Article

Estimating Models of Financial Market Behavior During Periods of Extensive Structural Reform: The Experience of Chile

Author(s):
International Monetary Fund. Research Dept.
Published Date:
January 1983
Share
  • ShareShare
Show Summary Details

In designing a financial program, a sample of data taken from the prestabilization program period is often used to obtain empirical estimates of the parameters of a variety of behavioral relationships. Implicit in the use of these estimated empirical relationships is the assumption that, at least in the short run, the financial program will affect the performance of a given economic structure but will not fundamentally alter the structure itself. While this assumption may be quite reasonable for many economies, there is the problem of how to estimate these behavioral relationships for countries that are undergoing extensive structural change over an extended period as a result of sweeping trade, fiscal, and financial market reforms.

For the period 1973-81, Chile provides an example of an economy that underwent extensive trade, fiscal, and financial policy changes designed to open the economy to international influences. The trade policies encompassed the replacement of a system of quotas and high tariffs with a uniform tariff of 10 percent an exchange rate policy that evolved from a crawling peg complemented by occasional discrete adjustments, to a preannounced crawl, and, finally, to a fixed exchange rate. In financial markets, the program included the removal of interest rate ceilings, substantial reductions in required reserves in financial institutions, the payment of interest on bank reserves, sustained reductions in the Central Bank’s issuance of base money and subsidized lending programs, and opening up the financial system to increased domestic and international competition. Finally, fiscal reforms resulted in expenditures and tax changes that created a substantial fiscal surplus. Table 1 illustrates the sequencing and scope of some of the reforms that took place during 1975-80.

Table 1.Chile: Selective Listing of Reform Measures, 1975–801
Trade ReformFinancial System ReformsFiscal Policy
1974–75
Prior to the trade reform, there was an extensive system of import controls that included (1) a compulsory waiting period of 90 to 120 days for delivery of foreign exchange by central banks, (2) a list of some 2,400 products subject to a 10,000 percent deposit requirement, and (3) an unweighted average import duty of 67 percent at the end of 1974.

In August 1975 a new tariff program designed to establish (by the beginning of 1978) a tariff schedule with no rate higher than 35 percent was put in place. The program provided for a linear tariff reduction schedule composed of five stages with six months between each modification. The first modification was made in February 1976.
Prior to the start of the financial reform in 1974, the principal financial institutions consisted of the Central Bank, the Banco del Estado (a state bank), the commercial banks, and a system of savings and loan institutions (SINAP). This system was characterized by interest rate ceilings, high required reserve ratios, and limited competition between financial institutions.

In May 1974 short-term ceiling interest rates were increased, full inflation adjustment of principal was authorized on credits with maturities of at least one year, and investment companies (financieras) were allowed to operate beside commercial banks at interest rates up to 50 percent higher than maximum bank rates. In November 1974 interest rates paid by banks on time deposits were freed, and proceeds of such deposits could be used to purchase commercial paper at free rates.
To improve the consolidated government budget position, the authorities undertook a series of measures in April-May 1975 including (1) a 10 percent surcharge on income and property taxes, (2) the elimination of most exemptions to the value-added tax (VAT) and increases in excise taxes, (3) a 15 percent reduction in real terms in local currency expenditure and a 25 percent real cut in foreign currency expenditure other than debt payments, (4) a hiring freeze, (5) greater control over public enterprises, and (6) accelerated sales of assets of official entities. Despite these measures, there was a consolidated government deficit equal to about 2 percent of gross domestic product (GDP).
In 1975 there was rapid growth of the financieras but serious difficulties with the SINAP system, as it lost deposits to the financieras. Throughout 1975 various actions were taken to stem the outflow of deposits from SINAP, including limitations on deposit withdrawals and the extension of Central Bank regulatory policies to include the financieras. To improve the position of banks vis-à-vis the financieras and to further the liberalization process, bank lending rates were freed (in April 1975) and an 8 percent reserve requirement was established on financiera liabilities.
1976
The schedule of tariff reductions was speeded up. Revised schedules were issued for changes to be made in December 1976, January 1977, and April 1977. The unweighted average tariff rate fell to 49 percent at the end of 1975 and to 27 percent at the end of 1976. A compulsory waiting period for delivery of foreign exchange was gradually reduced and then eliminated. A list of products subject to an advance deposit of 10,000 percent was reduced basically to imports of used motor vehicles.In May 1976 the Central Bank began to pay interest on bank reserves against time deposits. Initially this rate was pegged to the treasury bill rate. After November 1976 the rate was linked to the rate offered by banks on their time deposits. At first, the interest rate on reserves was equal to 70 percent of the time deposit rate during the same month. Subsequently, the proportion was raised to 90 percent, and in June 1977 to 100 percent. During 1976 the authorities began to lower reserve ratios in the financial system on a gradual but continuous basis. Between mid-1976 and mid-1979, for example, these changes lowered the reserve ratio on sight deposits from 85 percent to 52 percent and on 30-day to 89-day time deposits from 55 percent to 16 percent.The consolidated government accounts moved into a surplus equal to approximately 4 percent of GDP.
1977
Tariff reductions in 1977 brought the average tariff to 16 percent by the end of the year. The final stage of the tariff reform consisted of monthly tariff reductions between December 1977 and June 1979. At the end there would be a uniform tariff of 10 percent on all imports (except automobile items).In November 1977 the payment of interest on bank reserves was extended to 50 percent of excess reserves. The reduction in reserve ratios continued throughout the year.In January 1977 steps were taken to extend the use of indirect taxes and to reduce the role of direct taxes. The changes in indirect taxes included the elimination of exemptions from the 20 percent VAT and raising the VAT from 8 percent to 20 percent on a set of goods. Changes in direct taxes encompassed reductions in income tax rates and increases in basic nontaxable income allowances (especially on lower-income and middle-income groups) and reduction of certain direct business taxes. The consolidated government budget surplus was about 1 percent of GDP.
1978
Tariff reductions were continued.The above-mentioned reforms were continued.The consolidated government budget surplus was roughly 2 percent of GDP.
1979
Tariff reform (except for automobile products) was completed and the remaining prior import deposits of 10,000 percent were eliminated in June 1979.Reserve ratios reached 42 percent on sight deposits and 8 percent on time deposits by the end of 1979. As reserve ratios fell, the authorities ended the payment of interest on reserves. In April 1979 the interest rate on reserves fell to 50 percent of the time deposit rate, and all such payments ceased at the end of 1979.Special emphasis was placed on measures designed to reduce tax evasion. The consolidated government budget surplus was equal to about 5 percent of GDP.
1980
The reserve ratio on time deposits was lowered to 4 percent, and the incremental reserve ratio on sight deposits was lowered in steps from 21 percent to 10 percent.In August 1980 there were further increases in the basic nontaxable allowances, reductions in personal income tax rates, and a revenue-sharing program with the local governments. The consolidated government budget surplus was about 2 percent of GDP.
Exchange Rate PolicyControls on Capital InflowsOther Reforms and Measures
1975
Throughout the year there were small frequent changes in the exchange rate, with a sharp real depreciation of the exchange rate early in 1975.The most important capital inflows were subject to the conditions imposed by Article 14 of Decree 1272, which stipulated that capital brought into the country in the form of foreign exchange and registered with the Central Bank could be repatriated as long as the capital remained in the country for no less than 24 months from the date of registration. The 24-month period was established in August 1976; prior to that, it had been only 6 months.This was a period of gradual but continued liberalization of the price system. Prices were classified into three groups: (a) free prices, (b) freely determined prices subject to cost determination, and (3) controlled prices. At the end of 1975 there were less than 30 items in the controlled category, including basic foodstuffs, public utility tariffs, cigarettes, and petroleum derivatives. Price increases for category (2) goods were to be accompanied by the filing of evidence of cost increases. Public enterprise tariffs were increased sharply in April and May to reflect rapid cost increases that had not been reflected fully in utility prices. In September the authorities adopted a new wage-indexing formula based on the rate of increase in the consumer price index during the last two months plus an adjustment for the expected price rise in the month of the wage adjustment.
1976
Small adjustments of the exchange rate continued. In addition, on June 30, 1976 there was an 11 percent revaluation of the exchange rate. Until July 1976 the small adjustments in the exchange rate were announced as they occurred, but after that date the changes were announced approximately one month ahead.There was further liberalization of the price system.
1977
In March 1977 there was an 11 percent discrete appreciation of the peso, followed again by a period of small frequent changes in the exchange rate.The number of products subject to price controls was reduced to 15.
1978
In February 1978 the authorities published a schedule of daily exchange rates for the peso in terms of the U.S. dollar for the rest of the year. In December 1978 a new schedule was published for 1979.In January, commercial banks were authorized to contract external credits under Article 14 for relending in domestic currency but with the loan principal indexed to the U.S. dollar/peso exchange rate. Such loans could equal up to 25 percent of the paid-up capital and reserves of each bank. In addition, no more loans than the equivalent of 5 percent of capital and reserves could be converted into domestic currency in any one month. Finally, the global limit on bank indebtedness to foreign banks was raised from 150 percent to 160 percent of bank capital and reserves. If this limit was exceeded, the entire stock of such liabilities would be subject to an 8 percent reserve requirement.

