Journal Issue
Share
Article

The Relative Price of Nontraded Goods, Absorption, and Exchange Rate Policy in Chile, 1974–82

Author(s):
International Monetary Fund. Research Dept.
Published Date:
January 1988
Share
  • ShareShare
Show Summary Details

During the late 1970s and early 1980s the Chilean economy underwent major structural changes consisting in the partial liberalization of trade and capital flows.1 The macroeconomic policies pursued during the liberalization resulted first in an economic boom and afterwards in recession and crisis. From 1978 to 1981 Chile received a massive inflow of foreign financial capital, which allowed the relative price of nontraded goods to increase continuously, and real output and real absorption expanded at a rapid rate.2 The economic bonanza, however, proved to be unsustainable, and following the contraction in international credit flows to developing countries in 1982 a deep economic crisis developed. As a result, the country was left with a weak traded-goods industry, whose development had been restrained by the extremely high relative price of nontraded goods; a heavy foreign debt burden; and a deep financial crisis.

The sizable increase in the relative price of nontraded to exportable goods (appreciation of the real exchange rate) can be explained partly by the large capital inflows and partly by the use of the nominal exchange rate as an anti-inflationary instrument. Exchange rate policy was used to reduce inflation, from 1976 to 1979, by preannouncing a steadily decreasing rate of devaluation, and from 1979 to 1982, by fixing the exchange rate to the U.S. dollar. Inflationary inertia, however, produced by the indexation of wages and other contracts, kept inflation high—resulting in a sizable increase in the relative price of nontraded to exportable goods (see Table 1).3

Table 1.Macroeconomic Indicators of the Chilean Economy
Nontraded-

Goods

Pricesa
Importable-

Goods

Pricesa
Public

Sector

Wagesa
Real

GDPb
Real

Absorptionb
Unemployment

Rate
Current

Accountc
Capital

Accountcc
Inflation

Ratec
Devaluation

Rated
Year(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)
197462.1072.9454.751.00-0.9010.30-2.391.43619.13649.55
197554.15100.3754.92-12.62-20.6416.10-4.503.43377.30551.49
197672.09100.0270.703.430.1818.000.900.37234.14165.89
197798.9796.90107.239.6114.0713.00-3.703.81113.7864.92
197896.33112.24106.638.039.5912.80-5.309.4850.0247.04
1979104.2798.29121.708.1110.3912.50-5.5010.4035.8117.65
1980124.3989.91151.737.639.2011.70-7.8012.5335.884.71
1981154.47104.68197.245.6110.659.00-16.5016.5019.530.00
1982146.50125.93194.46-13.98-24.4623.20-9.605.069.5330.54
Sources: Column (1): Index 1977.4 = 100; nontraded-goods prices were obtained from selected goods in the consumer price index (CPI) (see Le Fort and Gillet (1986)); exportable prices were obtained from trading partners’ prices (see Le Fort (1985)). Column (2). index 1977.4= 100; source for importable goods prices: Moran (1983). Column (3). index 1977.4 = 100; source: Chilean Statistical Office (INE). Column (4), source: Cuentas Nacionales, Banco Central de Chile. Column (5), source: Cuentas Nacionales. Banco Central de Chile. Column (6), source: Department of Economics, University of Chile. Column (7). source; International Monetary Fund, International Financial Statistics (IFS), yearbook issue, line 77bzd. Column (8), source: International Monetary Fund, IFS, yearbook issue, line 77azd. Column (9), CPI annual average; source: Cortazar and Marshall (1980). Column (10), annual average peso-U.S. dollar rate; source: International Monetary Fund, IFS, yearbook issue, line rf.

Relative to exportable-goods prices.

Annual percentage change.

As a ratio of nominal GDP.

Annual rates.

Sources: Column (1): Index 1977.4 = 100; nontraded-goods prices were obtained from selected goods in the consumer price index (CPI) (see Le Fort and Gillet (1986)); exportable prices were obtained from trading partners’ prices (see Le Fort (1985)). Column (2). index 1977.4= 100; source for importable goods prices: Moran (1983). Column (3). index 1977.4 = 100; source: Chilean Statistical Office (INE). Column (4), source: Cuentas Nacionales, Banco Central de Chile. Column (5), source: Cuentas Nacionales. Banco Central de Chile. Column (6), source: Department of Economics, University of Chile. Column (7). source; International Monetary Fund, International Financial Statistics (IFS), yearbook issue, line 77bzd. Column (8), source: International Monetary Fund, IFS, yearbook issue, line 77azd. Column (9), CPI annual average; source: Cortazar and Marshall (1980). Column (10), annual average peso-U.S. dollar rate; source: International Monetary Fund, IFS, yearbook issue, line rf.

Relative to exportable-goods prices.

Annual percentage change.

As a ratio of nominal GDP.

Annual rates.

Keeping a misaligned relative price of nontraded goods (that is, that deviates from the value compatible with internal equilibrium in the labor and goods market and with a sustainable external position) has proven to be costly because it reduces long-run economic growth and exposes the economy to a costly adjustment process characterized by drastic changes in relative prices and reductions in absorption. To avoid the costs of a misaligned relative price, exchange rate policy could, in principle, be used to compensate for the effects of other policy actions and economic rigidities on the relative price of nontraded goods. The extent and duration of the effects of exchange rate actions on the relative prices of nontraded goods is, however, an unresolved empirical issue.

This paper examines the extent and duration of the effects of exchange rate policy on the relative price of nontraded goods in Chile. Two effects of exchange rate policy on relative prices are distinguished. The first is a direct effect, which is the result of changes in the product wage of the exportables sector, produced by changes in the nominal exchange rate that alter production costs and the price of nontraded goods relative to exportable goods. The second is an indirect effect, similar to that of demand-management policies, that results from changes in real absorption produced by changes in the nominal exchange rate.

Three empirical questions are addressed: (1) Do changes in absorption have significant causal effects on the relative price of nontraded goods? (2) Does the exchange rate have a significant direct effect on the relative price of nontraded goods? (3) For how long can the effects of absorption and exchange rate policy on relative prices be sustained? The methodology used to answer these questions involves the design and estimation of a model of an open economy that can accommodate the following alternative theories: (1) a sticky-wages theory, according to which the exchange rate policy can have direct effects on the product wage in the exportable sector and on the relative price of nontraded goods;4 (2) a specific-factors theory, according to which absorption and resource endowments can affect the relative price of nontraded goods; and (3) a purchasing-power-parity (PPP) theory, according to which relative prices of nontraded goods depend only on international prices of traded goods, domestic tariffs, and technology. A semireduced-form equation was estimated over the period 1974–82 using classical and Bayesian procedures and three alternative prior beliefs representing each of the theories.5 The results favored the sticky-wages solution of the model and indicated that expenditure-switching devaluations—changes in the exchange rate keeping absorption constant—have a significant effect on the relative price of nontraded goods.

Some limitations of this empirical study should be considered to assess adequately the conclusions obtained. First, owing to the unavailability of data for the measurement of sectoral technical change and sectoral capital accumulation, certain simplifying assumptions about the role of these factors were made. Second, the short period covered in the sample (eight years) may reduce the validity of the long-run implications of the conclusions. Finally, to some extent the results are conditional on particular characteristics of the Chilean case; the effects of exchange rate and wage policies on relative prices could have resulted in part from the suspension of collective bargaining and the high inflation in Chile during most of the sample period.

Section I presents a simple general equilibrium model for the relative price of nontraded goods in a small developing economy and its alternative solutions. Section II presents the empirical results obtained from the estimation of the semireduced form using classical and Bayesian procedures. Section III summarizes the results and draws out the implications of the exercise for the role of exchange rate and demand-management policies.

I. Relative Price of Nontraded Goods: Alternative Theories

The first theory, sticky wages, assumes that wages do not adjust to clear the labor market and that capital is sector specific; the second theory, specific factors, assumes that wages are flexible and that capital is sector specific; and the third. PPP, assumes that wages are flexible and that factors are mobile among sectors.

