Journal Issue

Mexico: Selected Issues

International Monetary Fund
Published Date:
August 2004
  • ShareShare
Show Summary Details

III. A Model of Inflation in Mexico1,2

This chapter examines the determinants of inflation in Mexico since 1997. A model of wage and price inflation is formulated using an expectations-augmented Phillips curve framework with both forward- and backward-looking inflation components, proxies for excess demand, and relative price shocks. The model is estimated using data from 1997 to mid-2003. Wage dynamics appear to be driven mostly by inflation expectations, lagged inflation and to a lesser extent the output gap. Price inflation is in turn driven by wage inflation and price shocks, including to administered prices and the exchange rate. The similarity of the inflation process in Mexico to that in countries with a longer history of low and stable inflation suggests that the disinflation strategy pursued since the mid-1990s has provided a credible nominal anchor.

A. Overview

59. Previous studies of inflation in Latin America have shown that episodes of high inflation tend to be associated with the monetization of large fiscal deficits, aggravated by exchange rate volatility and extensive wage indexation. To a lesser extent, these phenomena were characteristic of Mexico in the 1980s. With Mexican inflation declining to much lower levels in recent years, a fresh look at the inflation process is in order to examine whether it has become more similar to those in countries with a longer experience of relatively low and stable inflation.

60. Unsurprisingly, given Mexico’s previous history of high and variable inflation in the 1980s to mid-1990s, some observers initially discounted the decline in inflation after 1995 as transitory. Disinflation was, however, maintained through a long economic expansion (1996–2000) and in the presence of various types of inflationary shocks. Certainly, a shift to inflation targeting has played a role in preserving this relatively stable outcome, but it also appears that some types of shocks that previously had a strong influence on inflation now have less influence.3 For example, the experience of small open economies in Latin America and elsewhere has been that large currency depreciations usually generate a substantial increase in inflation. In recent years, however, despite sharp depreciations of the peso and rising import prices, growth in retail prices in Mexico has remained subdued, with only a limited effect on consumer price inflation. The apparent reduction in exchange rate pass-through provides an example of a direct and visible change in pricing behavior that has had a bearing on recent inflation performance.4

61. At the same time, the program of market liberalization that began with NAFTA entry in 1994, and extended from product and capital markets to labor markets, has had important consequences for price behavior. Domestic competition has increased, centralized wage setting has been replaced by a decentralized system, and trend productivity growth has risen. Each of these developments is conducive to achieving lower and more stable inflation, at least in the short to medium term, and may have reinforced the effects of reduced exchange rate pass-through. Moreover, these changes have occurred against the background of more disciplined policy formation in Mexico. Macroeconomic management since the 1994–95 crisis has focused on achieving medium-term price stability, lower fiscal deficits, prudent debt management to reduce vulnerabilities, implementation of structural reforms, and integration with NAFTA. In this regard, an important step undertaken after the collapse of the Mexican peso at end-1994 was to focus monetary policy on attaining medium-term price stability, and to address banking sector problems by specific programs, with the cost to be assumed by the fiscal authority.

62. Monetary policy has subsequently evolved into an inflation targeting framework in the context of a flexible exchange rate system. Importantly, the inflation objective has now become the nominal anchor for the economy.

63. We analyze wage and inflation dynamics in Mexico for the period 1997–2003. The framework consists of two relationships. The first is an augmented Phillips curve that relates contractual wage inflation to a measure of the output gap, inflation expectations, and lagged inflation. The domestic price level is set as a mark-up over production costs, with the latter represented by wages and import prices. The mark-up varies according to the state of aggregate demand. Finally, changes in “administered” prices are assumed to affect the aggregate price level.

64. The chapter is organized as follows. First, we present descriptive statistics on inflation, showing how the variability of inflation and its determinants has changed over time. Second, we briefly discuss developments in the main determinants of inflation in Mexico. Third, we develop an empirical model of wage and price inflation and estimate it for the period 1997–2003. The estimation period is then truncated and the model is used to generate out-of-sample inflation forecasts. Policy implications are discussed in the final section.