In April, banks were authorized to surpass both the limit on Article 14 borrowings and the global limit on foreign indebtedness by 20 percent, provided that this excess borrowing represented foreign credits having an average maturity of at least 36 months.

In December, banks and financial institutions could use Article 14 credits to fund domestic lending in amounts equal to 5 percent of capital and reserves or US$2 million, whichever was larger. Also, the limit on Article 14 borrowings by such institutions was raised from 25 percent to 60 percent of capital and reserves, provided that the excess borrowing represented credits having an average maturity of at least 36 months. The limit on total external indebtedness of banks to other banks was raised from 160 percent to 180 percent of capital and reserves. This limit could be raised to 215 percent if excess borrowing had an average maturity of at least 36 months.
1979
On June 30, 1979 the peso depreciated to 39 Chilean pesos to the U.S. dollar; it was announced that this level would be left in effect until the end of February 1980.In April, the limit on total foreign indebtedness of banks was raised to 225 percent, and the limit on Article 14 borrowing went up from 60 percent to 70 percent. At the same time, financial institutions that obtained credits under Article 14 were required to deposit at the Central Bank the equivalent of 25 percent of loans with a maturity of 24–36 months and 15 percent for longer maturity. In May, the 15 percent deposit requirement was made applicable to loans with maturities of three to four years with 10 percent on loans of four to five and a half years and 0 percent for longer maturity.

In June 1979 the monthly limit on the rate of utilization of foreign loans was lowered to US$1 million or 5 percent of capital and reserves, whichever was larger. The 225 percent limit on total external debt was abolished.
1980
The ceiling on monthly conversions of Article 14 borrowings was raised to US$2 million or 5 percent of capital and reserves in January. This restriction was eliminated in April. The deposit requirement on loans of 24-48 months was set at 15 percent.

The objective of this analysis is to empirically estimate the response of inflation, the balance of payments, and financial markets to the opening of the economy and other reform elements within a framework that is consistent with the usual financial programming structure. The rest of this paper has four sections. Section I presents a discussion of the financial programming framework used to describe the linkages between inflation, the balance of payments, interest rates, and financial aggregates. Section II analyzes the empirical results for a sample of data from July 1976 to June 1980. Section III contrasts these estimates with those obtained recently for Argentina and also examines the reasons for the persistence of high nominal and real interest rates. Section IV sets forth the conclusions.

I. Basic Model

To describe the channels by which the trade, fiscal, and financial reforms influenced the Chilean domestic economy, it is convenient to first describe the determinants of financial system behavior and then turn to the determinants of domestic inflation and the balance of payments. In the financial markets, this model distinguishes between the behavior of banks, financieras, and the nonfinancial sector. The financieras are financial institutions that accept short-term deposits and make short-term consumer and enterprise loans. The deposits and loans of financieras are viewed to some degree as substitutes for comparable bank assets and liabilities. The nonfinancial sector consists of households and firms. In each sector, it is assumed that wealth holders attempt to achieve some desired stock and composition of wealth or net worth but that adjustment costs make it unlikely that actual and desired portfolio holdings are equal at each instant.

NONFINANCIAL SECTOR

The nonfinancial sector is taken as holding currency, demand deposits, bank time deposits, financiera time deposits, and real capital as assets, and bank, financiera, and foreign loans as liabilities. Since the bank loan rate and the time deposit rate are taken as representative domestic market interest rates, one can begin with the specification of nonfinancial sector demand for bank loans and time deposits.

The nonfinancial sector’s demand for bank loans is assumed to be a positive function of the expected real return on capital, the real cost of foreign borrowing, the cost of borrowing from financieras, and the level of permanent income, and a negative function of the real loan rate. At each instant, the nonbank sector will attempt to eliminate any difference between its desired and actual holdings of bank loans. Thus, the desired stock of bank loans equals

where

  • B = bank loans

  • P = general price level

  • rB = domestic bank loan rate (in percent a month)

  • πe = expected rate of inflation (in percent a month)

  • rF = cost of foreign borrowing (in percent a month)

  • rFL = cost of financiera loans (in percent a month)

  • YT = permanent income

and the stock adjustment behavior is given by

where

L = lag operator (LXt = Xt-1)

D = 1 – L

(Appendix I presents a summary of notation.) Since the expected real return on capital is assumed to be fixed in the short run, its effect has been subsumed into the α1 term. The level of permanent income (as measured by the trend income level) is taken as a proxy for the level of wealth. The In (Y/YT) reflects the cyclical component of the demand for bank loans associated with financing extra output during an upturn in economic activity. The exact sign on γ4 depends on the degree to which firms tend to anticipate cyclical movements. If firms do not anticipate these cyclical movements, they will be forced to obtain bank credit during a cyclical upturn to finance the unanticipated increase in their output. This situation would argue for a positive γ4. Whenever firms anticipate cyclical movements, they may follow a policy of borrowing when the demand for bank credit is relatively low in order to finance accumulation of inventory during a slack period. In this case, γ4 would be negative. The actual value of γ4 is thus an empirical question.1 The financiera loan rate is included to reflect the fact that firms and households can borrow from financieras as well as from banks.2 In an economy without controls on capital inflows or outflows, the cost of foreign borrowing (rF) is composed of three elements: the nominal foreign borrowing rate (rF), any risk premium (θ) attached to lending to domestic nationals, and the expected rate of depreciation of the exchange rate (e). In Chile, the cost of foreign borrowing has also been affected by the capital controls that the authorities have applied in an attempt to limit capital inflows. These controls have taken a variety of forms, including minimum maturity requirements on foreign loans, required deposits at the Central Bank (at a zero or low interest rate) of up to 25 percent of foreign borrowing, ceilings limiting both total foreign borrowing by banks and loans in foreign currency to a certain percentage of a bank’s capital and reserves, and restrictions on the amount of foreign borrowing that could be converted into domestic currency each month. Even though the required deposits on foreign borrowing declined from 25 percent as the maturity of the loan lengthened, the effect of the decline can be approximated by taking the 25 percent deposit requirement as the representative rate during the period April 1979 to June 1980. This deposit requirement was applicable to short-term foreign borrowing. If δ0 represents a vector that takes on the value one between April 1979 and June 1980 and is zero otherwise, the cost of foreign borrowing would become

ignoring for the moment the effects of all other capital controls.

The impact of the minimum maturity requirements and the changing limits on conversion of foreign currency are much more difficult to capture. To simplify, this analysis uses a series of dummy variables (zi) to represent changes in these restrictions.3 The cost of foreign borrowing is thus given by4

The nonfinancial sector’s desired stock of bank time deposits is assumed to be positively related to the level of permanent income and the expected real return on bank time deposits, and negatively related to the expected real return on currency and demand deposits, the expected return on financiera deposits, and the ratio of current to permanent income.

where

  • T = nominal stock of bank time deposits

  • rT = nominal yield on bank time deposits (in percent a month)

  • rFT = nominal yield on time deposits in financieras (in percent a month)

  • Y = real income

The negative of the expected rate of inflation is taken as a measure of the own rate of return on currency and demand deposits, since interest was not paid on demand deposits during the sample period.5 The effect of the financiera deposit yield is important, since nonfinancial entities could always hold financiera rather than bank time deposits. The ratio of current to permanent income represents the fact that a relatively high current level of income may increase the proportion of wealth held as transactions balances, and hence may reduce the proportion held as savings or time deposits.

