2.1 The setting

Consider a country where:

Commodity producers produce at no cost a volume X of export commodities at a world price P _X .

Formal manufacture producers produce a volume Y _F of consumption goods at the price P _C , using imported intermediate inputs M at the world price P _m , and labor L _F at a wage rate ω, under a Cobb-Douglas production function:

where A is the total factor productivity of the formal manufacturing sector.

Workers have the option of staying in an informal (subsistence) sector where they produce and consume a volume YH of consumption goods (at the same price PC ) based only on their own labor:

where v , the labor productivity in the informal sector determines the informal wage.

As long as the economy remains perfectly open, the price of the consumption good is set externally:

Instead, under an import substitution regime, import permits limit to a fraction 1 - q of total consumption the consumption goods that can be imported. Thus, commodity exporters must use a fraction q of their export income to purchase locally produced goods, sold at a price PC>PC* .

Local manufacture producers then use the foreign exchange from the commodity exporters’ purchases to import intermediate goods; hence:

The total working population is L- , and there is perfect labor mobility at the ongoing wage:

2.2 The commodities trap

Assume for now that firms are perfectly competitive. The first order profit maximization conditions may then be written as:

where PC is the price of the consumption good in terms of the price of the imported inputs. Substituting the factors ratio from (7) into (6):

This condition defines an upward sloping relationship between the real wage (and hence marginal labor productivity) and the price of the consumption good, shown as the A curves in Figure 1.

Source: Authors’ elaboration.

Figure 1 Trade protection, the home price of consumption and the real wage  

As the TFP of the formal manufacturing sector rises, the A curves rotate clockwise. Suppose that initial TFP is A0 . In an open economy, the price of consumption is set externally. Hence, to be able to compete globally (i.e., to export manufactures), the real wage would need to be ω0* at the point of intersection, E0* , between the horizontal line pC=pC* and the A0 curve. However, as long as the real wage offered by the formal industrial sector, ω0* , remains below the informal wage, v , workers remain informal and industrialization cannot take place. This will be the case as long as the industrial sector’s productivity is insufficient to allow it to compete globally:

In this case, the economy remains in the commodities-informality trap E0* , where it only exports commodities, imports but does not produce manufactures, its labor force remains informal, and its growth rate is set by its exports.

2.3 ISI labor market equilibrium

Suppose now that the economy is partially closed through import permits, so that commodity exporters must use part of the proceeds of their exports to consume locally produced manufactures. As a result, pC rises to p^C0 in Figure 1, at the intersection of the A0 schedule and the informal wage, v , and the economy settles in the ISI equilibrium E^0 . At this point, as long as commodity exports (hence, the demand for local manufactures) are insufficient to shift the whole labor force away from the informal sector, formal industrialization will start, but the economy will remain dual, with an informal sector coexisting with the formal sector. Noting the dual labor market equilibria with a hat, to differentiate them from the single labor market equilibria where the informal sector has been fully absorbed (noted with a tilde), the economy will remain dual as long as:

The labor market dynamics can be represented as in Figure 2. As long as L^FD ≤ L- , an increase in protection (q) raises formal output by pulling in workers from the informal sector at the informalwage. Once L~FD≤ L- , informality disappears and further increases in protection raise the formal wage, ω , above the informal wage, v . In Figure 1, the economy will thus shift from E^0 to E~0 , at the intersection of the A0 schedule with the full employment hyperbolic curve M0 obtained from (4), (6), and (7) as:

Source: Authors’ elaboration.

Figure 2 Labor market: Dual vs. single equilibria 

Several important points are worth stressing. First, as increases in protection (or commodity exports) raise real wages (i.e., as the M0 and LFD schedules move rightward in Figure 1 and Figure 2, respectively), the higher costs of production pull away the ISI economy upward along the A0 schedule, away from its global competitiveness frontier. Thus, for the economy to become globally competitive, productivity needs to be raised all the way to A2 > A1 , so as to shift from E~0 to E2* in Figure 1. Closing the productivity gap may, therefore, become harder for economies having experienced a lengthy ISI regime with increasing protection and/or abundant and rising commodity exports. As we will discuss further below, this helps explain why Latin American countries could not manage a successful transition to manufacture exports-led growth the way Asian countries did.

Second, protection may serve different objectives depending on where the economy lies in Figure 2. On the vertical portion of the labor supply schedule (with informality), increasing protection will raise the size of the formal industrial sector, possibly bringing it closer to its global competitiveness frontier through learning and scale effects that improve productivity. On the horizontal portion (where informality has disappeared), increasing protection may serve a redistribution objective by raising real wages.

Indeed, as rising protection or commodity exports push the local wage above the informal wage, workers become better off under ISI than under the commodities trap. Instead, because the price of local consumption goods keeps rising (or their quality deteriorates), commodity exporters’ welfare worsens increasingly in relation to what it would have been under the open commodities trap economy. Thus, at least in this perfectly competitive economy where there are no monopolistic rents, one would expect the income distribution to improve under ISI, both because the real purchasing power at the lower tail of the distribution improves and that at the upper tail worsens.

