** p<0.01, ** p<0.05, * p<0.1
C. A Supply or Demand Story?
9. The cross-country analysis of the previous section can be deepened for Uruguay by distinguishing between bank credit in pesos and in U.S. dollars. This section looks for structural supply-side influences on credit growth by estimating a two-equation structural model of bank credit, separately for credit in peso and credit in dollar.
10. The model consists of a supply and a demand equation. Both the supply and demand of bank loans are a function of the lending rate and other variables.7 Banks will supply more loans if the interest rate is higher and borrowers will demand fewer loans if the rate is higher, so that supply (demand) depends positively (negatively) on the lending rate. The lending interest rate adjusts to “clear the market,” that is, to equate demand and supply. The main challenge for the estimation is to find variables that allow for the identification of the two curves (the so-called “shifters”). These shifters need to move either supply or demand without affecting the other. Nominal credit supply is thus assumed to negatively depend on the interest rate paid by banks on deposits (a measure of bank funding cost), on the loans-to-deposits ratio (a measure of the availability of funds to lend) and on the ratio of overdue loans over total loans (a measure of credit risk for banks). Nominal credit demand is expected to depend positively on business confidence (a proxy for future economic conditions and investment profitability) and negatively on unemployment.
11. The estimation of the model addresses the endogeneity of the lending rate by using a two-stage-least-square procedure. The exogenous variables in the system are used as instruments for the lending rate.8 The sample period covers January 2002–August 2015. The results of the estimation are presented below.
Structural Determinants of the Supply and Demand of Bank Lending in Domestic and Foreign Currencies
Structural Determinants of the Supply and Demand of Bank Lending in Domestic and Foreign Currencies
| Expected Signs | Bank Credit in Domestic Currency |
Bank Credit in Foreign Currency |
|
|---|---|---|---|
| Supply Equation | |||
| Lending Rate | + | 11082.237 | 4139.303 |
| Constant | +/− | 2370.824*** | −131.179 |
| Supply Shifters | |||
| Deposit Rate | − | −13678.986 | −5584.435 |
| Lagged Loan-to-Deposit Ratio | − | −1800.356 | 5.067 |
| Lagged Overdue Loans/Total Loans Ratio | − | −5538.635** | −699.909 |
| Demand Equation | |||
| Lending rate | − | −1844.774** | −3281.465*** |
| Constant | +/− | 1582.279*** | 236.612*** |
| Demand Shifters | |||
| Business Confidence Indicator | + | −1225.944 | −103.145* |
| Change in Unemployment | − | −1030.293*** | −18.096 |
| Number of Observations | 163 | 163 | |
Structural Determinants of the Supply and Demand of Bank Lending in Domestic and Foreign Currencies
| Expected Signs | Bank Credit in Domestic Currency |
Bank Credit in Foreign Currency |
|
|---|---|---|---|
| Supply Equation | |||
| Lending Rate | + | 11082.237 | 4139.303 |
| Constant | +/− | 2370.824*** | −131.179 |
| Supply Shifters | |||
| Deposit Rate | − | −13678.986 | −5584.435 |
| Lagged Loan-to-Deposit Ratio | − | −1800.356 | 5.067 |
| Lagged Overdue Loans/Total Loans Ratio | − | −5538.635** | −699.909 |
| Demand Equation | |||
| Lending rate | − | −1844.774** | −3281.465*** |
| Constant | +/− | 1582.279*** | 236.612*** |
| Demand Shifters | |||
| Business Confidence Indicator | + | −1225.944 | −103.145* |
| Change in Unemployment | − | −1030.293*** | −18.096 |
| Number of Observations | 163 | 163 | |
12. The estimation allows the identification of the supply and demand of credit in peso. Both the supply and demand equations include statistically significant “shifters” that affect one but not the other. The supply curve is identified by the ratio of overdue loans to total loans while the change in unemployment is a significant demand-shifter. On the supply side, the loan-to deposit ratio came out as significant at the 10 percent level in one specification but the result was not robust to specification changes (results not reported).
13. The supply curve for bank credit in U.S. dollars is not identified. This (non-)result is robust across many specifications including with alternative supply shifters.9 This can be related to the features of the credit market in dollar: Banks collect a lot of deposits in dollar, so dollar funding is not a constraint. Besides, as documented before, the market for credit in U.S. dollar is much more competitive than the peso market, so any increase in the demand for dollar loans can be accommodated without much change in the lending rate.
14. The negative sign on business confidence in the demand equation for bank lending in dollar may be explained by the reduction of outside financing options for firms in bad times. This negative sign is a priori counterintuitive. A possible explanation is that in good times, when business confidence is high, firms have more financing options, in the form of higher self-generated cash flows or easier access to capital markets, and less need for bank credit to finance investment.