In attempting to achieve this desired stock of bank time deposits, actual holdings change according to

One portion of the nonfinancial sector’s portfolio that is especially important for determining short-run economic behavior is its holdings of highly liquid assets (i.e., currency, demand deposits, and time deposits). While each of these asset demands could be modeled separately (as in equations (4) and (5)), the formulations here for inflation and balance of payments behavior can be simplified if there is a general representation of the excess stock of liquid assets. In addition, this will allow a direct comparison between these results for the demand for money and those obtained in previous studies that generally used ordinary least-squares regression analysis and ignored the endogenous nature of the money supply. Therefore, broad money (M) is defined as the sum of currency (C), bank demand deposits (N), and bank time deposits (T).6 The desired stock of liquid assets is taken as a positive function of the expected real return on holdings of currency and bank demand deposits, the real yield on bank time deposits, the level of permanent income, and the ratio of current to permanent income, and a negative function of the expected real yield on financiera deposits. Thus,

Since the demand for broad money reflects the underlying demands for currency, demand deposits, and bank time deposits, the signs of γ9 and γ10 are ambiguous. For example, a higher expected real yield on bank time deposits would increase the desired holdings of that asset but depress desired holdings of currency and demand deposits. A rise in the level of transactions will increase the demand for currency and demand deposits but lower that for time deposits. In taking γ12 as positive, it is also assumed that the transactions demand for currency and demand deposits implicit in γ12 outweighs the negative transactions effect associated with time deposits.

The stock adjustment mechanism for liquid assets is given by

BANKING SYSTEM

The banking system holds reserves, loans to domestic residents, and foreign securities as assets, and demand deposits, time deposits, foreign borrowing, borrowing from the Central Bank, and bank capital as liabilities. The banks’ willingness to supply loans or accept deposits reflects their decisions regarding what sources of funds they will utilize and what earning assets they will purchase. Since this analysis is concerned with the short run, the owners’ commitments of funds are taken as given and the use of other sources of funds is assumed to depend on the relative cost of these funds and any restrictions that the authorities impose on their use. Thus, the use of source Si as a proportion of total funds (F) will be a function of the vector [rB] of the relative nominal costs of the various sources of funds and the returns that can be earned on bank assets. Nominal returns are relevant, since bank profits depend on the differential between the cost of bank funds and the returns on the bank’s assets. The use of any given source of funds would fall as its cost rises relative to other sources. A bank’s use of a given source of funds will also be influenced by any government regulations that affect such portfolio selections (these restrictions being represented by the vector g). These considerations mean that Si/F = fiB(rB, g).

These desired sources of funds and holdings of assets are unlikely to be achieved at each instant. Just as for the nonfinancial sector, the cost of adjustment and uncertainties regarding the sustainability of any interest rate structure mean that banks may want to spread their purchases of assets or issuance of liabilities over time. In addition, banks in a less than perfectly competitive industry may have some influence over the level of market interest rates. For example, banks may attempt to establish a loan rate that reflects the cost of bank funds and ensures an “adequate” level of profits. This type of “markup” pricing scheme would imply that banks follow a mixed strategy of changing the quantity of financial instruments supplied or purchased and the yields that banks offer on deposits or charge on loans.

To allow for the possibility of both price and quantity adjustment in the financial system, it is assumed first that the banks’ desired proportion of total bank funds (F) supplied by time deposits is a negative function of the gaps between the cost of time deposits and the rates of return earned on domestic loans and government securities and the cost of foreign funds. Thus,

where

  • F = total bank funds

  • KT = required reserve ratio on bank time deposits

  • rE = interest paid by the Central Bank on reserves held against time deposits (in percent a month)

  • rG = yield on government securities (in percent a month)

The net cost to the bank of utilizing time deposits as a source of funds is the difference between the time deposit rate (rT) and the interest earned on reserves held against time deposits (rEKT) where KT is the required reserve ratio on time deposits. In comparing this net cost with the earnings on domestic loans, the bank recognizes that it can lend only (1 – KT) pesos of every peso of time deposits. The net addition to bank profits of one or more pesos of time deposits is thus rB(1 – KT) – rT + rEKT.

Since foreign borrowing is an alternative source of funds, the banks’ desired supply of time deposits will be influenced by the differential between the net cost of time deposits (rTrEKT) and the cost of foreign funds (rF). Similarly, sales of domestic government securities can be used as substitutes for additional time deposits as a source of funds so that the differential rTrEKTrG is also included in equation (8).

To reflect the slow adjustment of actual to desired time deposit issuance, one has

The pricing behavior of the banking system can be represented by assuming that the banks always adjust the interest rate on domestic loans to ultimately achieve a constant long-run return on bank equity. In the long run, the loan rate (rB*) that will yield a given rate of profit (α5) is7

where an asterisk denotes the long-run value. This relationship means that the banks’ desired loan rate will be positively related to the net cost of time deposits and foreign funds, and negatively related to the return that can be earned on government securities. Given the uncertainty regarding the sustainability of the current mix of interest rates, each individual bank will want to adjust its lending rate not only to reflect general market adjustments of interest rates (e.g., owing to changes in expected inflation) but also to maintain the bank’s share of market activity. While a bank’s oligopolistic position would allow it to raise its interest rates relative to other banks without losing all its loan business, its share of total market lending would decline, thereby reducing its level of profits.8 In such a situation, banks will move the prevailing loan rate only gradually toward the desired level. To reflect this behavior in a simple framework, it is assumed that the adjustment of the loan rate takes the form

where γ16, γ17, and γ18 are estimates of (T/B)*, (F/B)*, and (G/B)*

Since portfolio disequilibrium has been allowed in both the bank and nonbank sectors, this model necessarily contains adjustment equations for both its quantities of loans and deposits and the market interest rates on both of these financial instruments. The nonbank sector’s flow demands for bank loans and time deposits (equations (2) and (5), respectively) are determined by a number of variables, including the levels of the loan and time deposit rates. These demands thus determine the accumulations of bank loans and time deposits, given the values of the loan and time deposit rates. The time deposit rate is assumed to adjust until such time as the nonfinancial sector’s flow demand for bank deposits (equation (5)) is equal to the banks’ flow supply of time deposits (equation (9)). The loan rate is determined by the banks’ desire to gradually move the rate toward the long-run profit-maximizing level (equation (11)). The banks must therefore supply an amount of loans that is consistent with the nonfinancial sector’s flow demand for loans (equation (2)) at the level of the loan rate being quoted by banks. If there are sharp changes in the nonfinancial sector’s flow demand for loans, then the bank may have to vary its supply of loans equally sharply to keep the loan rate on its desired path. In this model, the required funds for these loans could be obtained via sales of government securities or increased foreign borrowing.

FINANCIAL MARKETS, BALANCE OF PAYMENTS, AND DOMESTIC MONEY CREATION

The linkage between financial market behavior, domestic money creation, and the balance of payments rests on three relationships. First, increases in the stock of base money (H) are equal to the weighted sum of the growth of central bank domestic credit (CR) and the conversion of international reserves (XR).

where

X = exchange rate

Second, the stock of base money can be held as currency (C) or bank reserves (CBR). Thus,

Bank reserves reflect holdings of banks’ required and excess reserves. In Chile, there are multiple ratios that vary with the maturity and type of deposit. These ratios have changed frequently in recent years, which makes the identification of required and excess reserves quite difficult. To simplify, an “effective” reserve ratio (ERR) is defined empirically, as follows:

where

  • N = bank demand deposits

  • T = bank time deposits

  • FD = foreign currency bank deposits (converted into domestic currency units)

  • GD = government bank deposits

ERR is thus the implicit average required reserve ratio in the Chilean banking system,9 and it will be taken as exogenous to this analysis.10

When the definition of broad money (M = C + N + T) is combined with equations (13) and (14), there is a nonlinear link between the overall stock of money and the stock of base via a money multiplier. While this model could be used to forecast values of this money multiplier, the objective here is rather to define the endogenous relationship between M, H, C, and T. Linearizing produces

where εi = constants with ε1, ε2≥0; ε3, ε4, ε5, ε6≤0. (See Appendix II for a definition of these constants.) This equation implies that the stock of broad money will rise as base money increases but will fall with any increase in holdings of currency, foreign currency deposits, and government deposits or an increase in the ERR.

The preceding definitions can also be used to specify the total stock of funds (F) available to banks.

where

  • CCB = bank borrowing from Central Bank

  • CA = bank capital

  • FL = foreign borrowing by banks

  • OI = other bank sources of funds

The bank funds available for the purchase of assets thus equal the sum of deposits net of reserve holdings, bank borrowing from the Central Bank, bank capital, bank foreign borrowing, and other bank sources of funds. This nonlinear relationship can be linearized (using M = C + N + T) to yield

where

ε7, ε8, ε9, ε10, ε11, ε12, ε13≥0

(See Appendix II for the derivation of these constants.)