B. Inflation and its Determinants: Stylized Facts

65. Mexico’s inflation performance over the past 2½ decades is illustrated in Figure 1. Two measures of inflation are shown—core and headline. Table 1 presents descriptive statistics for inflation and its determinants since 1981. Several interesting features are apparent from these data:

  • Inflation has been highly variable over the long run. Average inflation declined from about 75 percent in 1983–89 to 5¾ percent in 2002. High inflation in the 1980s was associated with the monetization of large and unsustainable fiscal deficits, aggravated by frequent devaluations of the exchange rate and a high degree of wage indexation.
  • Inflation fell in the early 1990s with exchange-rate-based stabilization, but increased in 1995 in the face of the economic crisis.
  • The authorities responded to the 1994–95 crisis by implementing a stabilization plan centered on reducing the fiscal deficit and orienting monetary policy toward disinflation in the context of a flexible exchange rate regime.
  • The authorities’ disinflation program was consolidated with the introduction of inflation targeting in 2001. Subsequently, inflation has been maintained below target, with the exception of 2002, when increases in administered prices caused a temporary rise in inflation.
  • The data show the increasing “resilience,” or lack of lasting response, of inflation to external and domestic shocks. Table 1 and Figure 2 compare the variability of inflation with that of various shocks. The standard deviation of inflation has fallen significantly since the 1980s, while the standard deviations of the exchange rate, wages, and administered and agricultural prices have fallen by less. As a result, the ratio of the variability of the shocks to that of inflation in the 2000–03 period generally rose relative to the 1980s, with the change being particularly notable in the case of the exchange rate and administered prices.

Figure 1.Mexico: Headline and Core Inflation, 1983–2003

(In percent)

Source: Bank of Mexico.

Table 1.Mexico: The Variability of Inflation and its Determinants, 1981-2003


deviation relative to

that of Inflation
1981:1 -1989:12
Headline inflation74.741.31.0
Contractual wages
Exchange rate77.859.81.4
Import prices3.33.10.1
Output gap-
Administered prices90.664.21.6
Agricultural prices84.636.80.9
Inflation expectations
1990:1 -1995:12
Headline inflation19.411.21.0
Contractual wages
Exchange rate20.532.12.9
Import prices2.11.90.2
Output gap-
Administered prices23.914.01.2
Agricultural prices14.011.01.0
Inflation expectations
1996:1 -1999:12
Headline inflation22.29.11.0
Contractual wages18.51.70.2
Exchange rate11.110.51.2
Import prices-
Output gap0.21.00.1
Administered prices23.910.21.1
Agricultural prices25.014.61.6
Inflation expectations14.11.90.2
2000:1 -2003:6
Headline inflation6.72.01.0
Contractual wages8.73.01.5
Exchange rate2.57.13.6
Import prices1.31.80.9
Output gap-
Administered prices12.07.13.6
Agricultural prices4.53.41.7
Inflation expectations6.22.11.0
Source: Bank of Mexico; and staff estimates.
Source: Bank of Mexico; and staff estimates.

Figure 2.Mexico: Ratio of Inflation and its Determinants, 1996-2003

(Ratio of Standard Deviations)

Sources: Bank of Mexico; and Fund staff estimates.

C. Determinants of Inflation in Mexico: Recent Developments

66. We focus in this section on developments in some of the key influences on inflations in Mexico. The framework described above suggests a set of domestic and foreign inputs that are relevant to the inflation process in Mexico. These include wages and productivity, imported inputs to production, the exchange rate, and relative price shocks.

67. A key influence on inflation is the rate of wage inflation. For many years, labor unions were important in setting working conditions in Mexico for some sectors of the labor force.5 Wages were to some extent centrally determined and indexed to the cost of living. This increased the likelihood of direct transmission of price shocks to wages and the potential for a wage-price spiral, as occurred in the early 1980s. However, beginning in the 1980s and through the 1990s, changes were implemented that significantly increased the flexibility of the labor market. The government implemented a restrictive income policy to control inflation in 1983. In December 1987, labor, business and government reached an agreement called the Pacto, in which they agreed to coordinate closely wage demands with a view to moderating union wage demands and increasing employment. Subsequently, the privatization of state enterprises and entry into NAFTA in the 1990s coincided with a decline in unionization, and the increased pressure on domestic firms to compete internationally compelled them to reduce labor costs by hiring part-time workers and sourcing out segments of the production process. Reduced unionization and more decentralized wage bargaining lowered the tendency for wages to be indexed to past inflation, thus removing the direct transmission of prices to wages.6

68. Inflation performance in the 1990s has also been affected by labor productivity growth. Trend growth in productivity has averaged about 2 percent since mid 1990s. At the same time, wage earners seem not to have captured all of the productivity gains in the form of higher real wages, which have grown at a lower rate than trend productivity (Table 1). This has led to lower growth in unit labor costs (ULC), which has contributed to disinflation.