The model developed here is a simple general equilibrium model of the type developed by Jones (1965). All the variables are represented in terms of proportional change—a circumflex (ˆ) over a variable is used to denote proportional change—and all the parameters in the model can be interpreted as elasticities.6

The following assumptions are used in the model: (1) The economy produces and consumes three goods, an exportable (T), an importable (M), and a nontraded good (N). (2) Production activity in each of the three sectors takes place under a technical relationship with two factors, capital (K) and labor (L), and constant returns to scale. (3) Factor endowments are predetermined. (4) The economy is small, and the world prices of the traded goods (P*T and P*M) are given exogenously. (5) The domestic prices of traded goods follow the law of one price and are given by the nominal exchange rate (e), their world price, and the domestic nominal protection rates (zT and zM). (6) The price of nontraded goods is determined endogenously by market-clearing conditions. (7) The nominal exchange rate (e) is a policy instrument adjusted by the authorities. (8) Firms maximize profits in a competitive environment. (9) The nontraded good (N) is produced using labor-intensive techniques, and the importable good (M) using capital-intensive techniques. (10) Technological change is neutral (Ĉi represents the technological progress in sector i), is equal in both traded-goods sectors, and is slower in the nontraded-goods sector ̂CT = ̂CM = ̂CNC;c > 1).7 (11) Individuals are rational maximizing agents, and the model is neutral to a proportional change in all the nominal variables; neutrality, however, does not necessarily apply to changes in particular nominal variables. (12) The utility function is weakly separable; consumers maximize independently each period’s utility, by selecting the amounts consumed of each good, and intertemporal utility, by selecting the amount to be spent each period. (13) The demand for nontraded goods depends on the level of real absorption, but it is independent of the composition between consumption, investment, and public expenditure. The expenditure share of each good in investment and public expenditure is assumed to be equal to that of private consumption.

Sticky-Wages Model

The sticky-wages solution is considered the short-run solution of the model and is obtained by assuming that the wage rate does not adjust to clear the labor market and that capital is sector specific.8 Sticky-wages models have been widely used in the open-economy macroeconomic literature (Calvo (1982), Dombusch (1982), and Bruno (1978)) under different rationalizations. Among them are overlapping wage contracts, as in Taylor (1979), and institutional constraints that prevent a free operation of the labor market. In the case studied here, government intervention prevented the free operation of the labor market—collective bargaining was banned in Chile from 1973 to 1979—and the wage adjustments given to government employees became an important signal for the private sector wage structure.

The main implication of wage rigidity for the determination of relative prices is that an expenditure-switching devaluation can have effects on relative prices. Under the rigid-wages solution, wage policy, represented in this case by the wage rate in the public sector (WP), and the exchange rate policy, represented by the nominal price of exportable goods (PT). have direct effects on the determination of the product wage in the exportables sector and thus on the relative price of nontraded and exportable goods. The wage policy variable is a predetermined variable that could be subject to some indexation rule, and the nominal price of exportable goods (PT) and of importable goods (PM) is determined by the law of one price (equations (1) and (2)) and can be controlled through adjustment of the nominal exchange rate:9

and

where

According to the rigid-wages and specific-factor solutions of the model, changes in absorption could also cause changes in the relative price of nontraded goods. Real absorption is a behavioral variable that would respond to policy actions, including changes in the nominal exchange rate. A change in the nominal exchange rate can modify absorption through its effects on the level and composition of real wealth; in addition, a change in the nominal exchange rate can alter expectations regarding future asset returns.

The wage-setting equation was obtained by assuming that wages in the private sector are adjusted according to wages in the public sector (WP) and the excess supply of labor (LsL).10 The wage-setting equation implies that the model is neutral to a simultaneous increase in all the nominal variables and that the wage and exchange rate policy can affect the product wage in the exportable-goods sector. The equation is specified as

The aggregate labor demand is the weighted sum of sectoral labor demands derived from cost-minimization conditions. The weights are given by the shares of sectoral employment in total employment (λi). The sectoral labor demands are derived by equating the proportional change in the capital-labor ratio (̂Ki − ̂Li) with the rate of change in relative factor rewards (̂w − ̂ri) times the elasticity of substitution among factors (σi). The relative factor rewards are obtained from the zero-profit conditions, assuming that rental payments to capital differ across sectors.11 The demand for labor is given by

The third equation of the system is the nontraded-goods market-clearing condition. The supply function of nontraded goods is obtained by replacing the sectoral demand for labor in the production function. The demand for nontraded goods is obtained from the solution of a consumer optimization problem, assuming that the demand for non-traded goods is independent of the composition of absorption with respect to consumption, investment, and government spending. It is also assumed that the utility function is mtertemporally separable: thus the maximization of utility in each period is a second stage of a more general intertemporal optimization problem. The demand for nontraded goods12 is a function of real absorption and of the relative prices of importable and nontraded goods.13 The specification of the nontraded-goods market-clearing condition is given by

The general equilibrium solution for the relative price of nontraded goods is obtained by solving a system of three equations including the wage-setting equation (3), the aggregate demand for labor equation (4). and the nontraded-goods market-clearing condition in equation (5). The determinant of this system (det 1) represents the price elasticity of the excess supply of nontraded goods.

where

Equation (6) shows the equilibrium relative price of nontraded goods in the rigid-wages and specific-factors solution. The elasticity of the equilibrium relative price of nontraded goods with respect to the economy-wide capital-labor ratio is positive because an increase in labor supply reduces real wages. The elasticity of relative prices with respect to absorption is positive because all goods are normal and the quantity supplied of nontraded goods increases as the relative price increases. The elasticity with respect to the wage and exchange rate policy variable (WP/PT) is positive because of the effect of public sector wages on the economy-wide wage rate. The elasticity with respect to the price of importables can be negative if goods N and M are complements (gM < 0) and is positive if they are substitutes. The elasticity with respect to technical change is negative when the reduction in production costs in the nontraded-goods sector more than offsets the increase in the wage rate brought about by technical progress.

Specific-Factors Model

The second solution of the model is obtained by assuming flexible wages and sector-specific capital. Under the assumed conditions, the labor market clears, and rental payments to capital differ across sectors. This solution of the model is similar to the models of Corden and Neary (1982) and Sanyal and Jones (1982). The relative price of nontraded goods to that of exportable goods is a function of absorption, factor endowments, technology, and the relative prices of importable and exportable goods.14 In the specific-factors and flexible-wages solution the exchange rate does not have a direct effect on the relative price of nontraded goods. In this solution of the model, however, changes in the exchange rate can have indirect real effects through induced changes in absorption; a devaluation can reduce absorption, which in turn reduces the relative price of nontraded goods and the producer wage in the exportable-goods sector. The market-clearing conditions are

and equation (5).

The general equilibrium solution for the relative price of nontraded goods under sector-specific factors is obtained from a two-equation system that includes the market-clearing conditions for labor, equation (7), and for nontraded goods, equation (5). The market-clearing condition in the labor market is obtained by equating the rates of change of the supply (Ls) and the aggregate demand of labor, equation (4). The sector-specific-factors solution is

where15

The specific-factors and flexible-wages solution of the model for the relative price of nontraded goods differs from the rigid-wages solution only in the value of two parameters. The parameter that measures the response of wages to the wage and exchange rate policy variable is equal to zero (k0 = 0), whereas the parameter that measures the response of wages to the excess demand for labor is equal to infinity (k1 = ∞). Those two restrictions exclude the wage and exchange rate policy variable from the relative price equation but do not alter the signs of the other elasticities.

Purchasing-Power-Parity Model

The third solution of the model is obtained by assuming that wages are flexible and that capital is mobile across sectors. This solution is consistent with international factor-price equalization, the international equalization of all goods prices, and thus with the PPP condition. If perfect capital mobility across sectors, flexible wages, linear homoge neous production functions, incomplete specialization, and equal number of traded goods and factors of production are assumed, then the solution of the model corresponds to the traditional Heckscher-Ohlin trade model with the addition of anontraded good. The results obtained by Komiya (1967), and then used by Jones (1974a and 1974b), show that the addition of a nontraded good to the Heckscher-Ohlin trade model will not change its essential properties. In particular, the proportional change of prices of traded and nontraded goods—adjusted by exchange rate changes—is equal across countries, unless there are changes in the protection rates or differences in technical progress. (See also Learner (1984), Chapter 1.)