The relationships represented by equations (1)(17) imply that financial market developments are strongly influenced by events in international markets and such domestic factors as inflation, output, and banking system regulations. What is described next is how financial market developments in turn influence the balance of payments and inflation.12

BALANCE OF PAYMENTS

Since the domestic monetary base can increase as a result of central bank purchases of foreign or domestic assets, the overall state of the balance of payments will necessarily be closely related to monetary and portfolio disequilibrium. It is therefore assumed that the balance of payments is influenced by attempts to arbitrage the prices of goods and securities across countries, by portfolio disequilibrium, and by the state of the domestic business cycle. In the long run, goods and financial market equilibrium require that (i) the prices of domestic goods increase at the same rate as world prices adjusted for any changes in the exchange rate and trade restrictions, and (ii) domestic interest rates not differ from comparable foreign interest rates plus the expected rate of depreciation of the exchange rate by more than the risk premium attached to lending to Chilean nationals by foreign financial institutions (i.e., interest rate parity must hold). Short-run departures from either relative purchasing power parity or interest rate parity will result in arbitrage flows that will lead to changes in the authorities’ stock of foreign exchange reserves (assuming that the exchange rate is not perfectly flexible).

In addition, any excess flow demand for money implies an excess demand for base money. If this excess demand for base money is not satisfied by Central Bank creation of domestic credit, then the resulting portfolio adjustments will help to generate a balance of payments surplus. Finally, the balance of payments may be affected by the state of the domestic business cycle. As the level of output expands relative to the economy’s capacity output, imports may increase sharply, which will deteriorate the state of the balance of payments. Thus,13

where

  • PD = price vector for domestic goods

  • PF= price vector for foreign goods (in terms of domestic currency)

PRICE BEHAVIOR

The overall price level (P) can be defined as a log-linear weighted average of the levels of the prices of domestic (PD) and foreign (PF) goods. Thus,

Domestic goods are those produced locally, and foreign prices equal world prices adjusted for exchange rate and tariff (or quota) effects.

The prices of domestic goods are influenced by international price arbitrage, domestic monetary disequilibrium, and domestic cyclical developments. Thus,

The rate of increase in the prices of domestic goods rises relative to the foreign rate of inflation whenever the prices of domestic goods are sufficiently (as given by ε14) below foreign prices, an excess supply of real money develops, or the level of economic activity rises relative to capacity output. The presence of both monetary disequilibrium and the effects of international price arbitrage reflects the fact that during the Chilean reform period the Chilean economy was neither completely open nor closed in terms of international transactions.

EXPECTATIONS

In this model, the expected rate of inflation and expected rate of depreciation of the exchange rate have played important roles in determining the expected real returns on financial assets. To simplify, it is assumed that the private sector forms the expectations on the basis of its past experience with actual exchange rate and price movements. Thus,

where

x = actual monthly rate of change of the exchange rate, and

where

π = actual monthly rate of change in wholesale price index.

This study attempts to identify the values of β6 and β7 that best describe the formation of exchange rate and price expectations.

This type of adaptive-expectations structure is often regarded as “irrational” in the sense that there can be a significant gap between actual and expected price movements for an extended period. Brunner, Cukierman, and Meltzer (1980) and White (1981) have argued, however, that this expectations structure is “rational” (i.e., represents an optimal forecasting technique) whenever economic agents are uncertain about whether observed shocks to the economy are permanent or transitory. In an economy undergoing extensive structural changes, all past observations are useful in identifying the permanency of past shocks. Brunner, Cukierman, and Meltzer (1980) also argued that the size of β6 and β7 is positively related to the ratio of the variance of permanent shocks to the variance of temporary shocks. Thus, if the ratio of the variance of permanent shocks to the variance of temporary shocks is low, then the βi, will be low, giving important weight to past history.

II. Empirical Results

PARAMETER ESTIMATES

Table 2 summarizes the empirical results for the model obtained from a sample of monthly data for the period July 1976 to June 1980. The model was estimated using a full-information maximum-likelihood estimator,14 which allowed for the imposition of the appropriate cross-equation restrictions on parameters. The various behavioral relationships are generally well estimated.

Table 2.Chile: Parameter Estimates, July 1976–June 19801
EquationsDependent

Variable
Explanatory VariableParameterEstimatet-Ratio
(1) and (2)D ln (B/P)adjustmentβ10.0637.66
parameter
constant−0.0343.00
rB–πeγ138.9937.82
rFπeγ217.5914.82
rFL–πeγ3 = γ22
ln(Y/YT)γ420.00
Z1λ10.0112.21
(4) and (5)D In (T/P)adjustmentβ20.1823.50
parameter
constant−0.3114.89
–πeγ52.7650.35
rT–πeγ6102.2152.71
rFT–πeγ798.3942.99
ln(Y/YT)γ81.4041.86
z2λ20.1653.12
z3λ3−0.0113.31
z4λ40.0332.46
(6) and (7)D In (M/P)adjustmentβ30.3089.46
parameter
constant−0.2899.98
–πeγ911.75426.13
rT–πeγ10−1.1241.40
rFT–πeγ114.7634.92
ln(Y/YT)γ12−0.0910.92
(8) and (9)D ln(T/F)adjustmentβ40.2195.79
parameter
constant−0.2717.10
rTrEKTrB(1 – KT)γ135.5161.71
rTrEKTrFγ144.1322.58
rTrEKTrGγ151.6850.78
z2λ50.0141.65
z5λ60.0213.45
(11)D rBadjustmentβ50.2185.35
parameter
constant−0.0156.00
rTrEKTγ163.7964.01
rFγ170.4581.40
rGγ180.8671.54
(15)In Cautocorrelationρl50.95992.41
parameter
constant−0.23844.22
ln Mε10.596
ln Hε21.0
ln Cε3−0.076
ln ERRε4−0.070
ln FDε5−0.043
ln GDε6−0.187
z6λ70.0303.46
z7λ80.0304.51
(16)ln Fautocorrelationρl61.01127.88
parameter
constant0.0060.27
ln Fε71.0
ln Mε80.466
ln Cε90.087
ln ERRε100.325
ln FDε110.020
ln GDε120.086
ln[FD + GD+ CCB +CA +OI + FL]ε130.752
(17)(XR/H)D ln Rconstant−0.2329.26
D ln PDD ln PFγ190.1953.29
rBrFγ201.5797.85
21{ln MdL ln M}γ210.99773.08
–D ln CR]γ220.2809.88
ln(Y/YT)γ230.491.35
z2λ90.0050.31
z3λ10−0.0021.97
z8λ110.0235.15
(18)ln Pconstant−0.0046.59
ln PDγ240.811306.58
(19)D ln(PD/PF)constant0.4815.39
L ln(PDε14/PF)γ250.3385.46
L ln(M/P)
–ln(M/P)γ260.3584.85
ln(Y/YT)γ27−0.2512.84
z3λ12−0.0247.24
z6λ13−0.0562.37
z1λ14−0.0435.91
(20)xeβ60.1
(21)πeβ70.1

The parameter estimates for the nonfinancial sector indicate a relatively slow adjustment of actual to desired portfolio holdings. The mean time lags involved in the adjustment of actual to desired holdings range from 3 months for the demand for broad money to 6 months for the demand for time deposits and to 16 months for the demand for bank loans. The relatively rapid adjustment of broad money holdings quite likely reflects fast adjustment of currency holdings. These adjustment speeds imply a sharp difference between the short-run and long-run portfolio responses to movements in interest rates, inflation, and income. Table 3 illustrates, for example, that an increase of 1 percent in the loan rate would bring about a fall of 0.16 percent in the demand for bank loans in the short run, versus a decline of 2.48 percent when all portfolio adjustments are complete.