69. Exchange rate movements can affect inflation through both direct and indirect channels. The direct channel reflects the impact of exchange rate movements on the domestic currency price of imports. The effect on final prices will depend on the extent to which importers and distributors change their profit margins in response (which may itself be endogenous to the process driving the exchange rate). In any event, except in the extreme case where changes in import prices are fully offset by changes in margins, some proportion of import price changes will be passed through to final prices.7

70. The indirect channel reflects the impact of the initial rise in final prices on other costs, including wages, and inflation expectations. The importance of these indirect effects is likely to depend on the process believed to driving the exchange rate, and in turn the underlying credibility of the monetary policy framework. To the extent that inflation is believed to be well-anchored by a credible policy commitment, the second-round impact of exchange rate changes is likely to be limited. Absent such a commitment, however, exchange rate changes may well be interpreted as signaling long-lived changes in prices, implying a greater effect on other prices and inflation expectations.

71. Changes in other relative prices can also influence inflation. In Mexico, important components of the overall CPI basket have prices that are set by various governmental organizations in ways that may not reflect the forces determining market-based prices. These are collectively referred to as “administered” prices, and consist primarily of energy prices and those set by other public utilities. These are incorporated in the inflation model via a term reflecting changes in administered prices relative to the overall CPI.8

D. Model Estimation

72. In this section, we analyze the inflation process in terms of a two-equation model consisting of an expectations-augmented Phillips curve for wages and a mark-up equation for prices. The model incorporates both backward- and forward-looking elements in the inflation process.9 It is derived from the following 2 equations. All variables are in logs, w stands for wages, p for the domestic price level, p* for foreign export prices, z for foreign prices expressed in domestic currency, e for the exchange rate (defined as the domestic currency price of foreign currency), ygap for the output gap, π is inflation and E is the expectations operator.

73. Equation (1) defines wage dynamics as a function of expected inflation, past inflation, and the output gap. Equation (2) indicates that consumer prices are a function of wages, foreign prices, the administered prices and the output gap. The derived Phillips curve will be vertical in the long run if we assume that the sum of the coefficients on expected and past inflation is equal to one in equation (1). This ensures that no long-run trade-off exists between the level of inflation and excess demand pressures.

74. As noted previously, we model consumer inflation as a mark-up over unit costs of production, which we assume to be determined by contractual wages and foreign prices, as well as movements in administered prices. Movements in foreign and administered prices are expressed in real terms by defining them as the deviation of the 12-month percentage changes in the variable from previous period inflation.10 To solve the model for inflation, we first difference equation (2) such that:

The final version of equation (3) for estimation is described in equation (4), where L is the lag operator:

75. We estimate wage dynamics and price inflation by ordinary least squares following a general-to-specific approach. All data except the output gap are obtained from the Bank of Mexico website. Potential output and the output gap are estimated by applying the Hodrick-Prescott filter to the monthly global economic activity index (IGAE) for Mexico obtained from INEGI.11 The gap measures the proportionate deviation of actual from potential output. The estimation period covers 1998:09–2003:06. Wages growth is measured as the increase implied by contractual wage settlements; overall inflation and the change in foreign prices are measured as the 12-month changes in the respective variable, while inflation expectations are the 12-month ahead definition based on BOM surveys. All parameters are expected to be positive. The overlapping observations in our measurement of the dependent variables are likely to induce moving average processes in the residuals—these are accounted for by explicitly introducing moving average terms in the estimation process.

76. The estimation results for the wage equation exhibit the expected properties. Table 2 shows that the weight on backward-looking inflation is about one third, while that on the forward-looking component is about two thirds. The relatively large weight on forwardlooking inflation suggests a comparatively low degree of inflationary inertia The fact that the coefficients on backward- and forward-looking inflation sum to roughly one also suggests that the long-run Phillips curve is vertical—i.e. that there is no long-run trade off between activity and inflation.12 The intercept term is positive and significant, proxying for the effect of productivity growth on real wage gains, while the sensitivity to the output gap is borderline significant at the 10 percent level. The regressors explain most of the movements in contractual wages, with an adjusted R-squared of about 0.96. In addition, the White test for Heteroskedasticity and Breusch-Godfrey test on the residuals allow rejection of the presence of heteroskedasticity and autocorrelation, respectively.