This being said, the strict stepwise relationship between the real wage and informality, such that the real wage remains constant as long as there remains some informal labor, may look extreme and can easily be smoothed out. Suppose that instead of a single type of labor, there are two types, perhaps one relatively more skilled than the other. In this case, an increase in protection could raise the demand for skilled labor above its supply, hence pushing the skilled labor wage up. If so, the average wage (skilled and unskilled) should also start rising. With more labor categories, the average wage could thus become a smoothly rising function of protection (or commodity exports) and be inversely related to the informal labor share in the economy. In the narrative for the rest of the paper, we will, therefore, associate increases in protection a bit more loosely with both higher wages and lower informality.

2.4 ISI welfare and growth

How would the aggregate welfare and growth of a closed ISI economy compare to that of the open economy? To check this, let Y be the economy’s GDP, expressed in terms of consumption goods. Y should equal the sum of the value added by the informal sector, ωLH , the formal industrial sector, ωLF , and the commodities sector, qpXXpC+1-qpXXpC* :

Instead, the economy’s GDP under the commodities trap would be:

Consider first the static welfare implications. As shown in Appendix 1, protection is “inefficient” because it lowers GDP (it destroys value). Protection taxes commodities exports while promoting domestic industrialization and raising wages. Thus, while it improves workers’ welfare, it reduces commodities exporters’ welfare, and the latter more than offsets the former. Hence, unless industrialization produces positive externalities by pushing the economy closer to its global competitiveness frontier through scale and learning effects, instead of industrializing, it would be more efficient to tax commodity exports and use the proceeds to import consumption goods and give these away to workers (see the proof in Appendix 1).

Consider next the growth implications. In this simple model setting without investment and capital and in the absence of changes in TFP and for a given labor supply, the rate of growth of the economy will be determined by the growth of commodity exports. As shown in Appendix 1, protection is again inefficient: it promotes the growth of the formal industrial sector (and in a model with capital would no doubt promote investment in that sector); but it reduces the growth response of the economy as a whole to an increase in the growth of export commodities. There is, therefore, an apparent inconsistency between the rapid observed growth of Mexico under the rising protection of the ISI years and the lower growth predicted by our model.

This inconsistency could be resolved if the rapid growth was a reflection of manufacturers’ success in raising their TFPs. However, should this have been the case, the rise in productivity should have rotated the A schedule clockwise in Figure 1, gradually reducing the price of the locally produced goods until the formal sector’s productivity reached A2 . At that point, the productivity gap would have closed, and local manufacturers should have started to export at the world price. Clearly, this did not happen.

This second inconsistency can be resolved in two ways. First, the downward shift toward the competitiveness frontier could have been systematically offset by an upward shift away from the frontier associated with increases in protection or commodity exports: the economy could then have followed the orange arrow trajectory in Figure 1, with steady prices and no real exchange rate appreciation. Second, even if the downward shift had dominated the upward shift sufficiently to bring the economy down to its competitiveness frontier, the quality of the domestically produced goods could have differed from that of imports because the local manufacturers lacked the know-how, sophistication, or motivation to match the quality of their foreign counterparts. In either case, the domestic goods would have remained globally uncompetitive, notwithstanding the strong increases in productivity. Protection, therefore, allowed local manufacturers to capture domestic economies (of scale or learning) by selling to a captive clientele without having to worry about the foreign competition, thereby granting them the benefits (at least for a while) of high productivity gains and rapid growth. But at the same time it prevented local manufacturers from reaching their global competitiveness frontier because it raised their costs of production or prevented them from reaching the required export quality. In either case, it continued to block their access to foreign markets.

In turn, this enduring wedge between home and foreign goods can help explain both what happened during the ISI years in terms of inflation and what happened with trade liberalization in terms of real GDPs. The relatively good inflationary record notwithstanding the increases in protection may be explained (at least during the early ISI years of the Desarrollo Estabilizador) by both a mix of productivity gains and lower-quality goods with steady prices. Similarly, the GDP gains predicted by our model when protection is reversed would a priori appear to be inconsistent with the initial GDP losses incurred by the Mexican economy when it liberalized its trade (more on this below). However, this inconsistency could be resolved again by addressing quality differences. If the productivity gains obtained under trade protection were strictly associated with lower quality goods, the flood of better quality imports under trade liberalization would have wiped out local producers’ capacity to compete no matter how productive they had become in producing “LADA cars”.

2.5 Open growth

Given its cost and quality impacts on production, the ISI regime can therefore also become a trap and coming out of it requires opening up. In the short run, absent any changes in A , trade liberalization should drive the economy in Figure 1 from the ISI equilibrium, E^0 , back to its pre-ISI starting point, E0* . This would imply a massive de-industrialization and labor shift back from the formal to the informal sector. Leaving this new equilibrium would only be possible for firms that manage to raise their TFP from A0 to A1 ≥ A* , so as to reach E1* , at which point they could start exporting. With a perfectly elastic external demand for exports and a constant returns production function, full industrialization (the complete absorption of the informal sector) would be reached immediately. From there onward, additional increases in productivity, A2 > A1 , would raise the formal wage over and beyond the informal wage.