15. The plots of the estimated demand and supply curves as a function of the lending rate in the two markets show how the curves shifted over the last ten years. The curves are constructed using the estimated coefficients reported in the table above and the means of the explanatory variables over the two years.10 Because of the way the curves are constructed, the shifts represent changes in the explanatory variables and not changes in the relationships between the variables. The light-shade of blue indicates that the slope is not statistically significant. The credit supply curve in U.S. dollar is not drawn as it is not identified.
Fitted Supply and Demand Curves for Bank Credit in Peso
Citation: IMF Staff Country Reports 2016, 063; 10.5089/9781498384223.002.A002
Sources: Staff estimates.Fitted Demand Curve for Bank Credit in Foreign Currency
Citation: IMF Staff Country Reports 2016, 063; 10.5089/9781498384223.002.A002
16. The absence of a significant relationship between credit supply in peso and the lending rate can be interpreted as the result of the quasi-monopolistic nature of the market, in which BROU controls nearly half of the market. In a monopolistic market, there is indeed no supply curve as the monopoly as the monopoly adjusts the quantities depending on its costs to maximize its profits. The story for bank lending in dollar is different. The dollar lending market is demand-determined, as supply can accommodate whatever level of demand, while the lending interest rate is determined by the international interest rate in U.S. dollar plus a risk premium.
D. Policy Conclusions
17. The above analysis suggests that policies can increase bank lending by addressing supply-side constraints, in particular in the peso market. In particular, the 2014 Financial Inclusion Law can help stimulate credit in pesos through two channels:
First, the law helps to increase the peso funding of private banks (one of the supply shifters in the above credit model) and may thus facilitate more peso lending. The increase in peso funding for private banks may come from the generalization of payroll deposit accounts and also from the new limitations on cash transactions, which should support the use of bank accounts and electronic means of payments.
Second, by instituting the right of employees to choose the bank where they want to have their salaries deposited (a selection that was previously made by employers), the law will likely increase competition, which in turn should lead to efficiency gains and a reduction in spreads. Both would have a positive effect on the depth of the credit market as seen in the previous paper.
18. The government could further encourage competition in the peso market by expanding opportunities for private banks to compete for public accounts. More competition in the provision of banking services to the public sector could allow for better borrowing terms and a reduction in financial costs for the government.
19. Higher competition in the peso lending market would enhance the credit channel of monetary policy. The lack of a significant relationship between bank lending in pesos and the lending rate limits the effectiveness of monetary policy, since changes in the policy rate have little effect on banks’ lending behavior and therefore on economic activity and inflation. Increased competition in the peso credit market could help foster a stronger link between lending rates and bank lending, and thereby make monetary policy more effective.
References
Claessens, Stijn and Luc Laeven. 2005. “Financial Dependence, Banking Sector Competition, and Economic Growth,” Journal of the European Economic Association, 3(1), pages 179–207.
IMF. 2013. “Assessing Policies to Revive Credit Markets.” Global Financial Stability Report, Chapter 2, October.
Prepared by Frederic Lambert and Diva Singh (WHD).
In December 2014, Scotiabank signed an agreement to buy Discount Bank Latin America (owned by Israel’s Discount Bank), Uruguay’s ninth largest bank, for US$65 million. The deal is expected to be completed in 2015.
Payroll loans are uncollateralized loans whose installments are directly debited from workers’ pay.
From a scale of 0 to 1, 1 being a perfect monopoly.
Kaufmann, Kraay, and Mastruzzi (2004), included in the World Bank Worldwide Governance Indicators.
Jimenez, Lopez and Saurina (2013), Fisman and Raturi (2004).
The analysis uses changes in loan stocks or net transaction flows as a proxy for new loans, although repayments of previously granted loans should not in theory be deducted from new loans. A detailed description of the model can be found in Chapter 2 of the October 2013 Global Financial Stability Report.
Other endogeneity issues complicate the proper identification of the model. Most variables in the analysis are at some level more or less jointly determined. For instance, changes in business confidence may be affected by current changes in bank lending standards and credit provision. Similarly, changes in the ratio of non-performing loans will depend on the volume of new loans during the period. To alleviate the resulting endogeneity, some of the regressors are lagged by one period.
Among the other possible shifters is the yield on U.S. Treasury securities, as banks can arbitrage between domestic lending in dollar and investment of the collected dollar deposits in dollar-denominated securities abroad. The results were not significant.
The slopes of both curves are assumed to have remained the same between 2006 and 2015 (i.e., the elasticity of supply and demand to interest rates has not changed over time).