PPP models dismiss the role of absorption and of the nominal exchange rate in the determination of relative prices. The general equilibrium solution for the relative price of nontraded and exportable goods under flexible wages and mobile factors is a function of the domestic relative price of importable and exportable goods and of technical change in the different sectors.16 The sign and size of the effects of exogenous and policy variables on the relative price of nontraded goods depend on the relative factor intensity used in the production of the three goods, which determines the relative magnitudes of the share of labor in total production costs (θiL). In this model, given the assumed relative labor intensity of nontraded goods and relative capital intensity of importable goods, the share of labor in total production cost is largest in the nontraded-goods sector and smallest in the imported-goods sector, θNLTL > >θML

The zero-profit conditions are

When wages are flexible and factors are mobile, capital is reallocated so as to equate the rental rate in all the sectors (ri = r). Under these conditions the price of nontraded goods can be solved from the three-equation system formed by the zero-profit conditions. The three endogenous variables are the price of nontraded goods, the wage rate, and the rental payment to capital; the price of exportables is used as the numeraire. The PPP solution is given by

An increase in the domestic relative price of importable to exportable goods reduces the relative price of nontraded goods according to the PPP solution of the model. An increase in the price of importables relative to exportables reduces the wage rate relative to the rental rate (Stolper-Samuelson theorem), given the assumed relative factor intensities. The relative price of nontraded goods to that of exportable goods would be reduced as the factor rewards are changed in favor of capital, which is used less intensively in the production of nontraded goods than in the production of exportables. Neutral technical change can affect the relative price of nontraded goods to that of exportable goods only if it differs across sectors. If the rate of technical change is equal in the exportables and importables sectors and slower in the nontraded-goods sector, the relative price of nontraded goods to that of exportable goods is increased by technical change.17

Interim Multipliers and Lagged Explanatory Variables

Each of the three solutions presented for relative prices implies different equilibrium conditions and dynamics for the relative price of non-traded goods. A change in one of the explanatory variables produces an impact on the relative price of nontraded goods, followed by a dynamic process created by the adjustment of wages and prices toward full employment, and by the reallocation of capital toward the equalization of rental payments across sectors. To represent the dynamic process, lagged endogenous and exogenous variables are included in the relative prices equation.

Equation (13) is the semireduced form presented in the level of the variables and was obtained from the rigid-wages solution of the model, equation (6), by adding lagged explanatory variables. Given the small sample size, the number of lagged explanatory variables was limited to one lag for the relative price of nontraded goods and two lags each for real absorption and the exchange rate policy variables. In addition, a proxy for the economy-wide technical change (Tec) replaced sectoral technical change.18

The interim multiplier, equation (14), represents the accumulated effect on the relative price of nontraded goods at time t of a change in the explanatory variable x(h) that took place at time t–;j it is used to study the duration of the effects of absorption and the exchange rate on the relative price of nontraded goods. The impact of the variable x(h) will last over time only if the sum of the impact and lagged coefficients of that variable has the same sign as the impact coefficient. If the sum of impact and lagged coefficients of the x(h) variable is greater than zero, then the long-run multiplier, equation (15), is positive:

and

The interim multipliers of changes in the ratio of public wages to the price of exportables are used to estimate the size and duration of the effect of an expenditure-switching devaluation on the relative price of nontraded goods. However, the interim multiplier presented in equation (14) assumes rigid nominal wages in the public sector in the sense that the devaluation is not followed by adjustments in public sector wages.

If public sector wages are indexed, the duration of the effect of a devaluation on relative prices depends critically on the wage adjustment that follows. When public sector wages are indexed to the consumer price index, the long-run multiplier of a devaluation is a function of 7, the indexation coefficient; θN the share of nontraded goods in the price index; and B7, the long-term multiplier of a once-and-for-all change in (WP/PT). In particular, the long-run effect of a devaluation is smaller the larger is the degree of indexation; it is equal to zero when the indexation coefficient (γ) is equal to unity and is larger the larger is the weight of nontraded goods in the consumer price index (γN). The specification is shown in equation (16):

where

II. Empirical Analysis

This section presents the empirical results obtained from the estimation of a semireduced-form equation for the relative price of nontraded goods using quarterly data for Chile over the period 1974–82. The estimations were performed using classical and Bayesian procedures, and the results were used to calculate interim multipliers and the duration of the effects of changes in absorption and of expenditure-switching devaluations on relative prices, Bayesian posterior estimates were obtained using instrumental-variables estimates and three alternative sets of prior beliefs that represent the sticky-wages, the specific-factors, and the PPP solutions of the model.

Although the model is developed in terms of proportional changes, with the parameters representing elasticities, it was estimated in the level of variables. This is consistent because the estimated coefficients are elasticities around the values of the variables in the fourth quarter of 1977; the value of all the variables was normalized to unity in the last quarter of 1977.19

Instrumental Variables and Absorption

The estimation of a linear approximation to the reduced form implied by the model is used to obtain an instrumental variable for absorption. The explanatory variables in the reduced form are those included in the semireduced form, with the addition of a group of instruments for absorption that was obtained from a portfolio model of an open economy that faces an upward-sloping supply of foreign credit.20 The set of instrumental variables includes the previous period’s holdings of real monetary assets (H), domestic real nonmonetary assets (DN), and net external real assets (F); the predetermined elements in the assets rates of return—including the international real interest rate (ir*). the expected rate of change in traded goods prices [E(^PT)],21 and the expected rate of change of the relative price of nontraded and exportable goods [E(PN/^PT)];22 a shift variable for the supply of foreign capital flows, which represents the international creditworthiness of the country (Ip);23 the rate of domestic credit creation less the rate of currency devaluation (RDC); government spending (G), real tax revenues (Trib), and real net interest payment abroad (FS); and finally a qualitative variable to represent the structural break created by the elimination of restrictions on capital flows (CC).24

Estimation Under Different Prior Beliefs

The semireduced form, equation (13), was used to estimate individual coefficients and interim multipliers for absorption and the wage and exchange rate policies under different prior beliefs. The semireduced form coefficients (B) can be represented as functions of the structural parameters of the model derived in Section I. The functional relationships between reduced-form coefficients and structural parameters, however, are different under different theories defining alternative prior beliefs for the B coefficients.

The subjective probability distribution of the B coefficients is assumed to be a multivariate normal; thus prior beliefs require the specification of a vector of means and a covariance matrix of the coefficients. The vector of means represents the location of the prior distribution in the parameter space. The variances represent the degree of confidence in the value selected for the prior mean; a high confidence in the prior coefficients implies small variances; whereas a small confidence implies large variances and a diffused prior distribution.

Posterior estimates are found by selecting a point in the contract curve defined by the tangencies obtained from the restricted minimization of the sum of squared errors subject to the prior probability distribution of the estimated coefficients.25 A sensitivity analysis was performed to check the effects of the location of the prior distribution on the posterior estimates.

The specification of the prior probability distributions for the semi reduced-form coefficients was obtained from the functional relationships between semireduced-form coefficients and structural parameters. The three prior distributions were specified using the same values for the structural parameters. The mean values of the structural parameters were selected using other statistical information, when available, or subjective beliefs. The values assumed for structural parameters are presented in Table 2. The prior standard deviations for each of the coefficients were selected on the basis of judgments about the acceptable 95 percent confidence intervals for each coefficient under the different theories. This procedure implies that all the prior covariances of the coefficients are equal to zero.

Table 2.Parameters and Variables of the Structural Model
SymbolDefinition or Value
Structural parameters (i, T, N, M)
θiLShare of labor in the cost of production of good i
σiHicks-Allen capital-labor elasticity of substitution in sector i
λiShare of sector i employment in total employment
DWage elasticity of the aggregate labor demand
YiProportional contribution of sector i to the wage elasticity of aggregate employment
fnPrice elasticity of the supply of nontraded goods
cProportional difference in the speed of technical progress in the traded- and nontraded-goods sectors
gIncome elasticity of the demand for nontraded goods
gNPrice elasticity of the demand for nontraded goods
gMCross price elasticity of the demand for nontraded goods
θiShare of good i in the consumer price index
k0Response of the economy-wide wage rate to the wage rate in the public sector
k1Response of the wage rate to the excess demand for labor
αNSector N’s share in GDP
Values assumed for structural parameters in the prior beliefs
θNL0.75
θTL0.30
θML0.475
σN0.175
σT0.1
σM0.05
σNL0.5931
σTL0.2256
σML0.1813
C3
k01.32
k10.33
αN0.55
g1
-gN-1
gM-0.15
D0.4711
YN0.8813
YT0.0912
YM0.0275
fn0.525
Variables (i = T. N, M)
^CiHicks’s neutral technical change in sector i
ENominal absorption
eNominal exchange rate in domestic currency per unit of foreign currency
KiStock of capital in sector i
LiEmployment in sector i
LAggregate employment
LsLabor force
PiDomestic price of good i in domestic currency units
PDomestic consumer price index
P*iInternational price of good i in foreign currency
QiOutput in sector i
QdNQuantity demanded of good N
riRental payment to capital in sector i
WPNominal wage in the public sector
wNominal wage rate
ziDomestic nominal protection of good i

Prior distribution 1 represents the sticky-wages solution of the model for the relative price of nontraded goods presented in equation (6). The sticky-wages solution is the most general one considered, and all the prior means are different from zero. The coefficients of lagged variables in prior distribution 1 are specified in such a way that the long-run multipliers of (WP/PT) and of absorption on the relative price of non-traded goods are equal to zero. The prior standard deviations were selected so that the sign of both bounds of the 95 percent prior confidence interval of each coefficient is equal to the sign of the prior mean. The exceptions are the coefficients of (PT/PM) and of technical change, which have ambiguous signs in the rigid-wages solution. Prior means. standard deviations, and 95 percent normal confidence intervals of the coefficients are presented in Table 3.