Table 2:Mexico: Estimation Results for Wages, 1998–2003
Output gap(-1)0.1311.3870.171
Inflation expectations0.6973.4340.001*
Lagged inflation0.3252.0060.050*
Adjusted R-squared0.957
Diagnostic tests
Jarque Bera test for normality of the residuals0.8680.648
Serial correlation First order-Durbin Watson1.971
Serial correlation:-Higher order-Breusch-Godfrey LM Test2.1310.037
White Heteroskedascity Test (F-Statistics)0.1360.907
Wald F-Test0.2640.609
Normalized restriction0.0330.523
Source: Staff estimates.

Significant at the 5 percent level.

P-values give the probability that the null hypothesis is accepted.

Source: Staff estimates.

Significant at the 5 percent level.

P-values give the probability that the null hypothesis is accepted.

77. The estimation results for the inflation equation are also as anticipated (Table 3). The coefficients on contractual wage increases indicate that domestic input costs dominate total production costs. The restriction that the sum of the coefficients on wage growth is unity is easily accepted, consistent with dynamic homogeneity of the wage-price process. The change in the output gap also helps to explain inflation, presumably capturing cyclical changes in mark ups over variable costs. Foreign prices affect domestic inflation with a “passthrough” coefficient of about 0.07, suggesting an impact that is typical of other economies with trade shares similar to Mexico that have enjoyed an extended period of low and stable inflation. Changes in administered prices also have a significant impact. The overall fit of the equation is high, with an R2 of 0.99.

Table 3.Mexico: Estimation Results for Inflation in Mexico, 1998–2003
Change in output gap0.0873.5030.001*
Wages (-1)0.40012.0910.000*
Wages (-2)0.29912.3320.000*
Foreign prices (-1)0.0734.0290.000*
Admiistered prices (-1)0.1222.4020.020*
MA (1)1.96017.1530.000*
MA (2)1.60913.3860.000*
Adjusted R-squared0.993
Diagnostic tests
Jarque Bera test for normality of the residuals1.0410.594
Serial correlation First order-Durbin Watson1.920
Serial correlation: Higher order-Breusch-Godfrey LM Test2.7810.009
White Heteroskedascity Test (F-Statistics)0.9860.477
Wald F-Test0.1470.703
Normalized restriction0.0290.383
Source: Staff estimates.Significant at the 5 percent level.P-values give the probability that the null hypothesis is accepted.
Source: Staff estimates.Significant at the 5 percent level.P-values give the probability that the null hypothesis is accepted.

78. The estimated equation also passes the usual diagnostic tests. The Breusch-Godfrey tests statistics for up to 6 lags and the White tests are below their critical values, indicating absence of higher-order serial correlation or heteroskedasticity, respectively.

E. Model Simulation

79. The two-equation model was then simulated to generate dynamic forecasts for the endogenous variables, based on the historical paths for the exogenous variables.13 The results are shown in Figure 3. The ± 2 standard errors of the actual and simulated outcomes for wages and inflation are shown in the graphs. The results illustrate how the model would have tracked if we had used it in 1999 to make a forecast for the economy through the sample period, assuming we had used the correct paths for the exogenous variables. The graphs show that as a one-step ahead predictor, the model performs quite well, although the ability of the model to predict wages, in particular, deteriorates during the last 12 months of the sample. In particular, the equation tends to overpredict wage growth since late 2002. This overprediction could reflect two factors: (i) an underestimation of the effect of the output gap on inflation, given that actual output was below potential over this period; or (ii) an overestimation of underlying inflation expectations based on the survey measure, which responded quite strongly to price shocks over this period that turned out to be temporary.

Figure 3.Dynamic Forecasts of Estimated Model, 1998-2003

Source: Fund staff estimates.

F. Conclusions

80. The paper provides evidence that the inflation process in Mexico is driven by factors characteristic of those in other countries. The degree of inflation persistence is relatively low and is dominated by the highly significant forward-looking inflation variable. Taking at face value, this implies that reducing inflation is less costly (in terms of output loss) than in countries with much higher inflation persistence.14 We find significant coefficients for all the key determinants of wage and price inflation included in our model.

81. Contractual wage setting plays a central role in the inflation process. The estimation of the wage equation exhibited the desired dynamic properties, with about one third of wage inflation explained by backward-looking expectations, while the forwardlooking component, thought to be increasingly important in the transition period after the changes in monetary policy and exchange rate regime explains two-thirds. The coefficients on contractual wages in the price equation indicate that domestic input costs dominate total production costs.