A more realistic dynamic would assume imperfect competition and imperfectly elastic demands faced by each firm. Suppose the representative firm faces a partially elastic demand such that:

where pC is now the price of the individual good relative to a broader world consumption basket; B is a scale factor that depends on world demand and the quality of the good; Y* is the income of the foreign purchasers of the good, ε is the income elasticity and η the price elasticity of their demand. The firm’s mark-up, τ , being the inverse of the demand elasticity τ=1η , the usual condition for the existence of an equilibrium with market power τ ∈[0,1] implies η ≥ 1 . The profit maximization conditions can thus now be written as:

As long as an informal sector exists the real wage in the formal sector, ω , must equal the informal wage, v . In this case, the formal sector’s production function (1), the demand schedule (14) and the two first-order conditions (15) and (16) form a system of four equations in the four unknowns pC , YF , M , and LF , which can be solved for the formal labor demand, formal output and the price of the good as:

where YF=NyF and N is the number of firms in the economy.

Including firms’ earnings, GDP, expressed in terms of world prices, may now be written as:

where pM , pX , and pC are now the prices of intermediate imports, commodities, and local manufactures in terms of the price of the world consumption basket. For simplicity, informal labor is also assumed to produce a good whose price is constant in terms of the world consumption basket, so that the informal wage, v , is also constant in terms of that basket.

Differentiating (20) for a given η and τ while using (18) gives the GDP growth rate:

The first two terms on the right-hand side of the above expression account for factors’ contribution to growth (since there is no capital in this model, the only factors are commodities and labor).¹ The third term accounts for the economy’s TFP. It includes the usual supply-based productivity factors for both the formal and informal sectors ( A and v ) but it also includes demand effects ( τ,ε,B , and Y* ) and an entrepreneurial connection factor between supply and demand, the number of firms ( N ) with a market niche for exports. A boost in external demand (or in the number of exporting firms) raises TFP because the resulting rise in exports (which is accompanied by a rise in formal labor and imports of intermediate goods) raises formal firms’ output and earnings.

Although labor’s marginal productivity remains unchanged at the informal wage, a labor shift from the informal sector to the formal sector raises aggregate TFP because the formal sector is more productive than the informal sector. By producing goods with more efficient technology and selling them globally with demands that are not perfectly elastic, formal firms generate additional output that translates into additional value due to positive mark-ups. And it can indeed be easily verified with (18) and (19) that, provided that ApC > 1 , the less elastic the demand for these goods (i.e., the higher τ ), the more the country’s GDP rises.

This condition will hold if formal firms are sufficiently efficient and innovative: their income should exceed what would be expected based only on the quantities of labor and intermediate inputs they use.

In this setting, factor reallocations can, therefore, play a key role in driving changes in aggregate TFP, leading to boosts in GDP, as in Hsieh and Klenow’s approach. But these reallocations have nothing to do with removing policy distortions. They are instead the result of external demand shocks associated with increases in external income ( Y* ), changes in the demand for the goods the country is exporting ( B ), changes in demand elasticities ( ε or η ), or increases in the number of export market niches ( N ). The resulting TFP and growth fluctuations are “structural”: they are linked with a dual labor market that functions as a buffer and produces changes in aggregate output and productivity without any change in prices or wages (check equation 19). Thus, unlike endogenous TFP fluctuations caused by changes in capacity utilization along the business cycle, these are external, demand-triggered fluctuations in TFP that cannot be “cleaned out” and removed from the data through simpleaveraging over the business cycle. Instead, they may happen at any time and are not inherently reversible.

External demand-triggered changes in TFP would also happen in the absence of an informal sector, but to a much lesser extent and for a different reason. To see this, solve the above four equation system again, but this time allow the wage, ω , to be freely determined and instead formal labor to remain equal to its full employment value, L- . In this case, formal output would now be given by the following expression:

where β = 1/(1 - α + αη) . Comparing (18) with (23), it can be immediately seen that the formal output elasticity in response to a demand shock, B , will be higher in the dual economy than in the single labor sector economy if (1 - α)β < 1 , which will always be the case for η ≥ 1 .² Hence, in dual economies such as those of Mexico and most of Latin America, the higher supply elasticity response to world shocks should result in higher GDP volatility.

3.1 From the easy ISI to the hard post-ISI: growth, informality and welfare patterns

During the ISI years, Mexico grew substantially faster than the rest of LA (Figure 3). Instead, during the Crisis and Adjustment period, Mexico’s growth collapsed, much like LA’s and EE’s; yet, CG continued to grow at a fast pace. This section argues that this highly contrasted post-ISI response is explained by domestic dynamics brewing deep within the ISI period rather than by unfavorable world shocks, as suggested by ^{Bértola and Ocampo (2013)}.⁴

Source: ^{GGDC (2023a)}.

Figure 3 GDP growth: Mexico vs. LA, CG, and EE 

Similarly to the average for LA but in deep contrast with CG, Mexico’s industrialization (measured as manufacturing value added as a percent of GDP) developed at a very early stage during the ISI period (Figure 4a). Yet, industrialization was mainly inward-looking: Mexico and other LA countries remained basic commodity exporters (Figure 4b). The similarity between Mexico and the rest of the region ceased during the post-ISI period. While Mexico turned into a very successful manufacturing exporter, other LA countries sharply deindustrialized, and their exports remained focused on commodities (the Southern American countries) or shifted toward services (the Central American countries). This sharply differentiated impact of trade liberalization can be largely traced back to comparative advantages. The early end of Mexico’s oil boom during the early eighties made it quite evident that the country should not count on commodities to sustainably support its growth. It should instead capitalize on its unique geographical advantages sitting next to the largest and wealthiest consumer market in the world.

