Using industry-specific cost indexes to monitor sources of pressure on inflation

Understanding the causes of inflation is important from a policy perspective because it enables informed policy decisions. Measuring cost pressures may allow policymakers to distinguish whether inflation is being driven by higher wages or higher input prices (cost-push inflation) or stronger demand (demand-pull inflation), and to understand if pressures are generalized or specific to certain goods or services.¹

In this article, we propose a method to approximate the evolution of production costs by industry. We argue that changes in production costs in each industry correspond to changes in the prices of the inputs used, weighted by the importance of each input in total costs. This implicitly assumes a constant technology and that the changes in relative prices do not change the composition of inputs. While this is a strong assumption that would generally not hold, this simplification allows us to construct a time-varying measure of costs with information that is publicly available. Moreover, we show that our measures are useful in understanding the sources and extent of cost pressures at different points in time.

The cost structure (or the importance of each input in total costs) is obtained from input–output (I–O) tables. We distinguish four types of inputs: commodities, domestically purchased inputs, imported inputs, and labor. The price evolution of each category is ideally captured by international commodity price indexes, domestic producer price indexes (PPIs), import prices, and average industry wages, respectively. We use this methodology to estimate cost indexes (CIs) for industries in Mexico at the 6-digit level of the North American Industry Classification System (NAICS), which is the narrowest industry classification in the country.

To illustrate the usefulness of CIs, we link each item in Mexico’s Consumer Price Index (CPI) basket with the 6-digit industry that produces it using a custom correspondence table. Then, we show the evolution and composition of estimated production costs of Mexico’s core CPI basket between January 2006 and July 2024. Despite the peak occurring during the COVID-19 pandemic, Mexico has undergone several episodes of inflation. The decomposition of costs into different types of inputs illustrates how the drivers of cost pressures have changed over time and how the sources of those pressures differ across goods or services.

The CIs proved useful for understanding the inflation surge during 2021–23. The accumulated increases in costs estimated using the CIs reflected the widespread rise in production costs across industries, driven by the shocks associated with the COVID-19 pandemic and the Russia-Ukraine war. The decomposition of these cost changes illustrates the specific sources of pressure at different points in time. While a depreciated exchange rate seems to have influenced costs at the onset of the COVID-19 pandemic, the increases in the costs of inputs became more widespread during 2021, despite a differential impact across CPI subindexes. Food items were particularly affected by commodity prices and domestic inputs. Nonfood merchandise was pressured by the cost of imported inputs, and services inflation was more exposed to labor costs.

Consumer prices generally increased more slowly than producer costs. We argue that demand weakness during the early stages of the COVID-19 pandemic and severe cost increases made it difficult for businesses to fully and immediately pass on these higher costs to consumers. However, the extent of cost pass-through differed between goods and services. For instance, by 2023, the cumulative increase in prices exceeded the increase in costs for goods, whereas the opposite was true for services. It is possible that businesses in the services sector had less scope to pass cost increases through to prices, since the demand for services took longer to recover relative to the demand for goods. This lagged cost pass-through for services may help explain the persistence of services inflation during 2023 and 2024.

More generally, we argue that CIs may be advanced indicators of inflation. To formally show this, we analyze whether they have predictive power over inflation. Our results indicate that monthly changes in the cost indexes provide useful information to predict monthly changes in consumer prices.

These findings raise the question of the extent to which higher costs (as measured by CIs) are passed through to consumer prices (as measured by the CPI) in Mexico, including both the magnitude and the lag with which this happens. The degree to which this happens will depend on

• the strength of demand and thus the scope to raise prices;

• demand elasticity, because greater sensitivity to price increases may force firms to absorb a larger proportion of cost changes;

• supply elasticity, which may partly reflect whether firms experience increasing or decreasing returns to scale;

• cost changes and whether they are generalized or idiosyncratic to particular firms; and

• inflation expectations, because firms may be more willing to pass on cost increases when inflation expectations are high.