Table 3.Mean, Standard Deviations, and Normal Confidence Intervals for Alternative Prior Beliefs on the Semireduced-Form Coefficients

(Dependent variable: (PN/PT))

95 Percent
ConfidenceStandard
SymbolMeanIntervalDeviation
Prior 1: Sticky wages
Constant1.14(+/-4E+04)4 E+04
Tec-0.51(0.52,-1.6)0.528
(PT/PM)0.01(0.30.-0.1)0.103
(PT/PM)(t-1)0.4(0.80,0.0)0.201
K-0.68(0.0,-1.4)0.346
(K/Ls)0.10(0.20,0.0)0.049
(E/P)0.68(1.35.0.02)0.339
(E/P) (t - 1)-0.15(0.45,-0.7)0.306
(E/P) (t - 2)-0.53(0.80,-1.9)0.681
(WP/PT)0.41(0.80,0.02)0.199
(WP/PT)(t-l)-0.3(0.90,-1.5)0.612
(WP/PT)(t-2)-0.1(0.44,-0.7)0.281
Prior 2: Specific factors
Constant1.14(+/-4 E + 04)4 E + 04
Tec-0.43(0.44,-1.3)0.443
(PT/PM)0.13(0.39,-0.1)0.130
(PN/PT)(t-l)0.40(0.80,0.0)0.201
K-0.94(0.0.-1.88)0.478
(K/Ls)0.49(0.98,0.0)0.249
(E/P)0.94(1.86,0.0)0.470
(E/P)(t-1)-0.15(0.45,-0.7)0.306
(E/P)(t-2)-0.79(1.19,-2.8)1.009
(WP/T)0(0.20,-0.2)0.102
(WP/T)(t-1)0(0.20,-0.2)0.102
(WP/T)(t-2)0(0.20,-0.2)0.102
Prior 3: Purchasing power parity (PPP)
Constant1.10(+/-4 E + 04)4 E + 04
Tec0.74(0.60,0.0)0.373
(P/T/P/M)1.10(2.19,0.0)0.559
((P/N/P/T)(t-1)0.3(0.60,0.0)0.150
K0.0(0.20,-0.2)0.102
(K/Ls)0.0(0.20,-0.2)0.102
(E/P)0.0(0.20,-0.2)0.102
(E/P)(t-1)0.0(0.20,-0.2)0.102
(E/P)(t-2)0.0(0.20,-0.2)0.102
(WP/Pt)0.0(0.20,-0.2)0.102
(WP/PT)(t-1)0.0(0.20,-0.2)0.102
(WP/PT) (t - 2)0.0(0.20,-0.2)0.102

Prior distribution 2 represents the specific-factors and flexible-wages solution of the model for the relative price of nontraded goods presented in equation (8). The specific-factors theory excludes the effect of the wage and exchange rate policy on the relative price of nontraded goods. The prior means for current and lagged (WP/PT) coefficients are, thus, equal to zero. All the other features of prior distribution 2 are similar to those of prior distribution 1, but the values of the coefficients differ somewhat because they are derived from slightly different functional relationships between the coefficients and the structural parameters.

Prior distribution 3 represents the PPP solution of the model presented in equation (12). The PPP solution does not involve the effects of the exchange rate policy and of absorption on the relative price of nontraded goods. The prior means for the coefficients of (WP/PT), absorption, and factor endowments are all equal to zero. The standard deviation of each coefficient with a prior mean different from zero was selected so that the sign of both bounds of the 95 percent confidence interval is equal to the sign of the prior mean.

The results of the semireduced-form estimation (see Table 4) were obtained using instrumental variables and Bayesian posterior estimates. The posterior estimates were derived from the instrumental-variables estimates and three alternative prior distributions of the coefficients.26 The instrumental-variables estimates can also be interpreted as posterior estimates obtained using a diffuse prior distribution. Posterior estimates under a diffuse prior distribution are obtained by assuming absolutely no confidence in the prior beliefs, then sigma 1—the ratio between the variance of the prior distribution and the variance of the data disturbances—is considered to be a very large number or close to infinite. The larger is sigma 1, the larger is the variance attached to the prior beliefs relative to the variance of the disturbances produced by the data, and the weaker is the confidence in the prior beliefs. Sigma 1 is equal to zero when the posterior estimates are represented by the prior distributions, and infinite when the posterior estimates are represented by the instrumental-variables estimates, Sigma 1 was selected to be equal to unity for the three posterior estimates reported in Table 4.

Table 4.Semireduced-Form Coefficients

(Dependent variable: (PN,/PT)

InstrumentalPosterior Estimatesa
SymbolVariablesPrior 1Prior 2Prior 3
ESSb0.02520.02770.03370.0313
¯R2c0.99160.99070.98870.9895
DW(h)d1.97(-)2.28(-l.5)2.07(-0.3)1.70(1.1)
Sigma 1e1.001.001.00
Constant-0.5260.0860.047-0.469
(-0.7)(0.3)(0.0)(-1.7)
Tec1.3050.4650.4710.686
(1.6)(1.5)(1.5)(2.7)
(PT/PM)0.1160.0900.0690.096
(3.1)(2.8)(2.0)(2.9)
(PN/PT)(t-1)0.5060.3900.4360.450
(2.4)(2.8)(3.7)(4.4)
K-0.816-0.533-0.659-0.036
(-1.0)(-1.8)(-1.9)(0.4)
(K/Ls)0.1920.0960.208-0.027
(0.4)(1.9)(1.1)(-0.3)
(E/P)-0.0150.1100.1670.055
(-0.1)(1.3)(2.0)(0.9)
(E/P)(t -1)0.0610.0120.0140.020
(0.7)(0.1)(0.1)(0.3)
(E/P)(t - 2)-0.141-0.184-0.181-0.107
(-1.6)(-2.4)(-2.3)(-1.8)
(WP/PT)0.5380.5410.4530.461
(10.6)(12)(11)(11)
(WP/PT)(t-1)-0.163-0.070-0.019-0.072
(-1.4)(-0.8)(-(0.3)(-1.2)
(WP/PT)(t-2)-0.102-0.054-0.05011.112
(-1.9)(-1.3)(-1.2)(-3.0)
Note: The sample period was first quarter ot 1974 through the end of 1987; t-values are reported below each coefficient.

Each posterior estimate was obtained using as inputs the instrumental-variables estimates, the covariance matrix of those estimates, and the corresponding prior distributions; the computations were performed using the program SEARCH (Learner and Leonard (1983b)).

The sum of squared errors.

Adjusted coefficient of determination.

The Durbin h-test statistic is in parentheses; it is omitted when it is not a real number. The DW and h tests for posterior estimates were directly calculated from the residuals of the semireduced form including absorption (and not the instrumental variable) as one of the explanatory variables.

Sigma 1 is equal to the ratio of variances (s*2/s2), where s2 is the variance of the disturbance term and (1/s*2) measures the confidence given to the prior beliefs.

Note: The sample period was first quarter ot 1974 through the end of 1987; t-values are reported below each coefficient.

Each posterior estimate was obtained using as inputs the instrumental-variables estimates, the covariance matrix of those estimates, and the corresponding prior distributions; the computations were performed using the program SEARCH (Learner and Leonard (1983b)).

The sum of squared errors.

Adjusted coefficient of determination.

The Durbin h-test statistic is in parentheses; it is omitted when it is not a real number. The DW and h tests for posterior estimates were directly calculated from the residuals of the semireduced form including absorption (and not the instrumental variable) as one of the explanatory variables.

Sigma 1 is equal to the ratio of variances (s*2/s2), where s2 is the variance of the disturbance term and (1/s*2) measures the confidence given to the prior beliefs.

The estimates obtained under a diffuse prior distribution (instrumental-variables estimates) indicate that the wage and exchange rate policy variable (WP/PT) has a significant impact on the relative price of nontraded goods, whereas the impact of absorption is not significantly different from zero. The inference obtained for the wage and exchange rate policy coefficient is robust. The value and significance of B70, the current (WP/PT) coefficient, is not sensitive to the selection of prior beliefs; it varies from 0.45 to 0.54 and is always significantly different from zero. The inference obtained from B60, the absorption current coefficient, is weak. Although B60 is positive in the three posterior estimates, its value fluctuates from 0.06 to 0.17 and is significantly different from zero only in the posterior 2 estimates.