Notes: All observations are calculated as ten-year backward-looking moving averages of yearly growth rates. To facilitate historic comparisons between the three sets of countries, this figure is derived from the Maddison Project Database instead of the World Bank’s Word Development Indicators (WDI).

Source: WDI (^{World Bank, 2023a}).

Figure 4 Industrialization: Mexico vs. LA, CG, and EE 

Mexico’s real wage and economy-wide labor productivity were deeply procyclical, rising very rapidly during the ISI years. They then collapsed during the post-ISI Crisis and Adjustment period, and recovered, albeit not yet fully, during the China and Globalization period (Figures 5 a and 5b). The income distribution (measured through the GINI index) went through a broadly similar cycle. While ISI period data are too spotty to draw firm patterns during this period, the GINI clearly worsened during the Crisis and Adjustment period before recovering during the China and Globalization period (Figure 5c).

Source: Archivo Histórico (^{INEGI, 2023}) and WDI (^{World Bank, 2023a}). Notes: Labor productivity is obtained as GDP divided by the working-age population. The real wage is obtained as the hourly nominal wage in the industrial sector in Mexico City, deflated by the consumer price index. The GINI series is intrapolated using historical data for 1984 and 1989 as given in ^{Londoño and Szekely (1997)} and put together for later years with the higher frequency data provided by the World Bank since 1992. GINIs for the ISI years are not shown because the data is too erratic and fails to provide a clear trend

Figure 5 Mexico: Labor market indicators and income distribution 

The industrialization, real wage, employment, and distribution patterns observed during the ISI cycle are, therefore, all consistent with the dynamics highlighted in the industrialization model of the previous section.⁵ As long as sufficiently abundant commodity exports and appropriately tight import permits provided sufficient room to manage the balance of payments constraint, trade protection continued to promote a labor shift from the informal sector (rural or urban) to the formal manufacturing sector. This enabled inward looking industrialization and raised the real wage. At the same time, by allowing for scale and learning economies, the captive local demand enabled local manufactures and labor productivity to grow rapidly.

However, once fiscal pressures started to build up and capital outflows further pressured the balance of payments (more on this below), import permits were no longer able to equilibrate the current account. As the balance of payments turned unmanageable, trade liberalization became unavoidable. It led to an opposite pattern of steep falls in real wages and a worsening income distribution. Given the quality wedge referred to in the previous section, it also depressed GDP, hence reducing aggregate labor productivity. While industrialization started to decline as a share of GDP, it recovered in the case of Mexico thanks to its proximity to the United States (US), which allowed manufacture exports to rise rapidly.

Regarding informality, the only data available on a yearly basis since 1960 is the number of workers insured by the Mexican Social Security Institute (Instituto Mexicano del Seguro Social, IMSS), which formed the core of the formal sector. Dividing this series by the working-age population provides a first approximation to the formality rate and, hence, to its complement, the rate of informality. As shown in Figure 6a, the formality rate has followed a steeply upward trend that would suggest a steady decline in informality as the economy developed. However, this is not consistent with survey data for recent years, which shows a rather steady rate of informality. The discrepancy most probably reflects the accumulation in the IMSS accounts of workers who became formal for only short periods of time. Instead, as shown in Figure 6b, based on yearly rates of change, the rate of formality has fluctuated in a more meaningful way around this upward trend. During the ISI years, formal employment grew at very high rates before growing at much slower rates during the Crisis and Adjustment period and recovering partially during the China and Globalization period.

Notes: Figure 6a shows the number of IMSS beneficiaries divided by the working age population. Figure 6b shows the ten-year backward moving average of the yearly rate of change of the series in Figure 6a. Figure 6c plots the rate of GDP growth together with the rate of growth of formality, calculated as % of the working age population. Source: Memoria Estadística (^{IMSS, 2023}) and WDI (^{World Bank, 2023a}).

Figure 6 Mexico: Employment formality (as % of working age population)  

Figure 6c shows that the rates of growth of formality (informality) have been closely related to the yearly growth rates of real GDP.

Thus, higher GDP growth has been indeed associated with systematic labor transfers from the informal to the formal sector. But formality increased by more under booms than it decreased under busts, a downward rigidity that is consistent with the labor market distortions that penalize labor dismissals. Thus, after accounting for the statistical distortions associated with the trend in the formality level and for the impact of labor market rigidities, the shorter-term rate of change relationship between growth and informality also does match that described in the model of the previous section.⁶

As for other countries enjoying trade protection in each of the three peer groups, Mexico’s high GDP growth was enabled until the mid-70s by high TFP growth (Figure 7a). However, the TFP gains started trending downward in the early to mid-70s and collapsed during the 80s.⁷ Thus, Mexico’s growth during the late 70s and early 80s was no longer fed by productivity increases but instead by the large investments of the oil boom (Figure 7b). These patterns are again consistent with the narrative coming out of the model presented in the previous section: while trade protection initially promoted rapid, domestically-induced TFP gains, these eventually ran out of steam. And they turned negative after Mexico liberalized its trade.