Table 3.Chile: Short-Run and Long-Run Elasticities, July 1976–June 19801
EquationDependent

Variable
Explanatory

Variable
Short-Run

Elasticity2

(t-ratio)
Long-Run

Elasticity

(t-ratio)
(2)ln(B/P)–πe−0.012(0.84)−0.192(0.83)
rB−0.157(4.81)−2.478(7.82)
rF0.054(3.62)0.853(4.82)
rFL0.073(3.62)1.157(4.82)
ln YT0.063(7.66)1.03
(5)ln(T/P)–πe0.010(0.25)0.053(0.27)
rT0.866(4.96)4.758(2.71)
rFT−0.881(5.61)−4.841(2.99)
ln Y−0.256(2.31)−1.404(1.86)
ln YT0.438(3.69)2.404(3.18)
(7)ln(M/P)–πe0.091(5.13)5.867(8.73)
rT−0.016(1.30)−0.052(1.40)
rFT−0.072(5.40)−0.234(4.92)
ln Y−0.028(0.93)−0.091(0.92)
ln YT0.336(7.74)1.091(11.08)
(9)ln(T/F)rT−0.115(4.30)0.528(4.35)
rEKT0.029(4.30)0.132(4.35)
rB(1 – KT)0.114(1.98)0.523(1.71)
rF0.044(2.45)0.200(2.58)
rG0.013(0.72)0.058(0.78)
(11)rBrT0.606(7.48)2.781(4.01)
rEKT0.151(7.48)0.694(4.01)
rF0.076(1.21)0.349(1.40)
rG0.102(1.93)0.474(1.53)

The nonfinancial sector’s holdings of bank loans increased whenever there was a higher expected real rate for financiera loans, a higher expected real rate for foreign loans, or an increase in permanent income.15 In contrast, a higher expected real loan rate depressed real borrowing from banks.16 As noted earlier, the short-run interest rate elasticity of the demand for bank loans is much smaller than the longer-term elasticities. This evidence provides further support for the McKinnon (1973) hypothesis that the short-run demand for bank funds is dominated by the demand for working capital. As is discussed later, the interest-inelastic nature of the demand for bank loans is one factor in explaining the persistence of high real interest rates in the Chilean reform period.

Nonfinancial sector holdings of real time deposits increased whenever there was an increase in the expected real return on time deposits or in permanent income. A higher real return on currency and demand deposits (given by −πe) or financiera deposits worked to reduce real time deposit holdings. As income rose relative to permanent income, holdings of time deposits declined, reflecting the shift toward currency and demand deposit holdings to satisfy the need for transactions balances.

Since the adjustment parameter β2 implies a mean time lag for the adjustment of actual to desired holdings of real time deposits of slightly over five months, the flow demand for time deposits has interest rate elasticities that are less than unity in the short run and considerably larger in the longer term (Table 3). The results also indicate that time deposits in financieras are viewed as quite close substitutes for time deposits in banks. The demand elasticities for short-run and long-run bank time deposits with respect to the expected real yields on bank time deposits and financiera time deposits are of roughly comparable size, although of opposite sign. And, despite the interest-elastic nature of the long-run demand for bank time deposits, the short-run demand must be characterized as relatively interest inelastic.17

The results for the demand for broad money reflect the fact that the estimated interest rate parameters are composite terms that are influenced by the underlying interest rate elasticities for currency, demand deposits, and time deposits. As noted earlier, a higher rT – πe that raises the demand for time deposits also lowers that for currency and demand deposits, therefore implying an ambiguous sign for the parameter on rT – πe in the demand for broad money. As shown in Table 2, the demand for broad money responds positively to a fall in the expected rate of inflation, a lower expected real yield on financiera deposits, a higher ratio of current to permanent income, and a higher level of permanent income. In contrast, a higher expected real rate on time deposits does not have a significant effect and is negative in sign. This result suggests that changes in the demand for currency and demand deposits have significantly influenced the demand for broad money.

The contrast between the insignificant and negative coefficient on the rT – πe terms and the significant and negative coefficient on rFT – πe in the demand for broad money reflects the type of portfolio substitutions generated by interest rate changes. A rise in the expected real yield on financiera deposits will lead to portfolio substitution away from currency, demand deposits, and time deposits. Since all three substitution effects work in the same direction, it is not surprising that γ10 is a significant coefficient. An increase in rT – πe however, leads to substitution out of currency and demand deposits and into bank time deposits; therefore, its effect on the demand for broad money is ambiguous. The combination of the highly significant coefficient for rT – πe in the demand for time deposits (γ6) and the insignificant coefficient for broad money (γ9) suggests that significant substitution has occurred between the various components of broad money.

Three general conclusions emerge from these equations for nonfinancial sector portfolio adjustment. First, portfolio adjustment generally occurs gradually, with the longer-term assets and liability holdings adjusting most slowly. Second, as a result of these relatively slow speeds of adjustment, the short-run nonbank demand for financial assets or liabilities tends to be highly inelastic even though there is considerable evidence of much higher long-term interest rate elasticity. Third, the demands for bank loans and time deposits both show considerable interest rate elasticity with respect to their own interest rates and to those of close substitutes.

The proportion of bank funds derived via time deposits was most significantly related to the differential between the net cost of time deposits and the cost of foreign funds (γ14). Although the coefficients on the differential between the net cost of time deposits and the return on loans (γ13) and the return on government securities (γ15) are of the correct sign, they are not highly significant. The results also indicate that the banks’ adjustment of actual to desired holdings of time deposits has a mean time lag of more than four and one-half months, which is not much faster than the speed of adjustment associated with the nonfinancial sector’s demand for time deposits. This relatively slow speed of adjustment implies that the banks’ supply of time deposits has been relatively interest inelastic in both the short run and the long run. The slow adjustment speed and low interest rate elasticities may reflect the banks’ ability to have some influence over domestic market interest rates, the possibility of credit rationing, or high adjustment cost.18

Although the results obtained for the loan rate adjustment equation are generally consistent with the hypothesis that banks attempted to achieve a long-term profit-maximizing loan rate, the estimates of certain parameters appear to be biased by the absence of data on key variables that are important for determining the desired loan rate. The estimate of the adjustment parameter β5 indicates that banks adjusted the loan rate rather slowly to changes in the determinants of the desired loan rate. The mean time lag in this adjustment process is slightly greater than four and one-half months. The coefficients on the net cost of time deposits (γ16), the cost of foreign funds (γ17), and the government securities rate (γ18) are all of the correct sign, although only the net cost of time deposits is statistically significant. In addition, γ16 implies an elasticity of the loan rate relative to the net cost of time deposits (2.8) that seems too large. This estimated elasticity quite likely reflects the fact that the specification here of the determinants of the desired loan rate has excluded a number of variables for which the information is not available. For example, the cost of operating the bank or of issuing demand deposits was not included. These excluded variables may have biased the estimate of the γ16, γ17, and γ18 parameters (and others as well).

The two linear approximations that were used to define the relationship between base money and broad money (equation (15)) and the determinants of movements in total bank funds (equation (16)) were evaluated at the sample means. Since the sample period witnessed rapid structural change and growth, it was found that the residuals from the original linear approximations exhibited positive serial correlation. To minimize the biases induced by such correlation, it was decided to treat each of the identities as if it were a stochastic equation with error term μi, t, where i denotes the equation and t, time. It is assumed that μi, t = ρiμi,t-1 + Vi where Vi is white noise with mean zero and variance σi2. The identities were then transformed so that the ρi, could be estimated directly. These resulting estimates for ρ15 and ρ16 both indicate that obtaining first differences for these identities was the appropriate transformation.19

While equations (12) and (15) illustrate the dependence of domestic monetary growth on the Central Bank’s accumulation of foreign exchange reserves, equation (17) also indicates that the state of the balance of payments is strongly influenced by domestic monetary disequilibrium. The results for parameters γ21 and γ22 in equation (17) show that, whenever Central Bank domestic credit creation was in excess of the flow demand for money, it led to a deterioration in the balance of payments. In addition, an increase of 1 percent a month in the domestic inflation rate relative to foreign inflation led, ceteris paribus, to a decline of 0.2 percent a month in the rate of growth of Chile’s international reserves. In contrast, an increase of 100 basis points in the differential between domestic and foreign interest rates resulted in an increase of 1.5 percent a month in the growth of international reserves. While the coefficient on the difference between current and trend output is insignificant, it is of the wrong sign. Ordinarily, one would expect a level of current income that is high relative to trend income to lead to a deterioration in the current account and, hence, to the state of the balance of payments.20

The results for the price equations (18) and (19) imply that both monetary disequilibrium and international price arbitrage have affected domestic price behavior. Equation (18) indicates that domestic prices receive a weight of 81 percent in the overall price index. The rate of increase in domestic prices relative to foreign inflation responded significantly to the lagged ratio of domestic to foreign prices and to the excess demand for money. The size of γ25 implies that there was a mean lag of approximately three months between any change in foreign prices and the resulting change in domestic prices. The coefficient on the ratio of current to trend income is significant but of the wrong sign. One would have expected greater price pressure as domestic demand rises relative to trend output. It may be, however, that this variable is more indicative of the availability of domestic goods than of demand pressure; ln (Y/YT) would then be high during periods when the supply of goods is high relative to trend output, thus helping to depress prices.21

The likelihood function for this estimator attains its maximum value when the adaptive expectations coefficients for both exchange rate and price expectations are assigned the value 0.1. The mean time lag in the adjustment of actual to expected price and exchange rate changes is thus 11 months. These relatively long lags could imply that the ratio of the variance of permanent shocks to the variance of temporary shocks is low and that important weight has been given to past history in determining underlying trends in inflation and exchange rate movements. This type of behavior could reflect private sector uncertainty about the sustainability of certain reforms.