Duration of the Effects on Relative Prices

The estimation of interim multipliers using classical and Bayesian estimation methods allows one to conclude that, whereas a sustained change in absorption has a transitory effect on the relative price of nontraded goods, a sustained change in the ratio of public sector wages to exportable-goods prices tends to have a permanent effect. The values of the interim multipliers indicate that—regardless of the prior belief used—the effect of absorption on relative prices is completely eliminated over time. The interim multipliers for absorption (see Table 5) are equal to zero after one quarter using instrumental variables, after two quarters using posterior 1 and posterior 3 estimates, and equal to zero after four quarters using posterior 2 estimates. The interim multipliers for the ratio of public sector wages to exportable-goods prices (Table 6) are positive even in the long run using instrumental-variable estimates and any of the three posterior estimates. In addition, the 95 percent confidence intervals for the sum of current and lagged (WP/PT) coefficients, constructed using posterior 1, 2. and 3 estimates, all lie in the positive quadrant.27

Table 5.Semireduced-Form Interim Multipliers of Absorption (E/P)

(Dependent variable: (PN/PT))

InstrumentalPosterior Estimatesa
ItemVariablesPrior 1Prior 2Prior 3
Sum of current and lagged absorption coefficientsb-0.09-0.060.00-0.03
95 percent confidence intervalsc(0.05,-0.23)(0.07,-0.18)(0.13,-0.12)(0.08,-0.15)
Interim multiplier after j periods [M(h, j)]d
Periods (j)
0-0.010.110.170.05
1-0.050.170.250.10
2-0.060.000.110.01
3-0.12-0.060.05-0.03
4-0.15-0.090.02-0.04
7-0.17-0.100.00-0.06
-0.18-0.100.00-0.06

Each posterior estimate was obtained using as inputs the instrumental-variables estimates, the covariance matrix of those estimates, and the corresponding prior distributions; the computations were performed using the program SEARCH (Learner and Leonard (1983b)).

The sum of current and lagged coefficients of (E/P) in the (Pn/Pn) semireduced-form equation.

Confidence interval for the sum of coefficients. Obtained from the t distribution and the posterior estimates.

M (h,j) measures the effect, after j periods, of a once-and-for-all change in variable h on (Pn/PT).

Each posterior estimate was obtained using as inputs the instrumental-variables estimates, the covariance matrix of those estimates, and the corresponding prior distributions; the computations were performed using the program SEARCH (Learner and Leonard (1983b)).

The sum of current and lagged coefficients of (E/P) in the (Pn/Pn) semireduced-form equation.

Confidence interval for the sum of coefficients. Obtained from the t distribution and the posterior estimates.

M (h,j) measures the effect, after j periods, of a once-and-for-all change in variable h on (Pn/PT).

Table 6.Semireduced-Form Interim Multipliers of the Wage and Exchange Rate Policy Variable, (WP/PT)

(Dependent variable: (Pn/PT)

InstrumentalPosterior Estimatesa
ItemVariablesPrior 1Prior 2Prior 3
Sum of current and
lagged wage-
exchange rate
policy variable
coefficientsb0.270.450.380.27
Confidence
intervals of
95 percentc(0.61,-0.07)(0.67,0.23)(0.57,0.20)(0.43,0.11)
Interim multiplier after j periods [M(h, j)]d
Periods (j)
00.540.540.450.46
10.650.680.630.60
20.600.680.660.55
30.580.680.670.52
40.560.680.680.51
70.550.680.680.50
0.550.680.680.50

Each posterior estimate was obtained using as inputs the instrumental-variables estimates, the covariance matrix of those estimates, and the corresponding prior distributions; the computations were performed using the program SEARCH (Learner and Leonard (1983b)).

The sum of current and lagged coefficients of (WP/PT) in the (PN/PT) semi-reduced-form equation.

Confidence interval for the sum of coefficients. Obtained from the t distribution and the posterior estimates.

M(h,j) measures the effect, after j periods, of a once-and-for-all change in variable h on (PN/PT).

Each posterior estimate was obtained using as inputs the instrumental-variables estimates, the covariance matrix of those estimates, and the corresponding prior distributions; the computations were performed using the program SEARCH (Learner and Leonard (1983b)).

The sum of current and lagged coefficients of (WP/PT) in the (PN/PT) semi-reduced-form equation.

Confidence interval for the sum of coefficients. Obtained from the t distribution and the posterior estimates.

M(h,j) measures the effect, after j periods, of a once-and-for-all change in variable h on (PN/PT).

The interim multiplier of (WP/PT) represents the medium-and long-run effects of an expenditure-switching devaluation on relative prices (with the opposite sign) when nominal wages in the public sector are rigid.28 If public sector wages are indexed to the price level, then the impact of a devaluation on relative prices is eroded over time by the increase in wages and nontraded-goods prices. The larger the portion of the impact eliminated, the more complete is the indexation of public wages; in the extreme, when real public sector wages are rigid (γ= 1), the effect of an expenditure-switching devaluation on relative prices would be completely eliminated in the long run (see Table 7). Even in this case, however, the effect of the devaluation on relative prices would last for three quarters, according to instrumental-variables and posterior 3 estimates, and for six or seven quarters, according to posterior 1 and posterior 2 estimates.

Table 7.Semireduced-Form Interim Multipliers of an Expenditure-Switching Devaluation Under Indexed Public Sector Wages

(Dependent variable: (PN/PT))

InstrumentalPosterior Estimatesa
ItemVariablesPrior 1Prior 2Prior 3
Long-run multiplier [B7 [(γ,θN)]b
Index (γ)
0.75-0.18-0.23-0.23-0.16
0.90-0.07-0.100.10-0.07
Variablec-0.14-0.20-0.20-0.13
Interim multiplier after j periodsd
(indexation coefficient γ = 1)
Periods (j)
0-0.54-0.54-0.45-0.46
1-0.26-0.29-0.28-0.24
2-0.05-0.12-0.13-0.04
30.00-0.05-0.070.00
40.00-0.02-0.040.00
70.000.00-0.010.00
0.000.000.000.00

Each posterior estimate was obtained using as inputs the instrumental-variables estimates, the covariance matrix of those estimates, and the corresponding prior distributions; the computations were performed using the program SEARCH (Learner and Leonard (1983b)).

B7(γ,θN) = (1 - γ)/(l - γθNB7), where B7 is the long-run multiplier of a devaluation with rigid nominal wages, and θN is the nontraded-goods share in the CPI and is equal to 0.5.

The variable γ is equal to 0.75 the first quarter after the devaluation, to 0.9 the second, and to 1 afterward.

Obtained simulating the effects of an infinitesimal change in PT on (PN/PT). Public sector wages are fully indexed to last-period inflation:WP^=P^T(t1)+θN(PN/^PT)(t1).

Each posterior estimate was obtained using as inputs the instrumental-variables estimates, the covariance matrix of those estimates, and the corresponding prior distributions; the computations were performed using the program SEARCH (Learner and Leonard (1983b)).

B7(γ,θN) = (1 - γ)/(l - γθNB7), where B7 is the long-run multiplier of a devaluation with rigid nominal wages, and θN is the nontraded-goods share in the CPI and is equal to 0.5.

The variable γ is equal to 0.75 the first quarter after the devaluation, to 0.9 the second, and to 1 afterward.

Obtained simulating the effects of an infinitesimal change in PT on (PN/PT). Public sector wages are fully indexed to last-period inflation:WP^=P^T(t1)+θN(PN/^PT)(t1).

If the indexation is incomplete (γ < 1), an expenditure-switching devaluation has long-run effects on relative prices according to the four different estimates used (see Table 7). A simulation performed assuming a variable degree of indexation indicated that the long-run multiplier of an expenditure-switching devaluation is 0.14, according to instrumental variables and posterior 3 estimates, and to 0.20 according to posterior 1 and posterior 2 estimates.29

III. Concluding Remarks

The answers to the three empirical questions stated at the beginning of the paper can be summarized as follows: (1) The inferences regarding the effects of absorption on relative prices are not robust; however, in general, the effects tend to be weak and short-lived. (2) Devaluation can have a significant direct effect on the relative price of nontraded goods, (3) The effect of the exchange rate on relative prices is permanent if public sector wages are rigid in nominal terms or if indexation is partial, and the effect lasts for three to seven quarters, depending on the estimates used, when wages are fully indexed to the consumer price index.