In terms of the speed of cost pass-through, there could be a delay if businesses do not want their selling prices to respond immediately to increasing costs, if changing prices is costly, if price rigidities prevent firms from adjusting prices as regularly as desired, or if there is a time lag before higher costs at the factory level materialize in higher retail prices. Using regression analysis, we find a lagged and incomplete pass-through of estimated costs to consumer prices.

The rest of this article describes the construction of CIs and how they are linked to Mexico CPI items. It also presents descriptive results that examine the sources of cost pressures during and after the COVID-19 pandemic, evaluates the predictive power of CIs over inflation, and estimates the pass-through of costs to consumer prices.

Related literature

This article explores cost-related pressures on inflation, particularly cost-push factors in the aftermath of the COVID-19 pandemic. Understanding the drivers of inflation and whether these stem from demand or supply (cost) factors has long been a topic of economic interest.²

Several statistical agencies construct indexes to monitor the cost pressures of specific industries. Different versions of cost indexes are available for the United States.³ Similarly, Eurostat publishes a quarterly cost indicator for new residential buildings.⁴ Statistics Canada calculates the Farm Input Price Index, which tracks the price change of goods and services used by Canadian farmers.⁵ The Official Statistics of Finland calculates a cost index of civil engineering works, which follows changes in input costs relative to a base year.⁶ The U.S. Bureau of Labor Statistics (BLS) satellite series of inputs to industry price indexes tracks cost pressures across a broader range of industries. These indexes are most similar to those constructed in this article.⁷

There are, however, important differences between CIs and the indexes published by BLS. First, CIs correspond to gross indexes while the BLS satellite series are net cost indexes.⁸ Second, BLS indexes consider domestically produced and imported input prices, but they do not consider labor costs. Third, the Mexico input–output table used for the CIs provides information about the share of imports in domestic costs, whereas the BLS satellite series use an approximation, assigning weights to domestically produced or imported commodities according to the value of domestic production of each commodity relative to its imported value. Finally, while the CIs are available at the NAICS 6-digit level disaggregation, the BLS satellite series are available at a more aggregated level (3-digit NAICS). The BLS data is plausibly more precise in terms of the approximated cost of imported inputs, as it directly uses the U.S. Import Price Index, whereas we primarily use the U.S. Export Price Index as a proxy for Mexico’s import price indexes.⁹

Several attempts have been made to measure the pass-through of costs to prices using different tools to monitor cost pressures. Empirical studies have primarily focused on the effects of specific cost shocks on prices (e.g., exchange rate, tax incidence, energy prices, imported input prices, minimum wages, etc.), often restricting their analyses to particular industries or goods. Consequently, it is difficult to compare the wide range of available evidence because of the different empirical strategies employed and the varying types of pass-through measures estimated, such as absolute pass-through or elasticities.¹⁰ The price transmission between intermediate goods price indexes and indexes for final demand has also been explored.¹¹

Some authors have exploited price microdata to estimate cost pass-through. Gopinath and Itskhoki study the relationship between price adjustment frequency and the magnitude of exchange rate pass-through (conditional on price adjustments) for U.S. imported manufacturing goods.¹² They find that firms above the median in price adjustment frequency pass through 44.0 percent of the cost shock, while for those below the median, the estimated pass-through is only 21.0 percent. Similarly, Kochen and Sámano estimate the exchange rate pass-through to consumer prices for a subset of goods and services in the Mexico CPI basket.¹³ Their results indicate that a 1.0-percent increase in the nominal exchange rate leads to a 0.073-percentage point pass-through when firms adjust their prices. This, in turn, results in a 0.043 percent increase in aggregate inflation. Lafrogne-Joussier et al. estimate the pass-through of energy and imported input cost shocks into the French manufacturing PPI, finding a pass-through rate of 32.2 percent for import shocks and a full pass-through for energy shocks.¹⁴ In addition, they find asymmetries between positive and negative shocks, with positive shocks leading to larger pass-through rates. Their estimates suggest that during 2021–22, cost shocks from imported inputs increased the PPI by 1.9 percent while energy shocks raised it by 1.6 percent. For Denmark, Dedola et al. estimate the pass-through of import and energy cost shocks to producer prices.¹⁵ Their findings suggest that the price adjustment to an import cost shock is gradual over the first 12 months, reaching a medium-run elasticity of approximately 0.2. By contrast, energy shocks exhibit much greater pass-through, with a medium-run elasticity between 0.8 and 0.9. The magnitude and dynamics of these price adjustments are primarily driven by substantial heterogeneity in pass-through rates across firm sizes and sectors.