IN-SAMPLE FORECASTING EFFICIENCY

Table 4 provides the static and dynamic in-sample forecasts of this model. The static forecasts utilize the actual values of the exogenous variables and the lagged endogenous variables, whereas the dynamic forecasts use the lagged endogenous variables generated by the model. The mean-squared errors for the static forecasts are less than 1 percent except for currency (which is derived via the linear approximation of the relationship between broad money and base money) as implied by equation (15). The dynamic forecasts, not surprisingly, suggest somewhat larger errors, ranging up to 7 percent for currency. The relatively large error for currency reflects the fact that the linear approximation of the nonlinear money multiplier relationship is not entirely accurate.

Table 4.Chile: In-Sample Forecasts—Mean-Squared Errors, July 1976-June 1980
VariableMean-Squared Error

of Static Forecasts
Mean-Squared Error

of Dynamic Forecasts
1n B0.0570.643
1n T0.0961.053
1n M0.0540.298
rT10.00030.0003
rB10.0010.005
1n C1.9796.960
1n F0.0390.418
1n R0.0440.047
1n P0.0260.100
1n PD0.0400.153
1n H0.0440.329

III. Comparisons with Earlier Empirical Results

Before contrasting the results obtained in this study with earlier studies, it is important to stress that the estimates presented in this paper come from the early and middle phases of the Chilean reform period. Owing to breaks in the basic data series used in the model, 22 the analysis cannot be extended directly into the 1981 and 1982 period during which the Chilean financial system and economy experienced some difficulty. It would be interesting to contrast estimates from the period 1976-80 with those for the entire period 1976-82. Such a comparison would provide considerable evidence on the stability of portfolio demands and supplies.

There are certain similarities between the results obtained in this study and those from an earlier study, which was of Argentina (Mathieson (1982)). Although the model used in the Argentine study has characteristics quite close to those utilized in the current analysis, the Argentine model did not allow for portfolio disequilibrium in the banking system and did not fully identify the linkages between base money, the balance of payments, and broad money. In both cases, however, the nonfinancial sector exhibited a relatively slow adjustment of actual to desired stock-holding for time deposits, broad money, and bank loans. The mean time lags in the adjustment process for time deposits were 6 months for Chile and 16 months for Argentina. For broad money, the mean time lags were 3 months for Chile and 26 for Argentina; for bank loans, the comparable lags were 16 months for Chile and 33 for Argentina. The slower adjustment speeds for Argentina could reflect the fact that, in the Argentine model, banks were assumed to be in continuous portfolio equilibrium. If market adjustments in holdings of these assets really reflected the portfolio disequilibrium of both banks and nonbanks, then misspecifying the portfolio adjustment process for banks could lead to slower estimates of the adjustment process for the nonfinancial sector.

The relatively slow speeds of adjustment for nonbank portfolio holdings imply a sharp difference between short-run and long-run interest rate and income elasticities in both Chile and Argentina. The short-run demands for financial assets and liabilities therefore have been much more interest inelastic than the long-run demands. By far the most interest-inelastic short-run demand curves have been those for bank loans. In Argentina, the short-run interest rate elasticity was only -0.003, which was even lower than the elasticity in Chile (-0.16). As will be discussed shortly, these characteristics of interest rate elasticity have been one factor contributing to the persistence of high real loan rates in both countries.

In these studies, the overall state of the balance of payments was influenced significantly by both monetary disequilibrium and price and interest rate arbitrage. Since the Argentine model did not allow for the full simultaneous relationship between money and the balance of payments, the estimates are not directly compatible with those of the current model. It is interesting, however, that a greater proportion of any excess supply of money seems to have spilled over into the balance of payments in Chile than in Argentina. This result could reflect the fact that, during the time periods being considered, Chile had proceeded much further in its trade reforms than Argentina. In contrast, the overall state of the balance of payments seems to have been much more responsive to price and interest rate arbitrage in Argentina than in Chile. Argentina’s greater responsiveness to interest rate differentials is likely to have been related to the lower level of capital controls in Argentina. The more rapid response to inflation differentials in Argentina is somewhat puzzling because, during the time periods considered, Chile had fewer trade barriers than Argentina. This result could be modified by a more appropriate specification for the Argentine model of the linkages between the balance of payments and domestic monetary equilibrium.

The price equations indicate that, although international price arbitrage had roughly comparable effects on domestic inflation, the effect of any excess money supply was much more inflationary in Argentina than in Chile. These estimation results are again consistent with the view that Chile was a more open economy and, hence, that any excess supply of money would spill over more readily into the balance of payments than into the domestic inflation.

The results for the bank and nonfinancial portfolio demands and supplies also imply that the high level of nominal interest rates have encompassed quite different real interest rate behavior for loans and deposits. During the period July 1976-June 1980, the Chilean nominal loan rate averaged 6.4 percent a month and the nominal time deposit rate, 4.7 percent a month. Given the estimate of the adaptive expectations parameter (β7), the average expected rate of inflation was 5.0 percent. These averages imply an average real loan rate of 1.4 percent a month and an average real time deposit of -0.3 percent a month. For the period March 1977-December 1979, Argentina experienced an average expected real 30-day loan rate of 0.9 percent a month and an average expected real time deposit rate of -0.5 percent.23 Thus, for both countries, there was a relatively high ex ante real loan rate combined with a low or slightly negative expected real time deposit rate.24 These results raise two related questions: Why was there such a high real loan rate? Why was there such a large spread between the lending and deposit rates?

The high real loan rate seems to have reflected both an interest-inelastic demand for bank loans on the part of nonfinancial sector portfolio owners and a relatively slow adjustment on the part of banks toward increasing the real supply of bank loans. As noted earlier, the Chilean nonfinancial sector’s demand for bank loans had a short-run interest elasticity of only -0.16.25 This elasticity alone would mean that an exogenous increase in the demand for loans or a reduction in the supply of loans would ensure a sharp increase in the loan rate unless the supply of bank loans was quite interest elastic. Given the nature of Chilean controls on capital inflows, any short-run increase in the supply of bank loans would have to come via an increase in the issuance of bank time deposits. As shown in Table 3, however, the bank supply of time deposits had a short-run interest elasticity (with respect to the loan rate) of only 0.11. In the short run, both the nonbank demand for bank loans and the bank supply of such loans were therefore quite interest inelastic. These characteristics of portfolio interest elasticity thus provide some insights into why there might be considerable short-run instability of interest rates; but, by themselves, these characteristics do not explain the sustained high level of the real loan rate.

The combination of a persistent high real loan rate, the need to maintain an extensive system of capital controls (at least in Chile), and rapid real growth in the amount of real credit available can be explained in terms of the financial market conditions that prevailed at the start of the financial reform and the market’s response to the reforms undertaken. The Chilean financial system started its reform with a small initial real stock of credit and a large excess demand for that stock at the prevailing ceiling for loan and deposit interest rates. The small initial stock reflected the highly variable and negative real returns on financial assets of the pre-reform period. When this small real stock of credit was combined with an interest-inelastic demand for credit, it is not surprising that a high real loan rate prevailed. This excess demand persisted over time because neither capital inflows nor larger real holdings of time deposits grew sufficiently rapidly to eliminate the excess demand for credit.26 Capital inflows were inhibited by capital controls that had been established to assist monetary control; and, at least in the short run, the supply of bank time deposits was relatively interest inelastic with respect to both the loan and time deposit rates. Thus, to encourage banks to undertake continued issuance of time deposits, a loan rate was needed that was high relative to the cost of time deposits. It was still true in Chile, however, that there was a significant change in the relative real returns on bank loans and time deposits. During the first half of the sample period (July 1976-June 1978), the ex ante real loan rate averaged 2 percent a month, whereas the ex ante real time deposit rate had an average value of –0.6 percent. For the second half of the sample period, the ex ante real loan rate declined to only 0.7 percent a month, while the ex ante real time deposit rate rose to –0.2 percent. Thus, the expected real loan rate fell by two thirds during the two subperiods, but the expected real deposit rate continued to rise.