The interim multipliers derived from the estimates indicate that a sustained change in the public sector wages relative to the price of exportables has a persistent effect on relative prices. A similar result was obtained using the four alternative estimates considered. The effectiveness of a devaluation, however, is reduced when public sector wages are adjusted after the devaluation. An expenditure-switching devaluation still has long-run effects on relative prices when the indexation of public sector wages is not complete, but the effect of a devaluation lasts only for three to six quarters when public sector wages are fully indexed to the price level of the previous quarter. This is consistent with the findings of most studies on the relationship between devaluation and real exchange rates.

In addition to the estimation of a semireduced-form equation under different prior beliefs, a test of the alternative theories was performed using ordinary least-squares estimates of the reduced form. The restrictions on the reduced-form coefficients implied by the different theories were tested using standard F tests. The results allow us to reject the restrictions implied by the specific-factors and PPP theories and to favor the rigid-wages solution, which allows for direct effects of devaluation on relative prices.

The main policy implication of the results obtained is twofold. First, demand management alone is not an effective tool to control relative prices of nontraded goods; the exchange rate policy needs to be used for that purpose. Second, devaluations and incomes policies, particularly public sector wage adjustments, should be consistently managed to avoid creating deviations of the relative price of nontraded goods from its target value.

The empirical results cannot be interpreted as a rejection of the effects that changes in international capital flows can have on the relative price of nontraded goods. Changes in net capital flows affect the relative prices of nontraded goods through changes in absorption—as the current account deficit is widened—and they also accelerate the accumulation of international reserves, allowing the authorities to pay less attention to exchange rate policy. Consequently, capital inflows could have been a major factor behind the exchange rate policy and the real appreciation observed in Chile in the late 1970s and early 1980s.

APPENDIX I

Derivation of the Model

This Appendix provides alternative solutions for the model, as well as a derivation of the interim multipliers.

Dependent Economy Model

The principal elements of the dependent economy model are given in the following subsections.

Zero-Profit Conditions

The prices of goods and factors obey the zero-profit condition. In proportional rate form,

where ^X represents the proportional change of variable X and the subindex i represents the sector of production.

Solving equation (17) for the proportional change in the relative factor rewards in sector i, where (ri/^w)=(r^iw^), one obtains

Labor Demand

Assuming that the production functions show constant elasticity of substitution and applying the proportional change operator to the first-order condition of cost minimization yields

where σi is the elasticity of substitution among factors (Li, and Ki) in sector i.

Replacing equation (18) in equation (19) to eliminate the rental rates and solving for the demand for labor in sector i (Li) yields

Using the exportable good (T) as numeraire and replacing in

gives

The proportional change in the aggregate demand for labor (L) is the weighted sum of equation (20a). The weights are the shares of each sector’s employment in total employment of labor (λi).

Assuming that (̂KiK) and (̂CTCM=ĉCN,c>1), then

Where Yi, represents the contribution of sector i to the wage elasticity of the aggregate labor demand (D), and YC is the elasticity of labor demand with respect to technical change in the traded-goods sector (^CT), then

with i = T, N, M.

Supply of Nontraded Goods

The proportional change in output of good N can be approximated by

The supply function of N is obtained by replacing equation (20), the sectoral demand for labor, in equation (22):

where fN represents the supply elasticity of nontraded goods with respect to the product wage in sector N.

The Demand Side and the Market of Nontraded Goods

The demand for goods in each period is obtained as the solution of the following optimization problem:

where E represents absorption. Consider the following utility function:

The first-order conditions expressed in terms of proportional change are @@Display Equation@@(24)

and

P is the rate of change in the consumer price index:

The demand for nontraded goods (QdN) is

where g,gN, and gM are the income, price, and cross-price elasticities of the demand for nontraded goods:

where

The market-clearing relative price of nontraded goods is obtained from equations (23) and (28):

Alternative Solutions of the Model

Three alternative solutions are given in the following subsections.

Solution with Rigid Wages and Specific Factors

If it is assumed that wages are rigid, then

The rigid-wages solution is obtained from the system formed by equations (21), (29), and (30). The three endogenous variables are L, (PN/PT). and (w/PT). The solution for (PN/PT) yields

where

Solution with Specific Factors and Flexible Wages

The wage rate is obtained by equating aggregate labor demand, equation (21), and Ls, assuming that wages are flexible:

The specific-factors solution is derived from the two-equation system formed by the market-clearing conditions in the labor market and in the nontraded-goods market, equations (32) and (29). The endogenous variables are (PN/PT) and (w/PT). The solution for (PN/PN/PT) yields

where

Solution with Flexible Wages and Mobile Factors

If flexible wages and mobile factors are assumed, the price of nontraded goods can be solved from the three-equation system formed by the zero-profit conditions expressed in equation (17). The three endogenous variables are PN, w, and r the solution for (PN/PT) yields

Interim Multipliers

Equation (35) represents the unrestricted semireduced-form equation for the relative prices of nontraded goods. X is the matrix of explanatory variables, B is the column vector of the corresponding coefficients (B′ is B transposed), and μ is the vector of residuals:

where

The matrix of explanatory variables X is partitioned into lagged endogenous variables (PN/PT)(t − 1); contemporaneous explanatory variables x (h,t); and lagged explanatory variables x(h, t−1),…,x(h, tm) where x(h, tm) represents the value of variable xh lagged for m periods.

The final form in equation (36) was derived by solving the semireduced form as a function of the value of the lagged endogenous variable in period 0 and of the value of the lagged explanatory variables x(h,t-j). Where x (h,tj) represents the value of the explanatory variable xh in the period tj,

The interim multipliers in equation (37) represent the effect on the relative price of nontraded goods of period t of a once-and-for-all change in variable xh that took place at time (t-j) Interim multipliers M (h,j) are obtained by deriving the final form and reordering terms:

The long-run multiplier of a sustained change in variable xh is presented in equation (38):

The long-run multiplier (B7) of (WP/PT) assumes a once-and-for-all change in the variable (WP/PT) and thus that nominal public sector wages are rigid. Allowing for indexation of public sector wages to the consumer price index to be lagged for one period gives

where WP (t) is the rate of change in nominal public sector wages in period t, γ is the indexation coefficient (0 < γ < 1). and θN is the share of nontraded goods in the price index (P). The long-run multiplier of a devaluation when public sector wages are indexed, B7-(γθN) is

APPENDIX II

Data Definitions and Sources

(Pn/PT) is the relative price of nontraded and exportable goods.

PN, is the nontraded-goods price index, constructed using selected prices from the consumer price index (source: Le Fort and Gillet (1986)).

PT is the domestic price of exportable goods, with PT=PT*e (zT is assumed to be equal to zero).

The symbol e denotes the nominal exchange rate, expressed as the value in Chilean pesos of a U.S. dollar (source: Banco Central de Chile, Boletín Mensual).

P*T the weighted average of trading partners’ wholesale price indices (source: Le Fort (1985)).

(E/P) is real absorption, private consumption, and investment plus government expenditure. Annual data were obtained from Banco Central de Chile, Cuentas Nacionales. Source for quarterly interpolation: Moran, Gutierrez, and Friedman (1983). The data were deseasonalized using the Jorgenson (1964) method.

(PT/PM) is the domestic price ratio of exportable and importable goods (source: Corbo (1983)). The series was completed using information contained in Moran, Gutierrez, and Friedman (1983).

Tec is a proxy for technology, constructed from a trend of the ratio of GDP to employment (source: Le Fort (1985)).

K denotes the stock of physical capital (source for annual totals: Banco Central de Chile, Indicadores Econónicos). The quarterly interpolation was performed using powers of a time-trend variable (source: Le Fort (1985)).

Ls represents the labor force in Santiago (source: Departamento de Econormía, Universidad de Chile, Encuesta de Ocupación).

(WP/PT) is the ratio of public sector wages to prices of exponahle goods.

WP is the index of wages in the public sector (source: Banco Central de Chile, Boletín Mensual).

The symbol ir* denotes the ex ante external real interest rate, with ir*= i* -E(^P*).

The symbol i* represents the quarterly average London interbank offered rate (LIBOR) in U.S. dollars for 180-day deposits (source: International Monetary Fund, International Financial Statistics (IFS)).

E(^P*) is expected international inflation, constructed using the rate of change of the international price index (P*^T) and four lags of the same variable (source: Le Fort (1985)).

E(^e) is the expected devaluation rate, E(e^)=p(ea^)+(1p)(ec^), where ea^ is the devaluation rate announced by the economic authorities, ec the devaluation rate conditional on the departure from the announced policy, and p the probability assigned to the first event (source: Le Fort and Ross (1985)).

E(^PT) is the expected rate of change of exportable-goods prices, E(P^T)=E(PT*^)+E(e^).