Another strand of the literature examines the relationship between production costs and price-setting behavior through markup estimation. In this context, researchers typically exploit firm-level data on production and total input expenditures and rely on the optimality conditions of the cost minimization problem.¹⁶ Although our article proposes a method to empirically approximate the cost pressures faced by producers within a specific industry, we do not adopt a structural approach nor attempt to estimate the underlying production function governing firms’ behavior.

Finally, our approach has limitations in capturing how cost shocks propagate and aggregate throughout the economy. As Baqaee and Rubbo show, the transmission of shocks in production networks is shaped not only by input–output interactions but also by wedges, substitution patterns, and complementarities across sectors.¹⁷ A framework that models the full general-equilibrium environment provides a richer understanding of how micro-level shocks propagate and aggregate, but such an approach lies beyond the scope of this article. Instead, our proposed cost indicators rely on input–output data and are aggregated using CPI expenditure weights. Although this approach does not incorporate the full set of mechanisms that influence how cost shocks affect the economy, it offers an intuitive, tractable, and policy-relevant measure of cost pressures.

Construction of industry-specific cost indexes

We classify inputs into four categories: commodities, imported goods or services, domestically sourced goods or services, and labor. Commodity prices are assumed to be determined in international markets and are quoted in dollars. Imported input prices are determined by foreign companies, may be proxied by import price indexes, and are referenced in foreign currency. Domestic input prices are determined domestically and may be proxied by domestic producer price indexes at the industrial level. The cost of labor is proxied using industry-specific wages. Each input in the I–O table is assigned to its type. If it consists of a raw material whose price will respond to international market forces, it is classified as a commodity (regardless of whether, according to the I–O table, the input is sourced domestically or internationally). For the rest of the inputs, the fraction classified as an imported or domestic input is determined by the share of expenditure on each input sourced domestically or from abroad, as reported in the I–O table. Labor is measured using the total payments to labor reported in the I–O table. The cost index for industry i at month t is constructed as the weighted sum of the imputed price indexes of each input k used by industry i in production, weighted by the share of k in the total expenditure of industry i. The total expenditure of i equals the value of all intermediate inputs plus payments to labor.

The imputed price index for each input corresponds to the best approximation feasible given the available data. We assume that both commodities and imported goods and services are referenced in U.S. dollars, so the price index in dollars is transformed to the local currency by multiplying its value by the exchange rate of the period (domestic currency per U.S. dollar). All indexes and the exchange rate should be normalized so that they have the same base. This homogenizes the units of all inputs so that they are given in local currency and so that the index is equal to 100 at a chosen base period.

We use gross I–O tables at the 6-digit NAICS level that account for the total value of production in each industry.¹⁸ Using the expenditure structure provided by the I–O table for each 6-digit code i, let $x_{i,k}^M$be the value in basic prices of the imported intermediate consumption of input k for the production of i, and let $x_{i,k}^D$represent the analogous value for domestic intermediate consumption, and $L_i$represent total labor payments. For each input, $x_{i,k}^O$, $O\in \{D,M\}$ corresponds to the total value and reflects both prices and quantities used, and K is the set of inputs. We define the total expenditure of the 6-digit industry i as the sum of all inputs and labor consumed in the production process:

${\mathrm{T}\mathrm{E}}_i=\sum _{k\in K}\left(x_{i,k}^D+x_{i,k}^M\right)+L_i\left(1\right)$

From the definition in Equation (1), we derive weights $\theta$ by origin that reflect the share of input k (sourced domestically or imported) and the share of labor in the total expenditure of i:

${\theta }_{i,k}^D=\frac{x_{i,k}^D}{{\mathrm{T}\mathrm{E}}_i},{\theta }_{i,k}^M=\frac{x_{i,k}^M}{{\mathrm{T}\mathrm{E}}_i},{\theta }_i^L=\frac{L_i}{{\mathrm{T}\mathrm{E}}_i}$

Because of the assumption that prices of commodities correspond to their international reference, regardless of origin, it is necessary to identify the inputs classified as commodities (),$\in C$ and define their weight as the sum of the domestic plus the imported weights:

$\forall k\in C,{\theta }_{i,k}^C={\theta }_{i,k}^D+{\theta }_{i,k}^M$

Because the price indexes of different inputs are obtained from different sources, we need to set a base year for all. We use the base year of the I–O table.¹⁹ In particular, we normalize the prices of all inputs to 1 in that year. This allows us to interpret weights as the quantity share of each input in the total quantity of inputs used in the base year.²⁰ However, this is a strong assumption that will distort the composition of costs in terms of input quantity, so the indexes might give information about the evolution of cost pressures but not about the level of costs. The weights we use are not informative by themselves.²¹

Under the assumption of a Leontief production function, where inputs are perfect complements so that a constant ratio between inputs is guaranteed regardless of scale or input prices, the cost function is given by the sum of the price of each input, weighted by the $\theta$s defined above.²² Then, estimated weights and price data are used to construct the CI for each 6-digit industry code i in month t in the following Equation (2):

${CI}_{i,t}=\sum _{k\in I}\left({\theta }_{i,k}^D{p}_{k,t}^D+{\theta }_{i,k}^M\left(p_{k,t}^M\bullet e_t\right)\right)+\sum _{k\in C}{\theta }_{i,k}^C\left(p_{k,t}^C\bullet e_t\right)+{\theta }_{i,L}{p}_{i,t}^L \left(2\right)$

where I is the set of imported and domestic inputs not classified as commodities, C is the set of inputs classified as commodities, where $p_{k,t}^D$$p_{k,t}^D$and $p_{k,t}^M$$p_{k,t}^M$denote the price index imputed to input k for domestic and imported inputs, respectively$O\in \{\mathrm{d}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c},\mathrm{i}\mathrm{m}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{t}\mathrm{e}\mathrm{d},\mathrm{c}\mathrm{o}\mathrm{m}\mathrm{m}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{t}\mathrm{y}\}$, and $e_t$corresponds to the index of the exchange rate at time t, and $p_{i,t}^L$is the index of the wage paid by industry i at month t. Notice that in Equation (2), the exchange rate enters mechanically, so that the prices of imported inputs and commodities, which are assumed to be invoiced in dollars, mechanically respond to the exchange rate. We construct CIs for the narrowest industry classification in Mexico (6-digit NAICS) for the period from January 2005 to July 2024. We use the 2018 I–O table published by Mexico’s Instituto Nacional de Estadística y Geografía (INEGI), which contains data, valued at basic prices, on domestic and imported inputs and labor consumption across 834 industrial classes considered by INEGI.²³ We use Banco de Mexico’s official FIX exchange rate (Mexican pesos per U.S. dollar) to convert the price indexes of commodities and imported inputs (referenced in dollars) to pesos. Indexes are normalized to 2018 = 100, which is arbitrarily done because 2018 is the base year of the I–O table. Each index can be interpreted as the cost of producing one unit of each good or service relative to 2018, considering that the production structure (composition of inputs) is fixed to that in 2018.²⁴

The proxy used for labor costs should approximate changes in average wages within each industry. For Mexico, the best approximation we have is the index of the average daily taxable income in the formal sector at the 3-digit level of the NAICS industry code that corresponds to each 6-digit industry.²⁵ This is obtained from the Mexican Institute of Social Security (IMSS). This proxy has two caveats. The first is that it is restricted to the formal sector, although we think that this is still a good proxy for the rate of growth of labor costs in each industry.²⁶ The other caveat is that average wages are given for more aggregated sectors (3-digit). This is because IMSS data do not use the NAICS classification system, and the correspondence used is available at this level.

