Borrower income and loan rates in the credit market of Lima

ABSTRACT

I analyse the effect of borrower income on loan rates in the credit market of Lima in 1840–65. I show that borrower income had a negative effect on interest rates. Borrower income influenced interest rates mostly through the impact on loan sizes: richer borrowers received larger loans and larger loans were associated with lower loan rates. The results are consistent with the influence of economies of scale on lending and differences in risk between large and small borrowers.

KEYWORDS:

• Credit
• moneylending
• loan rates
• firm size

JEL CODE:

• G21
• E43
• L25
• N26

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 Notary José de Selaya, 711 (f. 152), March 15, 1854.

2 Notary Felipe Orellana, 485 (f. 305), June 21, 1854

3 Empirical evidence for several economies suggests that smaller firms usually pay higher loan rates (Neuberger and Räthke-Döppner 2015; Rostamkalei and Freel 2016).

4 Information costs were probably higher in the past than nowadays. Only a few decades ago, for instance, lenders did not count with credit scoring reports or credit bureaus that provided information on borrower riskiness.

5 Studies for other countries have also relied on notarized records to analyse the functioning of early credit markets (Levy 2012; Hoffman et al. 2000).

6 People could have accumulated savings to invest. However, by facing lower interest rates, richer people had better chances to fund investment.

7 Information was obtained from the following notary books (protocolos): 65–70, 224–227, 478–501, 563–567, and 701–734. Obligations were one type of contracts certified by notaries.

8 I do not include sales on credit; they usually do not include information on loan rates.

9 I estimate the number of cash loans by assuming that the ratio of cash loans to the number of obligations is the same in the population of contracts notarized by all notaries from Lima and in the contracts notarized by the five notaries.

10 Tax reports were also used to complement information on the occupations of lenders and borrowers.

11 To determine the taxes for each class, The State first calculated the highest earnings in the first class and the lowest earnings in the fourth class, denoting those amounts as $M$ and $m$, respectively. For the first class, taxes were 4% of $M$. For the second class, taxes were 4% of $+{\displaystyle \frac{2}{3}\left(M-m\right)}$. For the third class, taxes were 4% of $m+{\displaystyle \frac{1}{3}\left(M-m\right)}$. For the fourth class, taxes were 4% of $m$. There were some exceptions to the 4% rule. In 1846, a decree established that first-class consignees had to pay 5% of their earnings (http://www.leyes.congreso.gob.pe/Documentos/LeyesXIX/1846021.pdf). In 1851, a decree declared that workers earning less than 200 pesos per year were exempted from paying income tax (http://www.leyes.congreso.gob.pe/Documentos/LeyesXIX/1851028.pdf). Further information on the calculation of taxes can be obtained from http://www.leyes.congreso.gob.pe/Documentos/LeyesXIX/1852073.pdf

12 I collected information on more than 3,000 loans. Of these, 42% include information on lender income and 32% include information on borrower income.

13 For instance, if the nominal monthly rate was $R_m$ percent per month, then the nominal annual rate (in percentage points) $R_a$ was calculated as $100\left[{\left(1+R_m/100\right)}^{12}-1\right]$. Then INT (the real annual rate in percentage points) was calculated as $R_a-\pi$, where $\pi$ is the annual inflation in percentage points.

14 For this paragraph, the average lender (borrower) income was obtained using the taxable income in levels. So the figure is not equal to the exponential of the average of LENDER_INC (BORROWER_INC). The exponential of the average of LENDER_INC is 1,674 pesos, whereas the exponential of the average of BORROWER_INC is 789 pesos.

15 In addition, richer lenders made larger loans.

16 I added one to the number of loans, because for some borrowers this was the first time they borrowed from the same lender.

17 I rely on all loans secured by these notaries, even for those for which there is no information on lender and borrower income.

18 I control for several variables. I include gender dummies for lenders and borrowers (LEND_FEM and BOR_FEM, respectively). I also include dummies that are equal to one if the lender/borrower was a company (LEND_C and BOR_C, respectively). Using information on occupations, I constructed dummies for lenders’ and borrowers’ occupations. In particular, the following occupations were considered: merchants, agriculturists, rentiers, clergy, professionals and artisans. I include macroeconomic variables, and I include the inflation rate in percentage points (INF) to control for nominal rigidities. In the presence of nominal rigidities, one would expect inflation to have a negative effect on the real loan rate. As indicated, inflation rates come from Gootenberg (1990). I also include the growth rate of the gross domestic product in percentage points (GDPG) and fiscal revenues as percentage of GDP (REV) to measure other macroeconomic conditions. Information on GDP growth and fiscal revenues comes from Seminario (2015). Finally, I include a war dummy (WAR) to control for the effect of wars on loan rates. Information on wars comes from Zegarra (2016).