E(PN/^PT) is the expected rate of change in the price of nontraded goods relative to exportable goods, obtained from E(PN/PT)(t)/X(t − 1).

E(PN/PT)(t)/X (t − 1) is the one-period-ahead expected relative price of non-traded goods, obtained from a linear projection of the relative price in the exogenous variables of the model. The information set [ + X (t − 1)] included the value of all the variables in the previous quarter (see Barro (1978); source: Le Fort (1985)).

H/P is the real monetary base.

H is the monetary base (source: Moran, Gutierrez, Friedman (1983)).

P is the corrected consumer price index (source: Cortazar and Marshall (1980)).

DN is the index of the value of real domestic assets in units of consumption goods, DN = (SPK)/P.

SP is the index of stock prices (source: Banco Central de Chile, Boletín Mensual).

F is the value of the net foreign debt in units of consumption goods, F = FDe/P.

FD is the net foreign debt (private and public) in U.S. dollars (source: Banco Central de Chile. Indicadores Económicos). Quarterly interpolation was performed using the method of Denton (1971) (source: Le Fort (1986)).

FS is balance of payments net financial services in units of consumption goods, FS = FSBOPe/P.

FSBOP is balance of payments net financial services. Annual data were obtained from IFS. Quarterly interpolation was performed using full-information, maximum-likelihood (FIML) technique and Denton’s (1971) method (source: Le Fort (1986)).

RDC represents policy shocks to real base money, RDC = DDC/H(t − 1) − ^PT.

DDC is the change in nominal domestic credit (source: Banco Central de Chile, Sintesis Monetaria y Financiers).

G is government consumption (source: Moran, Gutierrez. Friedman (1983)), deseasonalized.

Trib denotes real tax revenue.

lp is the index of international creditworthiness, Ip = i*(l − pd)lpd.

The symbol pd denotes the probability of debt default, pd − exp(EC)/ (1 +exp(EC).

EC is the expected cost of default for the borrower, EC = 0.314 + 0.633 FD/Y − 1.152 RES/Y − 1.186 Y + 0.353 BOPFS/X. The coefficients were obtained from Edwards (1984).

Y is GDP in U.S. dollars (source: Banco Central de Chile, National Accounts).

RES denotes international reserves (source: IFS).

EX represents exports (source: Banco Central de Chile, Boletín Mensual).

CCA is a qualitative variable for the capital account liberalization. CCA is equal to zero for the period extended from the first quarter of 1974 to the first quarter of 1980 and equal to unity thereafter.

REFERENCES

    BalassaBela“The Purchasing-Power-Parity Doctrine: A Reappraisal,”Journal of Political Economy (Chicago) Vol. 72 (December1964). pp. 58496.

    BarroRobert J.“Unanticipated Money, Output, and the Price Level in the United States,”Journal of Political Economy (Chicago) Vol. 86 (August1978) pp. 54980.

    BrunoMichael“Exchange Rates, Import Costs, and Wage Price-Dynamics,”Journal of Political Economy (Chicago). Vol. 86 (June1978). pp. 379403.

    CalvoGuillermo“Real Exchange Rate Dynamics with Fixed Nominal Parities: On the Economics of Overshooting and Interest-Rate Management with Rational Price Setting,”Working Paper 5 International Economics Research Center (New York: Columbia UniversityAugust1982).

    CorboVittorio“An Overview of the Macro economic Development in the Last 20 Years in Chile” (unpublished; Santiago, Chile: Universidad Catolica1983).

    CorboVittorio (1985a) “Reforms and Macroeconomic Adjustments in Chile During 1974–84,”World Development (Oxford) Vol. 13 (August1985) pp. 893916.

    CorboVittorio(1985b) “International Prices, Wages and Inflation in an Open Economy: A Chilean Model,”Review of Economics and Statistics (Cambridge, Massachusetts) Vol. 67 (November1985) pp. 56473.

    CorboVittorio and Jaimede Melo“What Went Wrong with the Recent Reforms in the Southern Cone,”Economic Development and Cultural Change (Chicago) Vol. 34 (April1986) pp. 60740.

    Corden W.Max and J. PeterNeary“Booming Sector and De-Industrialization in a Small Open Economy.”Economic Journal (London) Vol. 92 (December1982) pp. 82548.

    CortazarRene“Salarios nominales e inflacion: Chile 1974–1982,”Estúdios C1EPLAN (Santiago) Vol. 11 (December1983) pp. 85111.

    CortazarRene and J.Marshall“Indice de precios al consumidor en Chile: 1970–78,”Estúdios CIEPLAN (Santiago) Vol. 4 (November1980).

    DentonFrank T.“Adjustment of Monthly or Quarterly Series to Annual Totals: An Approach Based on Quadratic Minimization,”Journal of the American Statistical Association (Washington) Vol. 66 (March1971) pp. 99102.

    Diaz AlejandroCarlos F.“Southern Cone Stabilization Plans” inEconomic Stabilization in Developing Countriesed. byWilliam R.Clineand SidneyWeintraub (Washington: The Brookings Institution1981) pp. 11947.

    DombuschRudiger“Devaluation, Money, and Nontraded Goods,”American Economic Review (Nashville, Tennessee) Vol. 63 (December1973) pp. 87180.

    DombuschRudiger“PPP Exchange-Rate Rules and Macroeconomic Stability,”Journal of Political Economy (Chicago) Vol. 90 (February1982) pp. 15865.

    EdwardsSebastian“Economic Policy and the Record of Economic Growth in Chile” (unpublished; Los Angeles: Department of Economics, University of California1983).

    EdwardsSebastian“LDC Foreign Borrowing and Default Risk: An Empirical Investigation (1976–1980),”American Economic Review (Nashville, Tennessee) Vol. 74 (September1984) pp. 72634.

    EdwardsSebastian“Monetarism in Chile, 1973–1983: Some Economic Puzzles,”Economic Development and Cultural Change (Chicago) Vol. 34 (April1986) pp. 53560.

    FoxleyAlejandroLatin America: Experiments in Neoconservative Economics (Berkeley: University of California Press1983).

    FrenkelJacob A.“The Collapse of Purchasing-Power Parities During the 1970s.”European Economic Review (Amsterdam). Vol. 16 (May1981) pp. 14565.

    HarbergerArnold“The Chilean Economy in the 1970s: Crisis. Stabilization, Liberalization, Reform,”Carnegie Rochester Conference Series on Public Policy (Amsterdam and New York) Vol. 17 (Autumn1982) pp. 11552.

    IntriligatorMichaelEconometric Models Techniques and Applications (Englewood Cliffs, New Jersey: Prentice Hall1978).

    JonesRonald“The Structure of Simple General Equilibrium Models,”Journal of Political Economy (Chicago) Vol. 73 (December1965) pp. 55772.

    JonesRonald (1974a) “The Small Country in a Many-Commodity World,”Australian Economic Papers (Adelaide) Vol. 13 (December1974) pp. 22536.

    JonesRonald (1974b) “Trade with Non-Traded Goods: The Anatomy of Interconnected Markets,”Economica (London) Vol. 41 (May1974) pp. 12138.

    JorgensonDale W.“Minimum Variance, Linear, Unbiased Seasonal Adjustment of Economic Time Series,”Journal of the American Statistical Association (Washington) Vol. 59 (September1964) pp. 681724.

    KomiyaRyutaro“Nontraded Goods and the Pure Theory of International Trade,”International Economic Review (Philadelphia) Vol. 8 (June1967) pp. 13252.

    LearnerEdwardSpecification Searches: Ad Hoc Inference with Nonexperimental Data (New York: Wiley1978).

    LearnerEdwardSources of International Comparative Advantage: Theory and Evidence (Cambridge, Massachusetts: MIT Press1984).

    LearnerEdwardand HermanLeonard (1983a) “Reporting the Fragility of Regression Estimates,”Review of Economics and Statistics (Cambridge, Massachusetts) Vol. 65 (May1983) pp. 30617.

    LearnerEdward(1983b) “SEARCH: A Software Package for Bayesian Inference and Sensitivity Analysis” (unpublished; Los Angeles: University of California1983).

    Le FortGuillermo R.“The Real Exchange Rate and Capital Inflows: The Case of the Southern Cone Countries” (doctoral dissertation; Los Angeles: University of California1985).

    Le FortGuillermo R.“Trimestralización de series de balanza de pagos,”Estúdios de Económia (Santiago) Vol. 13 (April1986) pp. 16789.

    Le FortGuillermo R. and R.Gillet“Indices de precios de bienes no transables: Chile 1974–84,”Estiidios de Economia (Santiago) Vol. 13 (April1986) pp. 7394.