We consider an input as a commodity if it is available in the commodity price series published by the International Monetary Fund (IMF). Commodity prices are assumed to be determined in international markets. The imputed price index for each commodity corresponds to the price index in the IMF commodity prices database, multiplied by the average exchange rate (pesos per U.S. dollars) during the month (index 2018 = 100).

For the remaining goods and services used as inputs in the production of i (i.e., inputs other than commodities and labor), we infer the share sourced domestically and the share imported using information provided by INEGI, which reports both total and imported values in the I–O table. We proxy the change in domestically sourced input costs using Mexico’s PPI for each industry at the 6-digit disaggregation level, available from INEGI.²⁷ For the imported share of inputs, we primarily use the U.S. export price indexes because Mexico’s import price indexes are unavailable at the disaggregated 6-digit level.²⁸

We use the U.S. Export Price Index for input k at the 5-digit NAICS level when U.S. exports of that good to Mexico exceed $1 million during the year, according to data on U.S. trade flows from the U.S. Census Bureau. The reason is that if the United States is an important exporter to Mexico, then U.S. export prices are informative about the prices faced by importers in Mexico. If U.S. exports to Mexico fall below this threshold, we use the U.S. Import Price Index, assuming that Mexico and the United States source these goods from similar international markets. Under this assumption, U.S. import prices approximate the prices that firms in Mexico pay for imported inputs.

Because the United States is a large economy, its export and import price indexes can serve as a reference for changes in international prices of manufactured goods. Export and import price indexes for the United States are available at BLS. These price indexes are multiplied by the average exchange rate during the reference month (index 2018 = 100).

Assumptions and limitations of the cost indexes

The cost indexes we constructed may be a useful tool to show how production costs change. However, there are some drawbacks. First, the construction of the indexes implicitly assumes that the input composition (in quantities) is constant over time and equal to the input composition captured by the 2018 I–O table. Therefore, although input prices change at different rates (which could potentially change the input share if inputs are substitutable), the weights used in the CIs remain fixed. As a result, the indexes do not account for technological changes that could alter the composition of inputs nor do they consider production functions that allow for substitutability of inputs. In reality, firms may substitute inputs as their relative prices change or modify the composition of inputs for other reasons, such as the adoption of new technologies or the availability (or affordability) of new or different inputs. The CIs will not capture these changes. This means that they may misestimate cost pressures arising from current economic conditions or overestimate costs if technological changes lead to more efficient production processes.²⁹

Second, the CIs assume that changes in input prices and the nominal exchange rate are immediately reflected in the cost pressures faced by firms. However, firms may experience these effects with a delay. This is particularly true for the changes in costs associated with exchange rate fluctuations, as cost indexes incorporate these changes mechanically. However, firms may have mechanisms that shield them from exchange rate volatility, such as contracts with their providers or financial instruments to hedge the risk.

Third, the indexes approximate changes in costs that are generalized within industries. However, each firm may have different costs. For example, some businesses may be importing a larger fraction than others or may have a larger scope to adjust the composition of inputs when an input becomes more expensive. Firms could also differ in their bargaining power with providers or in the duration of their contracts, resulting in heterogeneous unit costs across firms.

Fourth, like any empirical measurement, they are subject to measurement errors. One source of these discrepancies is the use of the IMF commodity price series as a proxy for the price of inputs classified as commodities, which may not accurately reflect the prices paid by firms in Mexico and do not account for transportation costs. In addition, we use U.S. export or import prices as a proxy for Mexico’s import prices rather than directly estimating the latter. Moreover, because wage data are available only at a more aggregated industry level, we assign those average wages to corresponding detailed industries. Measurement errors may be more severe in the earlier periods of the series because more missing information had to be imputed using alternatives. For example, industrial classifications change over time, and the PPIs for several industry classifications at the most disaggregated level are unavailable. Given the way these indexes are constructed, the evolution of the costs of these inputs is approximated by the evolution of price indexes of more aggregated industries or, in some cases, by the evolution of U.S. import or export price indexes of these industries.