The actual size of the required spread between the loan and deposit rates also reflected the effects of government regulation and a number of economic factors. First, the deposit and loan rate spread was affected by such government regulations as the required reserve ratio and the payment of interest on reserves. The resulting net spread equaled the loan rate minus the time deposit rate adjusted for the payment of interest on required reserves (rEKT) divided by one minus the required reserve ratio on time deposits (rB - [rT - rEKT]/(1 - KT)). The division by 1 - KT reflects the fact that banks could use only 1 - KT percent of each peso of time deposits that they received. The net spread for the period July 1976-June 1980 tended to decline over the sample period but still averaged 1.3 percent a month. During the period June 1979-July 1980, however, the mean spread declined to only 0.5 percent a month.

The continued existence of a positive net spread reflected the cost of financial intermediation, the impact of inflation, the desire of banks to earn an adequate return on their equity capital, the risks associated with interest rate and exchange rate variability, and the competitive structure of the financial system. Banks must earn some positive net spread to cover their operating cost. In addition, the required reserve ratio is a means of imposing an inflation tax on both borrowers, whose loan cost is driven up, and lenders, whose return on deposits is driven down as the tax is imposed. The gap between the nominal loan and deposit rates can be shown (McKinnon and Mathieson (1981)) to be related directly to the expected rate of inflation by a factor equal to KT/(1 - KT) in this model. While Chilean inflation and required reserves were both quite high at the beginning of this sample, they declined over time. Since banks must bear the risks associated with interest rate and exchange rate variability, the net spread will be larger, the greater are these risks. These factors provide some explanation for the spread between lending and deposit rates being large at the beginning of the period and declining over time. Finally, the spread also could have reflected some element of oligopolistic power on the part of financial institutions. The empirical results here do not allow one to establish the importance of this factor.

IV. Conclusions

This analysis has shown that it is possible to estimate a financial programming model for periods during which an economy is undergoing extensive trade, fiscal, and financial reforms. The current model reflects a modification of earlier models in that it allows specifically for endogenous interest rates and the continuous impact of programs designed to gradually open the economy to international trade and capital flows.

The estimation results for Chile indicate that domestic bank and nonfinancial portfolio owners have adjusted their actual to their desired asset or liability holdings only gradually. There is thus a sharp distinction between the short-run and long-run income and interest rate elasticities of demands for and supplies of financial assets. The nonfinancial sector’s demand for bank loans has been especially interest inelastic. The results indicate that both monetary disequilibrium and international price and interest rate arbitrage played significant roles in explaining the behavior of the domestic rate of inflation and overall balance of payments position. The presence of both monetary and arbitrage effects reflects the fact that the Chilean economy was in a period of transition from a closed to an open economy.

A comparison with earlier results obtained for Argentina suggests that certain common characteristics were found in both the Argentine and Chilean reform periods. In both countries, the reform periods witnessed average ex ante real loan rates of roughly 1 percent a month and an average ex ante real deposit rate that was approximately zero or slightly negative. The high real loan rates encountered during the early stages of the financial reform reflected the low initial stocks of real credit and deposits and the uncertainties created by high and variable inflation rates. While the real loan and time deposit interest rates have declined over time, these real interest rates have shown considerable variability and have remained high as a result of interest-inelastic nonfinancial sector demand for bank loans and banking system supply of time deposits. The short-run interest-inelastic nature of these porfolio demands and supplies reflected in part a rather slow adjustment of actual to desired portfolio holdings. The level of interest rates and the spread between deposit and lending rates has also proved sensitive to such government policy as the level of required reserve ratio, the payment of interest on reserves, controls on capital inflows, and policies governing entry into the financial system.

The characteristics of the price and balance of payments equations for the two countries also show evidence of the effects of both monetary disequilibrium and international price and interest rate arbitrage.

APPENDICES
I. Notation

B = nominal stock of bank loans

P = wholesale price index

rB = bank loan rate (in percent a month)

πe = expected rate of inflation (in percent a month)

rF = nominal interest rate on foreign borrowing (in percent a month)

rF = nominal interest rate on foreign borrowing plus expected rate of change in exchange rate plus any risk premium (in percent a month)

rFL = nominal interest rate on loans from financieras (in percent a month)

YT = permanent income

X = exchange rate (pesos per dollar)

x˙e = expected rate of change in exchange rate (in percent a month)

T = nominal stock of bank time deposits

Y = current output

rFT = nominal yield on time deposits in financieras (in percent a month)

N = nominal stock of bank demand deposits

C = nominal stock of currency

M = nominal stock of broad money (= C + N + T)

F = total bank funds

rE = nominal interest rate paid on bank reserves (in percent a month)

KT = required reserve ratio on bank time deposits

rG = return on government securities (in percent a month)

H = nominal stock of base money

R = foreign exchange reserves

CR = Central Bank domestic credit

CBR = commercial bank reserve

ERR = effective reserve ratio

FD = foreign currency bank deposits

GD = government bank deposits

CCB = commercial bank borrowing from Central Bank

CA = commercial bank capital

Ol = other bank sources of funds

FL = foreign borrowing by banks

PD = domestic component of wholesale price index

PF = foreign component of wholesale price index

II. Sources and Definitions of Variables

IFS = International Monetary Fund, International Financial Statistics. The line numbers reported in Variables refer to the page containing the Chilean data.

BCC = Banco Central de Chile, Boletín Mensual.

Variables

The figures used in the analysis are the monthly averages of the following end-of-month stocks.

Currency=Line 14a in IFS
Demand deposits=Line 24 in IFS
Time and savings deposits=Line 25a in IFS
Broad money=Sum of lines 14a, 24, and 25a in IFS
Bank loans=Line 22d in IFS
Government deposits=Line 26d in IFS
Foreign currency deposits=Line 25b in IFS
Base money=Line 14 in IFS
Foreign exchange reserves=Difference between lines 11 (Central Bank foreign assets) and 16c (Central Bank foreign liabilities) in IFS
Commercial bank reserves=Line 20 in IFS
Commercial bank borrowing from Central Bank=Line 26g in IFS
Commercial bank capital=Line 27a in IFS
Other sources of bank funds=Line 27r in IFS
Foreign borrowing by banks=Line 26c in IFS
Price level (wholesale price index)=“Indice de precios al por mayor, Indice General” (BCC)
Domestic component of wholesale price index=“Indice de precios al por mayor, Productos Nacionales” (BCC)
Foreign component of wholesale price index=“Indice de precios al por mayor, Productos Importados” (BCC)
Income—monthly industrial production index=“Indice de producción industrial manufacturera, Indice General” (BCC)
Permanent income=Time trend for monthly industrial production index
Loan rate (in percent a month)=“Tasas de interés efectivas mensuales cobradas en colocaciones a corto plazo, I. Bancos” (BCC)
Time deposit rate (in percent a month)=“Tasas de interés efectivas mensuales pagadas en captaciones a corto plazo, I. Bancos” (BCC)
Foreign interest rate (in percent a month)=Rate paid by domestic nationals under Article 14 borrowing (Source: Central Bank of Chile)
Government security rate (in percent a month)=Until December 1978, this yield equals the midpoint in the range on “Tasas de interés en el mercado secundario de pagarés de tesorería por bancos” (BCC). From January 1980, “Pagarés de tesorería” no longer issued. Then, used midpoint in monthly range on “Pagarés descontables del Banco Central.”
Interest paid on required reserves against time deposits (in percent a month)=Series provided by the Western Hemisphere Department of the International Monetary Fund
Financiera loan rate (in percent a month)=“Tasas de interés efectivas mensuales cobradas en colocaciones a corto plazo,

II. Sociedades Financieras” (BCC)
Financiera time deposit rate (in percent a month)=“Tasas de interés efectivas mensuales pagadas en captaciones a corto plazo,

II. Sociedades Financieras” (BCC)
Required reserve ratio on time deposits=“Tasas de Encaje del Sistema Financiero, Moneda Nacional, Depósitos y Captaciones de 30 a 89 días plazo” (BCC)
Exchange rate=Line ae of IFS

Dummy variables

Z1 Dummy for loosening of capital controls on foreign borrowing at end of June 1979. Included reduction in the percentage of any loan that was required to be deposited in the Central Bank and removal of ceiling on bank foreign borrowing, (one, June-December 1979)

z2 Dummy for impact of financiera crises in early and mid-1976 (involving the closure and merger of various financieras). This led to sharp increase in deposits at banks, (one, July 1976-April 1977)

z3 Dummy representing same event as z2 but containing time trend between July 1976 and April 1977.

z4 Dummy representing imposition of limitation on monthly conversion of foreign loans into pesos. An individual bank was initially limited to an amount equal to 5 percent of the bank’s capital and reserves, (one in September 1977 and April 1978)

z5 Dummy representing period prior to relaxation of capital controls at end of 1978. These changes included a higher ceiling on Article 14 borrowing, larger limits on the monthly conversion of foreign exchange, and a shorter minimum maturity on these borrowings, (one, September-November 1978)

z6 Dummy representing increase in minimum maturity on foreign borrowing to two years, (one in July 1976)

z7 Dummy representing initial effects of elimination in April 1980 of monthly ceiling on conversion of foreign loans into pesos, (one, May-June 1980)

z8 Dummy representing period between end of ceiling on total bank foreign borrowing under Article 14 (in late June 1979) and imposition of monthly limit on bank conversion of foreign borrowing into pesos (in mid-September 1979). (one, July-October 1979)

The constants given in equations (15) and (16) will equal (with m = MIH):

where a bar over a variable denotes the mean, 1n is the natural logarithm, and e is the exponential.