Notes: The rates of TFP and GDP growth are calculated as ten-year backwardlooking moving averages of yearly data. Source: ^{GGDC (2023b)} and WDI (^{World Bank, 2023a}).

Figure 7 GDP and TFP growth  

3.2 From the easy ISI to the hard post-ISI: a macro and trade growth decomposition

To track better the drivers of growth during ISI and beyond, a macro and trade perspective (instead of productivity and factor accumulation) is needed. Growth can be decomposed based on an accounting identity that focuses on the growth of exports and imports (all expressed in growth rates relative to the world):

Defining the growth of GDP as G , that of exports as EP (exports pull), that of the trade account as EL (external leverage), and the residual as DR (domestic response), the imports minus exports, the above identity can hence be written as:

Thus, as explained in more detail in Appendix 3, GDP growth reflects the sum of three driving forces: the “exports pull” ( EP ); the use of external finance or “external leverage” ( EL ); and the response of GDP growth to the flow of incoming imports or “domestic response” ( DR ). The EL term reflects domestic demand; the EP term may reflect external supply or external demand, and the DR term domestic supply or domestic demand.

Based on a variance decomposition of the growth decomposition between the countries of each group (i.e., how each component of the growth decomposition contributes to explain the sum), Figure 8 contrasts the growth of the LA and CG regions across the three periods of our study. The most striking feature of the figure is that while the growth of CG countries remained externally oriented, that of LA countries was externally driven during ISI (under trade protection) but became domestically driven after trade was liberalized. That growth during ISI was mostly driven by exports, even in LA countries, is consistent with the model of the previous section. Even though industrialization was inward-oriented, its growth depended on the availability of foreign exchange (i.e., commodity exports).

Source: Author’s own calculations, based on methodology presented in Appendix 3 and data from WDI (^{World Bank, 2023a}).

Figure 8 Growth variance decompositions: External vs. domestic growth  

Thus, the LA countries that grew faster were the ones that exported more, as commodity exports (the EP term) conditioned the growth of their domestic industrial sector. However, once trade was liberalized, the dependency on commodity exports ceased; instead, growth started to be affected by the economy’s domestic response to trade (the DR term). Unlike in CG countries, the integration of LA countries’ trade with the rest of the world was “imperfect” in that GDP growth did not respond “efficiently” to the incoming flows of imports. As a result, growth became DR -driven rather than EP -driven: the LA countries that grew faster became the ones that imported less (given their rates of growth), rather than those that exported more.

3.3 ISI’s unsustainability

Below, we will return to this key paradox. We need first to understand why the ISI growth miracle became unsustainable. For this, it is useful first to look at Mexico’s year-to-year growth decomposition (Figure 9). During the ISI years, growth dynamics can be broken down into two very distinct sub-periods. During 1965-1976, the steady decline of the EP component of growth suggests that the country’s capacity to keep growing based on commodity exports was becoming exhausted (Figure 9a). Growth could only be maintained because the decline in EP was offset by a rise in DR . The latter, in turn, was most likely associated with a gradual tightening of import permits that allowed domestic manufactures to keep expanding.

Source: Author’s own calculations, based on methodology presented in Appendix 3 and data from WDI (^{World Bank, 2023a}).

Figure 9 Mexico: Key features of the growth decomposition, selected periods  

Historical series on import permits, the likely linchpin of the ISI regime in Mexico, are unfortunately not readily available. However, the economy’s commercial openness may be used as a rough proxy (Figure 10a). And indeed, up to 1970 (the year of transition from the Desarrollo Estabilizador to the Desarrollo Compartido, following the matanza de Tlatelolco and the incoming Echeverría administration), the balance of payments was maintained in equilibrium thanks to prudent fiscal management (Figure 10b). Thus, during this period the economy was growing rapidly yet becoming increasingly closed, which is consistent with the increasing use of import permits.

Notes: The current account is expressed in current US dollars. The commercial openness is obtained as imports plus exports over GDP. The fiscal balance is that of the public sector as a whole (as % of GDP). Source: ^{Martin (2023)}.

Figure 10 Mexico: Key features of the macroeconomic stance, 1950-1982  

The Desarrollo Compartido and the oil boom (1977-1982) radically altered this dynamic. During 1970-1976, the loss of the fiscal equilibrium and the heavy public spending destabilized the current account (Figures 10 a and 10b). The oil boom temporarily relaxed the balance of payments constraint, boosting both exports and imports, and allowing growth to rise further. However, the boom eventually destabilized the current account much further, leading to the 1982 crisis. Post-ISI growth dynamics became then driven by demand (Figure 9b): a huge contraction from 1982 to 1991 (a fall in EL partly compensated by a rise in DR ) triggered by the debt crisis, followed by an equally large expansion (a rise in EL partly compensated by a fall in DR ) triggered by inflation stabilization policies during the Pacto de Solidaridad.