    Le FortGuillermo R. and R.Gillet and C.Ross“La devaluatión esperada una aproximación Bayesiana: Chile 1974–84,”Working Paper 72 (Santiago, Chile: Departamento de Economia, Universidad de Chile1985).

    MoranCristianGuillermoGutierrez and JorgeFriedmanEstadísticas trimestrales de producto y dinero para la economia Chilena 1960–1981 (Santiago, Chile: Departamento de Economia, Universidad de Chile1983).

    RamosJosephEstabllización y liberalización económica en el Cono Sur (Santiago, Chile: CEPAL1984).

    SanyalKalyan and RonaldJones“The Theory of Trade in Middle Products,”American Economic Review (Nashville, Tennessee) Vol. 72 (March1982) pp. 1631.

    TaylorJohn“Staggered Wage Setting in a Macro Model,”American Economic Review (Nashville, Tennessee) Vol. 69 (May1979) pp. 10813.

    ZahlerRoberto“Recent Southern Cone Liberalization Reforms and Stabilization Policies: The Chilean Case (1974–82)” (unpublished; Santiago, Chile: CEPAL1983).

    ZahlerRoberto“Las tasas de interés en Chile: 1975–1982.” inEl desarrollo financiero de America Latina y el Caribe (Caracas: Instituto Interamericano de Mercados de Capital1985) pp. 571623.

Mr. Le Fort, an economist in the Research Department when this paper was written, is now in the Asian Department of the Fund. He holds degrees from the Universidad de Chile and the University of California, Los Angeles.

This paper is based in part on the author’s Ph.D. dissertation (Le Fort (1985)). He is indebted to Edward Learner, Axel Leijonhufvud, and Sebastian Edwards for their guidance, and to his colleagues in the Fund for their valuable comments.

Similar structural reforms, ending in recessions and crises, were also attempted in Argentina and Uruguay. The experiences of these three countries, which have been analyzed at length in recent economic literature, provide important lessons on problems and policy mistakes during the process of economic liberalization.

Nontraded goods are goods that under the current technology, tastes, trade restrictions, transport costs, and international prices cannot be traded internationally, and thus their prices are determined domestically. Traded goods are goods that are actually exported or imported and their close substitutes produced domestically; their prices in international currency are determined abroad. The term “relative price of nontraded goods” refers to the price of nontraded goods relative to the price of exportable goods and can be interpreted as the real exchange rate.

The structural changes and the stabilization policy attempted in the Chilean economy during the late 1970s, as well as the results obtained, have attracted the interest of economists. Among the many studies produced are those by Corbo (1983, 1985a, 1985b); Corbo and de Melo (1986); Cortazar (1983); Diaz Alejandro (1981); Edwards (1983, 1986); Foxley (1983); Harberger (1982); Ramos (1984); and Zahler (1983).

Given the high inflation rate in the Chilean economy during most of the period analyzed, and that the direct effect of the exchange rate policy on relative prices takes place through real producer wages, the level of public sector wages relative to exportable goods prices was taken to represent the direct effect of the exchange rate policy on relative prices.

The posterior estimates were obtained using a program (SEARCH, Learner and Leonard (1983b)) that derives a contract curve between the prior beliefs and the least-squares estimates. The least-squares estimates were represented in this case by instrumental-variables estimates, and the prior beliefs were derived from each of the alternative theories.

The specification of prior beliefs, used in Section II below, is much easier when the parameters are elasticities.

This simplifying assumption is often used in the literature. According to the Balassa (1964) effect, technical change increases the price of nontraded goods relative to traded goods because technical progress is slower in the former sector. Limitations of the data prevent the use of a more general approach.

The implications of the sticky-wages model for the relative price of nontraded goods would not be different with intersectoral capital mobility.

A complete list of parameters and variables of the model is presented in Table 2.

A devaluation accompanied by an increase in public sector wages of the same proportion would not have any real effect according to this model. This restriction is consistent with rationality and helps to avoid estimation problems—that is. multicollinearity—that arise with the use of nominal explanatory variables in an inflationary environment.

It is also assumed that the rate of technical change is equal in the importables and exportables sectors and slower in the nontraded-goods sector, and that the rate of capital accumulation is the same in all the sectors. The derivation of the labor demand is presented in Appendix I.

In an open economy that can accumulate international assets, the domestic demand for goods in each particular period is restricted by real absorption, the sum of income, and the current account deficit. In this paper, the determination of absorption through intertemporal optimization is not modeled; explicit modeling is included in Le Fort (1985, Chapter 3).

E is nominal absorption; (E/P), real absorption; and P, the relevant price index. In the nontraded-goods market, fN is the supply price elasticity; gN, the demand price elasticity; gM, the demand cross-price elasticity; and g, the demand income elasticity.

A model with the same type of explanatory variables can be obtained with perfectly mobile factors if it is assumed that there are more factors than internationally traded goods. In the literature, several macroeconomic models of this type—that is, relating absorption and the price of nontraded goods—have been developed without explicit microeconomic foundations. See, for example, Dornbusch (1973).

The term (gN + fNfNYN) represents the price elasticity of the excess supply of nontraded goods after the indirect effects of wages (fNYN) have been taken into account.

The assumptions of the PPP solution are, in general, not met in the short run; PPP implications have been rejected empirically for developed countries with flexible exchange rates (see Frenkel (1981)). An empirical question that is still open, however, is the length of the period needed for PPP conditions to hold.

See Balassa (1964). These results refer to partial effects of technical progress, assuming that the international prices of importables and exportables are constant. Technical change could reduce the relative price of nontraded goods even when the progress is faster in both traded-good s sectors than in the nontraded-goods sector, provided that the technical change in the capital-intensive sector (M) is faster than in sector T by an amount large enough to compensate for the Balassa effect.

̂Tec=(1−αNN/CCT; ̂Tec is a proxy for technical change obtained from the economy-wide average labor productivity, αN represents the share of non-traded goods in GDP, ĈT is technical change in the ex portable-goods sector, and ĈT/c is technical change in the nontraded-goods sector. Equation (13) was estimated in the level of variables, with the values of the variables normalized to unity in the fourth quarter of 1977. The parameters represent elasticities around the value of the variables in that particular period.

This specification was selected because the rate-of-change linear model and the log-linear model indicated autocorrelation of the residuals that could not be corrected using a standard first-order autoregressive procedure. The results for the estimation that is linear in the variables did not show evidence of first-order autocorrelation.

To avoid excessive extension of the paper and a deviation from its main line of exposition, a structural equation for absorption is not specified, and only the set of instrumental variables used is presented.

The expected rate of change in the price of traded goods is equal to the sum of expected international inflation and expected currency devaluation. Expected international inflation was assumed to be formed adaptively. The expected rate of devaluation was constructed using a “peso problem” approach in which two events are possible for the exchange rate policy in each period: either the rate of devaluation is equal to the preannounced rate, or a major devaluation takes place. The probability of a policy break was obtained from a signal given by the change in international reserves and subjective information based on previous actions of the authorities.

The expected rate of change in the relative price of nontraded goods was obtained by assuming rational expectations and an information set limited to the value of the variables up to the last period. A linear projection of all the variables lagged one period was used to estimate the expected relative price for the next period; the expected rate of change of relative prices was calculated from that value.

The international creditworthiness index was constructed using variables that can affect the benefits and costs of debt default, including the debt-to-GDP ratio, the financial services-to-exports ratio, and the invcstment-to-GDP ratio, among others. The variables and coefficients were obtained from Edwards (1984).

Additional details and the information sources are presented in Appendix II.

For a theoretical presentation, see Learner (1978); for a practical application, see Learner (1984) and Learner and Leonard (1983a).

The posterior estimates were computed using the program SEARCH (Learner and Leonard (1983b)). Table 4 presents statistics for the residuals in the semireduced-form estimations, including the coefficient of determination corrected by degrees of freedom (¯R2), the sum of squared errors (ESS), the Durbin-Watson statistic (DW), and the Durbin h test.

The sign of the long-run multiplier is given by the sum of the current and lagged (WP/PT) coefficients.

A devaluation increases the denominator of (WP/PT). The interim multipliers were developed for a once-and-for-all change in the explanatory variable and consequently, in this case, for a constant WP.

The indexation coefficient was assumed to equal 0.75 in the quarter immediately following the devaluation, 0.9 in the second, and unity in the third and following quarters. Nontraded goods represent 50 percent of the price index weights. The simulation was performed assuming that public sector wages are indexed to inflation in the preceding quarter.

Other Resources Citing This Publication