Despite these drawbacks and the lower precision of earlier estimates, the constructed indexes remain informative about the evolution of cost pressures. A long time series is valuable because it allows for analysis of how cost pressures from different sources evolved across different inflationary episodes, despite the imprecision of these estimates.

Cost increases affecting the core CPI basket in Mexico, 2006–24

To illustrate the usefulness of CIs, this section examines how cost pressures have evolved for the goods and services that form the CPI basket in Mexico. We use a custom correspondence table to link each item g of the CPI basket with the industry or industries i (6-digit NAICS code) that produce it.³⁰ If an item is matched with more than one industry, we define $I_g$as the set of all 6-digit industries that are matched with the item, and the CI of the item is estimated by averaging the CIs of all $i\in I_g$ in the following equation:³¹

${\mathrm{C}\mathrm{I}}_{g,t}=\frac{1}{N_g}\sum _{i\in I_g}{\mathrm{C}\mathrm{I}}_{i,t}$

The following descriptive analysis covers the period from January 2005 to July 2024.³² In order to visualize the results, it is convenient to present them for some baskets that are more specific than the official subindexes published by INEGI. These custom CPI baskets group related CPI items into categories such as food merchandise, nonfood merchandise, and services.³³ CIs for these baskets correspond to the weighted average of the CIs of all the items in each group, using the items’ weights in the headline CPI basket and reweighted to sum to 1 for each aggregation. Therefore, if ${\omega }_g$is item g’s weight in the CPI, the cost index for each basket B is given by:

${\mathrm{C}\mathrm{I}}_{B,t}=\sum _{g\in B}\left(\frac{{\omega }_g}{\sum _{g\in B}{\omega }_g}\right){\mathrm{C}\mathrm{I}}_{g,t}$${\mathrm{C}\mathrm{I}}_{B,t}=\sum _{g\in B}\left(\frac{{\omega }_g}{\sum _{g\in B}{\omega }_g}\right){\mathrm{C}\mathrm{I}}_{g,t}$${\mathrm{C}\mathrm{I}}_{B,t}=\sum _{g\in B}\left(\frac{{\omega }_g}{\sum _{g\in B}{\omega }_g}\right){\mathrm{C}\mathrm{I}}_{g,t}.$

The evolution of the annual growth rate of the estimated cost indexes for the core CPI basket and its components is shown in chart 1.³⁴ This chart shows the contribution of each component (commodities, imported inputs, domestic inputs, and wages) to the annual growth of the cost indexes. We also plot the annual growth rate of the CPI of each basket.

Several inferences can be drawn from chart 1. First, the annual change of the estimated cost indexes is very volatile, much more so than that of the Consumer Price Index. Part of this volatility can be attributed to the fact that changes in the exchange rate are mechanically incorporated into the cost indexes (see appendix B).³⁵

Chart 1 shows how cost indexes identify episodes of inflation associated with rising production costs. The composition of cost increases differs across CPI baskets. While labor costs contribute relatively more to services, commodity prices are more important for processed food and imported inputs contribute more to nonfood merchandise. The chart also shows that cost pressures became more widespread during the COVID-19 pandemic, with all major components contributing to increases in estimated costs.

The annual increase of estimated aggregate costs for Mexico’s core CPI basket peaked during the COVID-19 pandemic. However, Mexico has gone through several significant episodes of rising costs. For example, costs rose sharply during the 2007–09 recession, particularly for the inputs used to produce nonfood merchandise, especially imported inputs. Moreover, during the period 2014–17, there was also a sustained increase in estimated costs that may be associated with the steady depreciation of the exchange rate.

Notably, elevated inflation persisted throughout much of the COVID-19 pandemic, and all core CPI components contributed to that increase, reflecting how broad-based and generalized cost pressures were during the period. Indeed, between 2020 and 2022, costs increased because of sharp rises in commodity prices (which had a particularly adverse impact on processed food costs), imported inputs (which mainly affected nonfood merchandise), domestic inputs (which affected all items), and wages (particularly for services). Following the COVID-19 pandemic period, the cost of commodities and imported inputs declined during 2023 and 2024, possibly reflecting the appreciation of the exchange rate during this period.