Mexico’s inflation started to rise in the early 70s with the expansionary fiscal policies of the Echeverría administration (Figure 11). It got a further kick with the public spending of the oil boom under the López Portillo administration. By the time of the debt crisis of 1982, the vicious circle between price inflation and exchange rate depreciation had already become well established; as a result, inflationary dynamics worsened for more than a decade. Similar dynamics affected the rest of the LA region. Thus, while Mexico’s inflation was initially below CG’s, it overtook it as early as in the mid-70s and kept rising from there on. And although the fiscal excesses of the Echeverría administration arguably reflected a crisis of legitimacy not directly related to the ISI regime, those of the López Portillo administration were the direct consequence of the need for a rapid expansion of commodity exports as the key to keep the ISI growth miracle going. But in both instances, by closing the door to imports, the strict trade protection of the ISI years exacerbated the inflationary impact of domestic demand shocks. Thus, in one way or another, the inflation problem in Mexico (and in the rest of the region) was the child of the ISI years, rather than the result of external shocks, as claimed by the structuralist narrative.

Notes: The inflation and real exchange rates are calculated as the log of 1 plus the ten-year backward-looking moving average of yearly inflation rates or real exchange rates against the US dollar using GDP deflators as price indices. Source: WDI (^{World Bank, 2023a}).

Figure 11 Inflation, demand and the real exchange rate: 1965-2000  

Why did Mexico (and the rest of LA) fail to achieve a smooth transition out of ISI into exports-led growth, the way South Asia did? As noted in recent academic research, firms’ incentives to innovate and learn new technologies vanish once they fall too far below the foreign competitiveness frontier, even more so when they enjoy comfortable rents in a protected environment.⁸ LA’s inability to transition reflected this problem; the wide availability of commodities no doubt exacerbated it. As Figure 12a shows, per capita commodity exports were far higher in one region than the other, particularly when setting aside Malaysia, a clear outlier. Thus, the need to rapidly find a way out toward the exports of manufactures was far more intense in Southeast Asia, which helps explain the much more export-oriented policy emphasis in one region than in the other. Instead, by contributing to the rise in real wages in the manner described in the model of the previous section, the rising protection and commodities abundance further distanced the Latin American countries from their global competitiveness frontier. Hence, it made the labor market implications of trade liberalization even more problematic.

Source: WDI (^{World Bank, 2023a}).

Figure 12 Exports: Selected indices 

To this “push” factor, one may add a “pull” factor. In one case, Japan was the local “locomotive” pulling the rest of the region forward. In the other case, it was the US. Thus, reflecting its more dynamic exports (Figure 13b), Japan invested heavily in the region to help produce the components of its export boom. Instead, US firms invested in Mexico and the rest of LA mainly to benefit from the rents associated with the heavy protection of the local markets for tradable goods.

Source: Author’s own calculations, based on methodology presented in Appendices 3 and 4, and data from WDI (^{World Bank, 2023a}).

Figure 13 Growth decompositions of countries undergoing trade liberalizations  

Finally, to the push and the pull, one needs to add a “catalytic” factor: industrial policies as a key vehicle to shorten the gap from the economy’s competitiveness frontier. The policies clearly differed, not only in terms of their orientation (inward vs. outward) but also their quality: in one case, pragmatism and a close public-private collaboration; in the other, a heavy ideological bias and much less working together to resolve collective action frictions.

3.4 The post-ISI adjustment costs

The post-ISI adjustment costs were of two types: 1) the costs associated with the loss of macroeconomic control, which culminated with the crises of 1982 and 1994-1995; and 2) the costs associated with the efforts to put things back on the right track through trade liberalization and inflation stabilization. To help assess the magnitude of these exit costs and the manner in which they depressed Mexico’s post-ISI growth potential we conduct two growth decomposition exercises that compare Mexico with other countries that went through similar experiences. The first exercise compares the growth of countries that underwent substantial trade liberalizations, as Mexico did during the mid-eighties. The second exercise does the same for countries that experienced severe and prolonged real exchange rate appreciations, as Mexico did during the first half of the nineties when it sought to stabilize its inflationary dynamics. The methodological details and country sample compositions for both exercises can be found in Appendix 4.

The trade liberalization sample covers 31 countries that experienced a sudden simultaneous divergence of their export pulls and domestic responses (their EP going up, their DR down), as should happen when trade liberalization stimulates both exports and imports (the timing of country dynamics is aligned, with the EP-DR widening starting at time zero). As shown in Figure 13a, the initial average response to trade liberalization is a decline in growth, as imports ( DR ) pick up faster than exports ( EP ). However, growth turns positive as early as in year 4, as exports continue to rise while imports reach a plateau. The full life cycle of trade liberalization extends over nearly three decades.

For Mexico (Figure 13b), the initial growth contraction lasted longer (7 years instead of 4) and was much deeper (down to -0.3 instead of -0.08). This was partly because exports initially fell instead of rising, mostly due to the reversal of the oil export boom that depressed Mexican exports. More importantly, however, this was because Mexico’s DR collapse was far deeper than that of the rest of the world (down to -0.57 instead of -0.19). But revealingly enough, that collapse was extremely similar to that of Eastern European countries. This is consistent with the fact that both Mexico and Eastern European countries opened very rapidly following decades of extremely tight trade protection. Thus, as predicted by the industrialization model, adjustment dynamics were much tougher for the countries that opened up when the quality of the goods they produced and/or their total factor productivities lay far below the world’s competitiveness frontier.