19 Some differences in risk may be not correlated with loan sizes.

20 In Equation (2´), loan size does not depend on maturity. In Equation (3´), maturity does not depend on loan size. However, they still both depend on loan rates.

21 With the system of Equations (1), (2´), (3´), (4), and (5), one can estimate the marginal effect of the borrower income on loan rates. As discussed, borrower income may affect loan rates for several reasons. A first possible reason is that richer borrowers demand larger loans. In the presence of fixed operation costs, one would expect lower rates for larger loans, and thus for richer borrowers. A second reason is that there were differences in the risk of default due to the size of capital: if poorer borrowers were more vulnerable to negative shocks, one would expect loan rates to be higher for poorer borrowers. A third reason is that richer borrowers demanded longer loans: if maturity influenced loan rates, then firm size would indirectly affect loan rates. A fourth reason is that there may be differences in information asymmetries between rich and poor borrowers: richer borrowers may be better known by lenders. A fifth reason is that there may have been differences in negotiation power: if richer borrowers had more negotiation power, they could have received credit at lower rates.

22 In this section, I assume that loan rates do not affect loan size and maturity (${\gamma }_R={\phi }_R=0$). Ordinary least squares (OLS) estimates may yield consistent and efficient estimators. It is possible, however, that loan rates influenced loan sizes and maturity. In the presence of mutual causality, endogeneity will arise and OLS will yield inconsistent and biased estimates. Instrumental variables could be used to obtain consistent and unbiased estimates. The appendix deals with the possible endogeneity using instrumental variables.

23 The results also show that the coefficient of lender income is negative and significant. This result implies that richer lenders charged lower loan rates.

24 It is, therefore, possible that ${\phi }_B$, ${\alpha }_B$, and ${\rho }_B$ are different from zero.

25 This expression is derived from Equation (8) in this section, under the assumption that loan rates do not affect loan size and maturity.

26 Also, the findings suggest that differences in risk not captured by loan sizes, maturity, information asymmetries, or other regressors did not have a significant effect on loan rates.

27 For instance, practically any adult could lend money.

28 For instance, borrowers needed to know somebody with extra savings, who was willing to lend to them. Further research is needed to understand the personal sources of financing. Much of what is known is limited to notarized loans, i.e. impersonal credit markets.

29 In Table 4, it is assumed that loan rates did not influence loan size and maturity. However, higher interest rates could reduce the size of the loans: facing higher loan rates, borrowers could have selected smaller loans. Loan rates could have also affected maturities: it is possible that borrowers would have opted to demand shorter loans if rates were higher.

30 I assume that relationship lending and negotiation power are not endogenous to loan rates, loan size, or maturity.

31 It should be noted that the female borrower dummy has a highly significant effect on loan size, but not on loan rate or maturity. Therefore, the female borrower dummy could be used as an instrument for loan size. Another possible instrument for loan size is the borrower-company dummy (BOR_C): this variable is more correlated with loan size (p value = .2) than with loan rate (p value = .9) and maturity (p value = .9). On the other hand, the collateral dummy has a significant effect on maturity, but not on loan rate or loan size. The collateral dummy could then be used as instrument for maturity. Meanwhile, the war dummy has a significant effect on loan rate, but not on loan size or maturity. In addition, the growth of GDP has significant effect on loan rates, but not on loan sizes or maturities. Therefore, the war dummy and the growth of GDP could be used as instruments for loan rates.

32 Table A2 reports the first-stage results.

33 A potential problem with the results refers to the selection of instruments. In particular, it might be argued that instruments for loan rates are not valid. For instance, although the OLS estimates in Table 4 show that war dummy does not have a significant effect on loan size and maturity, wars could have affected the capacity of lenders or willingness of lenders to make large and long loans. A similar argument could be made about the growth of GDP.

34 I rely on the 2SLS estimates from the equations of loan rates, loan size and maturity and on the OLS estimates from the equations of relationship lending and negotiation power. The direct, indirect and total effects of borrower income on loan rates are calculated using the formulas (6)–(8) from section 6.

35 Both the direct and indirect effects of borrower income on loan rates are negative and non-significant. However, the indirect effect of borrower income is far larger in absolute value than the direct effect.