Finally, Chart 1 reflects that over the last 6 years, high wage growth has translated into cost pressures. This component was relatively small before 2018 but has evidently gained importance in explaining rising costs in recent years.

Table 1 illustrates that consumer price inflation for the overall core CPI basket, food merchandise (processed food), and nonfood merchandise is positively correlated, albeit, weakly, with the annual change in their corresponding CIs. Specifically, the Pearson correlation of the series (shown in each graph) is statistically significant at the 5% level: about 0.14 for the CPI basket; 0.33 for processed food, and 0.15 for nonfood merchandise.³⁶ In contrast, the correlation between these series is not statistically significant for services. A plausible explanation is that services inflation is much less volatile than inflation of the other baskets.

Cost pass-through and inflation during the COVID-19 pandemic

To assess how costs evolved since the beginning of the COVID-19 pandemic, chart 2 illustrates the accumulated increase of each component of the cost index from February 2020 to July 2024 for the core basket and its components (processed food, nonfood merchandise, and services). To compare the estimated accumulated increase in costs with that of consumer prices over the period, we also show the accumulated increase in the CPI for each component.

Chart 2 shows a substantial increase in the production costs of items within the core CPI basket since the beginning of the COVID-19 pandemic. As of July 2024, the CIs of processed food, nonfood merchandise, and services registered a cumulative growth rate of 30.13 percent, 22.75 percent, and 27.63 percent, respectively, implying a 27.11-percent increase in the overall core basket from February 2020 to July 2024.

For processed food, the rapid and sustained price growth observed between early 2021 and 2022 was mainly driven by higher commodity prices, which were exacerbated in early 2022 with the onset of the Russia-Ukraine war, when grain prices rose substantially. Rising pressures from domestic inputs were also observed from the beginning of 2021 until mid-2024, countering the price reductions in commodities and imported inputs that began in 2023. Imported inputs were the primary contributor to cost growth for nonfood merchandise during 2020 and 2021, likely reflecting the exchange rate depreciation observed at the onset of the COVID-19 pandemic. Beginning in early 2022, however, domestic inputs and labor played an increasingly important role. Services exhibited a similar trend, with a decline in the contribution of imported inputs since mid-2022 and a persistent increase in the contribution of domestic inputs and wages to cumulative cost growth.

Comparing the cumulative changes in cost indexes and consumer price indexes explains differences in the extent to which estimated cost increases were passed through to consumer prices and may help rationalize why inflation has evolved differently across components. A negative gap—that is, when the cumulative increase in the CPI is smaller than that in the cost index—can be interpreted as evidence that businesses are absorbing a proportion of cost increases or that they are unable to fully pass through cost increases to consumer prices. In contrast, a positive difference indicates that the changes in the CPI exceed estimated cost increases.³⁷

The results suggest that the increase in consumer prices occurred at a slower rate than that of the cost indexes for all baskets. This could have been a consequence of a weakness in demand during the early stages of the COVID-19 pandemic (from February 2020 to late 2022) or of the severity of cost increases that made it difficult for businesses to fully pass through higher costs to consumers. Since 2023, however, the estimated gap reversed for nonfood merchandise and processed food, as accumulated increases in consumer prices (as indicated by the CPI) exceeded estimated cost growth (as indicated by the cost index). A decline in the factors putting upward pressure on costs contributed to this behavior. However, it is possible that the resilience of demand for goods allowed firms to continue increasing prices. In contrast, the estimated cost increases for services exceeded the increase observed in the CPI. Two factors may have contributed to this evolution. First, the pass-through of costs to consumer prices plausibly happened at a slower pace for services since businesses in this sector may have had less scope to translate higher costs into higher selling prices, given that the demand took longer to recover. Second, elevated wage growth during the period continued to pressure costs, particularly for services, because of their large share of production costs. The lagged pass-through and elevated growth rate may explain why services inflation in Mexico was particularly persistent through July 2024.