Turning now to the real appreciation exercise, the Mexican government agreed with the labor and business sectors on a social pact (the Pacto de Solidaridad) aimed at stabilizing inflation, running at the time at above 100%. The Pacto contemplated limiting further nominal exchange rate depreciations in exchange for wage and price controls. As is evident in Figure 11a above, the Pacto was indeed successful in breaking the upward inflationary dynamics. However, it resulted in a sharp and sustained real exchange rate appreciation, together with a private (not public) aggregate demand boom, a huge current account deficit, and a credit boom financed by short-term capital inflows. The whole experiment turned sour when the rising loss of confidence in the peso led to capital outflows that culminated in the 1994 currency crisis. In turn, the currency crisis was followed by the 1995 banking crisis when Banco de Mexico raised domestic peso rates up to the ceiling in order to prevent inflationary pressures from picking up again.

To compare Mexico with the rest of the world, we identify a set of 17 peer countries that underwent similarly sustained real exchange rate appreciations. Remarkably enough, the real exchange appreciations were accompanied by high and declining inflation rates, which is consistent with the underlying stabilization objective of these events (Figure 14a). Whether associated with managed nominal exchange rates (as in the case of Mexico) or freely floating exchange rates (as in the case, say, of Turkey), these real appreciations were accompanied by a sharp rise in the real domestic currency interest rate perceived by foreign investors and a sharp fall in the real foreign currency interest rate perceived by domestic borrowers (Figure 14b). Thus, the appreciation triggered an interest rate “wedge” that became very attractive for foreigners to lend to domestic borrowers and similarly attractive for domestic borrowers to borrow from foreigners. This “attractive to lend-cheap to borrow” syndrome was the basis for the partly dollar-based credit boom (Figure 14c) and the resulting private sector aggregate demand boom (Figure 14d).

Source: Author’s own calculations, based on methodology presented in Appendices 3 and 4, and data from WDI and Global Financial Development Database (^{World Bank, 2023b}).

Figure 14 Monetary dominance crises: Mexico vs. the rest of the world  

In Mexico, inflation came down from a much higher level, and the aggregate demand boom and real appreciation were stronger and longer lasting, hence more damaging than in the rest of the world(Figures 14 a and 14c); while the level of credit was substantially smaller (Figure 14d), the interest rate wedge (Figure 14b) and the credit boom (Figure 14d) were remarkably similar, in both magnitudes and durations. In both cases, the macro reversal toward a currency crisis and an exchange rate depreciation, lower real interest rates, and the end of the demand and credit booms took place around nine years after the gradual appreciation.

The bottom line of the above exercise is, therefore, that Mexico’s 1994-1995 Tequila Crisis was by no means a unique event. Its amplitude may have been exacerbated by local or external factors.⁹However, as in many other countries it was essentially the result of inflation stabilization policies that gave rise to a problem of “monetary dominance”. The interest rate wedges resulting from hikes in the local currency nominal interest rates or freezes of the nominal exchange rate caused large capital inflows, demand booms, and credit booms that, in many cases, ended badly, with a mix of currency and/or banking crises. Thus, the huge real exchange rate and financial instability ended up having high economic costs. In addition to shortly interrupting the growth recovery in the wake of the North American Free Trade Agreement (NAFTA), the 95 banking crisis had a particularly enduring adverse growth legacy through its long-run impact on the Mexican financial system. Coming on top of the 1982 crisis, it undermined in a lasting way the confidence in the Mexican banking sector, sharply and durably restricting the availability of credit and raising its cost. Yet the policy adjustments leading to these unfortunate events had become absolutely essential and could no longer be postponed. The resulting price tag was another “child of the ISI years”.

3.5 The strains of growing in a China-dominated globalized world

After Mexico stabilized its runaway inflation and opened up more fully to the global world through NAFTA, Mexico’s growth remained disappointingly low on average and went through three distinct phases (Figure 15). From 2000 (when China joined the World Trade Organization) to 2009 (after the Global Financial Crisis), Mexico’s growth declined, and so did its trade (as indicated by the converging EP and DR ). From 2009 to 2018 (prior to the start of the current administration and the irruption of the COVID-19 pandemic), both growth and trade (as indicated by the diverging EP and DR ) recovered. From 2019 onward, Mexico’s trade (as reflected in its EP ) reached a plateau, and its growth declined. Three complementary questions come to mind. Why did Mexico’s growth remain, on average, so low? What factors lay behind the marked fluctuations in growth and trade? And how can the model of section 2 help provide answers to the first two questions?

Source: Author’s own calculations, based on methodology presented in Appendix 3, and data from WDI (^{World Bank, 2023a}).

Figure 15 Mexico: Trade and growth; yearly growth decomposition, 2000-2022  

Regarding Mexico’s low average growth, as indicated by the growth decomposition for the entire 2000-2022 period, it resulted from its negative DR : the country did not grow faster because the domestic economy failed to respond more successfully to the flow of incoming imports (Figure 16a). Yet, in terms of growth fluctuations (Figure 16b), the culprits were exports, both volumes ( EP ) and prices ( ToT ). In other words, the disappointingly low average growth resulted from domestic constraints, yet the growth variations around this low average were trade-induced.