BIBLIOGRAPHY

Mr. Mathieson, Assistant Chief of the Financial Studies Division of the Research Department, holds degrees from the University of Illinois and Stanford University. He has been on the faculty of Columbia University.

The author has benefited from the comments of Fund colleagues but is naturally responsible for any remaining errors. An earlier version of this paper was presented at the XIX Meeting of Central Bank Technicians of the American Continent at Santiago, Chile in November 1982.

It must be stressed that the ln[Y/YT] term is not the only way in which cyclical developments affect the demand for bank credit. The levels of the interest rates are also affected by cyclical changes in economic activity; these effects are transmitted through the γ1, γ2, and γ3 terms.

While the nominal yields on financiera loans and deposits are taken as exogenous, they should be included as endogenous variables in this model of financial system behavior. Unfortunately, the absence of time-series data on the stocks of outstanding financiera loans and deposits makes identification of the appropriate demand and supply functions impossible.

The exact nature of these dummies is discussed in Section II, Empirical Results.

The true interest rate cost (ignoring capital controls) is given by (1+rF)=(1+rF)(1+θ)(1+x˙e). The linear approximation given in equation (3) has been used in order to have a variable elasticity formulation for the portfolio demands and supplies.

The real return on currency and demand deposits actually equals rcπe, where rc is the nominal return paid on currency and demand deposits. For Chile, however, rc=0.

The absence of consistent time series on financiera deposits prevents the inclusion of these holdings.

This relationship can be derived from the banks’ profit and loss statement. If E is defined to equal the owners’ equity in the bank and rQ is the yield on that equity, then rBB+rGG=rTTrEKTT+rFF+rEE. Thus, rB=(rTrEKT)(T/B)+rF(F/B)rG(G/B)+rQ(E/B). In equation (10), α5 = rQ(E/B).

The bank’s profits will decline because it will be in the elastic portion of the kinked demand that it faces in the market. This type of demand curve reflects the assumption that, at any moment, other banks would not match increases in loan rates above some prevailing market interest rate but they would match lower loan rates.

Once again, financiera holdings of reserves are being excluded because of the absence of data.

This simplification means that the effects of shifts between various classes of deposit on ERR are being ignored.

11

This last variable is used because CCB + CA + OI has negative elements, since OI is a variable based on a net definition.

Since this model is estimated over a sample of monthly data, the level of domestic output is taken as exogenous.

This balance of payments equation reflects the fact that Chile did not have a floating exchange rate during the sample period.

See Wymer (1978) for a description of the RESIMUL program.

The estimation program was unable to identify γ4 despite a variety of changes in the specification of the demand for bank loans; thus, γ4 was set equal to zero.

The demand for domestic bank borrowing was reduced when restrictions on foreign borrowing were loosened somewhat beginning in June 1979.

The nonbank demand for time deposits has also been affected by changes in capital controls and crises in the financiera system. In late 1976 and early 1977 holdings of bank time deposits increased as a result of the failures of some financieras. This shift is represented by two dummy variables—z2 and z3: z2 contains ones for the period July 1976 to April 1977 with zeros elsewhere; and z3 contains a time trend for the period July 1976 to April 1977 and zeros elsewhere. The estimated parameters (λ2 and λ3) indicate that this shift of deposits had its greatest impact during July 1976 and then gradually diminished over time. Between September 1977 and April 1978, z4 contains ones and zeros elsewhere; it also represents the impact of imposing a limit on commercial bank conversion of foreign loans into domestic currency. This no doubt influenced the willingness of banks to issue time deposits and quite likely resulted in banks improving the characteristics of as well as the yield on time deposits. The effect of the improvement in the characteristics of time deposits on the nonfinancial sector’s demand for these assets is represented by z4.

The banks’ issuance of time deposits was also affected by two exogenous events. During portions of 1976, the financiera system suffered a number of failures by financial institutions. These failures led to a flow of funds from the financieras to the banks that raised the proportion of time deposits to bank total funds. This situation was represented by a dummy variable (z2), which took on the value one between July 1976 and April 1977 and was zero otherwise. Second, at the end of 1978, the authorities relaxed capital controls in terms of a higher ceiling on borrowing under Article 14 of Chile’s foreign investment law, larger monthly conversion of foreign funds, and a shorter minimum maturity on those borrowings. In the period just prior to this relaxation (September 1978-November 1978), banks either found that the existing restrictions severely limited their foreign borrowing or they decided to delay some of their foreign borrowing to the period of lower restrictions. In either case, there was increased reliance on bank time deposits as a source of funds. This is represented by a dummy variable (z5) that has ones in September 1978-November 1978 and zeros elsewhere.

The linearization for the relationship between broad money and base money was also affected by changes in capital controls that altered the nature of the money multiplier. To capture these shifts, two dummy variables were used: (1) z6 has a one in July 1976 and zeros elsewhere, representing the effects of the increase in the minimum maturity on foreign borrowing by banks from 6 to 24 months; (2) z7 has ones in May and June 1980 and zeros elsewhere to represent the effects of the elimination of monthly limits placed on banks’ monthly conversion of foreign loans into pesos. The impact of these changes in capital controls on the money multiplier implicit in equation (15) is given by the estimated values of λ7 and λ8.

The state of the balance of payments was also affected by both financial system crises and changes in capital inflows. The impact of the financiera system failures in 1976 (described in footnote 17) is represented by z2 and z3; z8 contained ones in the period July 1979-October 1979, which basically represented the period between the end of the ceiling on bank foreign borrowing (in late June 1979) under the afore-mentioned Article 14 and the imposition of the monthly limit on banks’ conversion of foreign exchange into pesos (established in mid-September 1979). This relaxation of capital controls naturally resulted in a substantial capital inflow that raised the accumulation of reserves.

The linkages between domestic price and domestic monetary disequilibrium were influenced by domestic financial crises and changes in capital controls. The financiera system crisis of 1976 is represented by z3 and indicates that the real growth of broad money holdings (especially bank time deposits) that was observed during this period reflected, in addition to the ordinary demand determinant, a shift in the demand for bank deposits relative to financiera deposits. The term λ12 provides a rough estimate of the effects of the shift in the demand for broad money on price behavior during this period; z1 and z6 are described in footnotes 17 and 19, respectively. These variables represent effects of the establishment of the two-year minimum maturity on foreign borrowing (z6) and the temporary relaxation of capital controls during mid-1979 (z1) on the linkage between domestic price formation and monetary disequilibrium.

This is especially true for the series on banks’ external liabilities.

Although there is no accurate estimation of the average cost of funds for the banks, in Chile the nominal average cost of funds derived from time deposits (inclusive of interest on reserves) was 3.49 percent a month (-1.51 percent in real terms); from foreign loans (nominal rate plus expected rate of depreciation only), 4.85 percent a month (-0.15 percent in real terms); and from sale of government securities, 3.47 percent a month (-1.53 percent in real terms).

It must be remembered that these estimates of the real interest rates on loans and deposits are based on the assumption that price expectations are formed on the basis of an adaptive-expectations structure that uses previous experience with inflation to form future expectations about inflation. Since inflation was generally declining over the sample period, a more forward-looking expectations structure might imply somewhat higher real rates of return.

This low-elasticity estimate could also be affected by the fact that many bank loans in Chile are not made at the peso interest rate that has been used here but rather at a variety of other rates that are tied to the cost of borrowing in terms of U.S. dollars. Although an estimate (rF)

of the dollar costs has been included here, it may not have captured the full effects of the ability of banks and nonbanks to substitute dollar for peso loans.

An alternative explanation also consistent with this model is that the persistence of the high real loan rate reflects the loan rationing that banks have undertaken to achieve a desired path for the loan rate that maximizes profits over time.

Other Resources Citing This Publication