82. Of the other supply factor in the price equation, international prices appear significant, but appear to only have a small transitory effect on prices. It appears to be relatively unimportant compared with the other supply variables. One interpretation is that the disinflation process in Mexico since 1996 reduced the pricing power of firms and the degree of pass-through of exchange rate changes into inflation. The other supply factor, administered prices also play important roles in the inflation process. They tend to be more volatile due to the influence of weather and public pricing decisions.

83. Dynamic simulations have also been included in the paper to mimic the response of wages and prices. The model, while simple in structure, provides a useful framework to summarize and forecast wage and price inflation in Mexico. Future work could model explicitly the expectations process and include a policy reaction function to gauge the appropriate response of monetary policy to inflationary shocks.


    BallL.N GMankiw and D.Romer1988“The new-Keynesian economics and the outputinflation trade-off,”Brookings Papers on Economic Activity pp 182.

    • Search Google Scholar
    • Export Citation

    BallLaurence and N GregoryMankiw1995“Relative-price change as aggregate supply shocks,”Quarterly Journal of Economics vol. 110 pp 16293.

    • Search Google Scholar
    • Export Citation

    Bank of MexicoInflation Reports 1997–2003.

    BaqueiroA.Diaz-LeonATorresA.“Fear of Floating or Fear of Inflation? The Role of the Exchange Rate Pass-Through,”Bank of Mexico.

    • Search Google Scholar
    • Export Citation

    DwyerJ. and RamR.1994“Explaining Import Price Inflation: A Recent History of Second-Stage Pass-Through,”Reserve Bank of Australia Reseach Discussion Paper No. 9304.

    • Search Google Scholar
    • Export Citation

    FischerStanley1981“Relative shocks, relative price variability, and inflation,”Brookings Papers on Economic Activity No 2 pp 381441.

    • Search Google Scholar
    • Export Citation

    GruenD.A.Pagan and C.Thompson1999“The Phillips Curve in Australia,” Reserve Bank of AustraliaResearch Discussion Paper No 1999–01.

    • Search Google Scholar
    • Export Citation

    ChadhaB.MassonP.MeredithG.1992“Models of Inflation and the Costs of Disinflation,”Staff PapersInternational Monetary Fund Vol. 39 pp. 395431.

    • Search Google Scholar
    • Export Citation

    LaxtonD.MeredithG.RoseD.1995“Asymmetric Effects of Economic Activity on Inflation,”Staff PapersInternational Monetary Fund Vol. 42 pp. 344374.

    • Search Google Scholar
    • Export Citation

    RichR. W. and D.Rissmiller2000“Understanding the recent behavior of US inflation”,Federal Reserve Bank of New York Current Issues Vol. 6 No 8.

    • Search Google Scholar
    • Export Citation

    TaylorJ. B.2000“Low inflation, pass-through and the pricing power of firms”,European Economic Review 44 (7) pp. 13891408.

    WernerA.2001“Mexico’s Experience with a Floating Exchange Rate,” Bank of Mexico.


Prepared by E. Faal.


The study period was chosen to exclude the turbulent period leading to the 1994–95 crisis and its aftermath, and to also coincide with period of the authorities’ disinflation program.


Taylor (2000) suggests that the level of exchange rate pass-through declines as the level of inflation is lower, mainly because the pricing power of firms is eroded.


Baqueiro, Diaz-Leon, and Torres (2003) found that the level of exchange rate pass-through weakens as inflation falls.


Zapata (1992) observes that the rate of unionization in state enterprises was close to 100 percent since the Federal Labor Law fostered unionization among state workers.


Under the current system of wage setting, indexation of wages to a cost of living measure does not occur regularly, unless explicitly allowed by specific contracts.


Various studies, including Dwyer and Lam (1994), find a tendency for firms to partially but not fully absorb the effects of currency depreciation.


Ball and Mankiw (1994 and 1995) also argue that firms react proportionately more to large price shocks than to smaller ones. This is because they face adjustment costs, and would therefore be more inclined to change prices in response to large shocks than small shocks.


See Chadha, Masson, and Meredith (1992) for a discussion of the rationale for including both forward- and backward-looking components of the inflation process.


Defining shocks in this way avoids the potential problems of regressing inflation on its own components.


The latter is highly correlated with both quarterly and annual GDP— a regression of IGAE on quarterly GDP yields a coefficient of 0.9, and the fraction of variance of GDP explained by the regression is about 90 percent.


This restriction easily passes a Wald test.


The model can also be used to explore the linkages between wage and price inflation and the appropriate response of monetary policy. To do so requires augmenting the model with a policy reaction function, such as a Taylor rule, or an objective function for policy.

Other Resources Citing This Publication