Source: Author’s own calculations, based on methodology presented in Appendix 3 and data from WDI (^{World Bank, 2023a}).

Figure 16 Mexico: Full period growth decompositions, 2000-2022  

Indeed, as shown by Figure 17a, Mexico’s GDP growth closely followed the fluctuations of the country’s exports as a share of US imports. During 2000-2008, this largely reflected the inroads made by Chinese exports: the US imports from China following its accession to the WTA clearly displaced those from Mexico (Figure 17b). However, after the Global Financial crisis, Mexican exports started to rise again, but now in tandem with Chinese exports, suggesting that Mexican firms had found export niches that no longer competed so closely with China. Likewise, after the Trump-imposed tariffs and trade restrictions on China, the decline in the Chinese export share was fully offset by the rise of the export share of other Asian countries, with little change in the trend followed by the Mexican export share (at least up to 2022). This again suggests limited substitution between China and Mexico.

Source: ^{US Census Bureau (2023)} and WDI (^{World Bank, 2023a}).

Figure 17 Mexico: US vs. China growth effects, 2000-2022  

While Mexican exports no longer appear to compete so directly with China’s, the overwhelming presence of Chinese manufactures has nonetheless most likely contributed to limit Mexico’s export potential in many fields, hence its average GDP growth. Figure 18a compares Mexico’s growth decomposition for the 2000-2022 period with that of other world manufactures exporters, using the benchmarking methodology explained in Appendix 5 and normalizing the scale of the world manufacturers spectrum to match Mexico’s export pull. The two growth spectra follow very similar patterns. In both cases, they show negative ELs , largely associated with negative terms of trade ( ToTs ). By depressing the price of manufactures, China’s competition most likely affected the growth of manufactures exporting countries around the world, including Mexico.

Source: Author’s own calculations, based on the methodology presented in Appendices 3 and 4, and data from WDI (World Bank, 2023a).

Figure 18 Mexico: Growth decomposition components, 2000-2022  

However, Figure 18a also shows that Mexico’s GDP growth fell more than that of other manufactures exporters due to the much deeper contractions of its EL and DR components. Mexico’s underperforming DR reflects the limited pull exerted by its trade on its GDP. As shown in Figure 19, a 10% increase in the share of manufacture exports into total exports has only translated into a 1% increase in the value added by manufactures into GDP. While the back and forth of goods in the assembly process between each side of the frontier boosted the accounting value of exports, the value added by Mexican firms at each stage of the process remained limited.¹⁰And, independently of trade, the slow growth has also most likely reflected the lack of dynamism of the non-traded component of GDP, which in turn has had much to do with Mexico’s large informal sector, the lagging development of the country’s south, and the lack of sufficient competition in the non-tradable sector. As illustrated by the expression of GDP growth in equation (21) of the model of the previous section, a large informal sector with low productivity, v, will unavoidably slow down overall growth, even when the formal sector is making important gains in terms of productivity ( A ), the pull from foreign markets ( B ), or the multiplication of export niches ( N ).

Notes: To control for the different trends of the two variables in the figure, the averages for LA countries are subtracted from those for Mexico. Source: WDI (^{World Bank, 2023a}).

Figure 19 Mexico: Manufactures value added vs. exports  

As to Mexico’s EL contraction, it basically reflects the decline in investment. While the average manufactures exporters’ EL fell less than its ToT (i.e., the decline in aggregate demand was less than that warranted by the terms of trade-related income loss), in Mexico, it was the opposite: aggregate demand fell by more. Figure 18b illustrates why: until the post-2021 pick-up in consumption (largely reflecting the boost of direct transfers), the drop in aggregate demand in excess of the ToT income closely matched the drop in investment. In turn, the latter must have reflected both COVID and the politically induced drops in confidence.

To conclude, how could changes in total factor productivity have been responsible for the fluctuations in growth? The extremely close post-ISI correlation of TFP and GDP growth, notwithstanding the very high volatility of both variables, clearly challenges an interpretation according to which TFP solely “caused” GDP through efficiency fluctuations (Figure 8b). It is hard to see how pure efficiency shocks could have given rise to such abrupt breaks in productivity unless they were the reflection of deep underlying fluctuations in demand: on the way up, the domestic inflation stabilization boom of 1990 to 1994, followed by NAFTA induced external demand boom of 1995 to 2000; on the way down, the post-China WTO external demand collapse of 2000 to 2008; and on the way up again, the post Global Financial Crisis external demand recovery of 2009 to the present.

However, Mexico’s insertion into the global world has shaped the evolution of productivity not only over time but also across firms. While this paper only presented aggregate data, microeconomic data clearly supports the view that trade liberalization has opened a huge competition-driven gap in firm productivity between the tradable and non-tradable sectors. For example, a recent World Bank report on productivity growth in Mexico shows that the productivity of firms linked with GVCs far exceeds that of other firms (^{Iacovone et al., 2022}). This gap between the tradable sector firms that moved to their competitiveness frontier under the pressure of foreign competition and the non-tradable sector firms that continued to benefit from captive domestic demands is another unfortunate yet lasting feature of the Mexican economy, which also matches the key insights provided by the model in this paper.