SOURCES AND METHODS

The analysis that we carry out on the causes and consequences of related lending in Mexico draws on three bodies of evidence that we have developed. The first body of evidence consists of bank financial reports. These reports were published in the financial press, and allow us to estimate bank rates of return, share prices, dividend payments, and capital-asset ratios.

The second body of evidence focuses on bank lending strategies. For two of Mexico's largest banks, the Banco Nacional de México and the Banco Mercantil de Veracruz, we retrieved internal bank records that allowed us to estimate the extent of related lending over long time periods: 1884–1911 and 1898–1906, respectively. These records were located in the Archivo Histórico Banamex and the Archivo General de la Nación, both in Mexico City. For four other large banks, we developed a data set for a cross-section of the loans they made in 1908. These records were also retrieved from the Archivo General de la Nación.¹¹

The two banks for which we have collected time series information (Banamex and the Banco Mercantil de Veracruz) accounted on average for nearly half of total bank assets. When we add the four banks for which we have cross-sectional data, our sample of banks covers two-thirds of total bank assets.

The third body of evidence focuses on a downstream industry that received related loans from the banks—cotton textiles. We note that the cotton textile industry is an ideal natural laboratory with which to study the impact of related lending on the real economy. First, cotton textiles were Mexico's largest manufacturing industry. Second, the industry was finance-dependent, but at the same time it approximated the requirements of perfect competition to an unusual degree. There were not barriers to entry produced by patents, proprietary technology, control of raw materials, advertising, branding, or control of wholesale or retail distribution. The capital equipment was easily divisible and scale economies were exhausted at small firm sizes, compared to such industries as steel, cement, paper, and chemicals. The industry was also characterized by a high degree of entry and exit. Third, Mexico's cotton textile industry was protected from foreign competition by a high level of tariff protection.

We study the effect of related lending on this industry by employing the Razo-Haber textile data set. We draw seven censuses from their data set: 1888, 1891, 1893, 1895, 1896, 1912, and 1913. These censuses are enumerated at the mill level and contain information on inputs, outputs, firm location, and ownership. We also draw state and national data on textile inputs and outputs from their data set for every year from 1891 to 1913.¹²

This state and national data allows us to make certain that the years for which we have mill-level censuses are not outliers. Table 1 presents data on the overall size and growth of this industry.

We coded the data set in order to capture relationships between bankers and textile mill owners. Specifically, we coded for bank board members who were the sole proprietors of a textile mill, a partner in a firm that owned a textile mill, or served on the board of directors of a joint stock corporation that owned a textile mill. We denote such mills as “bank-related.”¹³

Table 2 presents aggregate data on the relationships between mill owners and bankers. In 1888, 21 percent of textile mills were owned by bank directors or their close relatives. By 1913 the proportion had grown to 54 percent. The percentage of installed capacity controlled by related mills increased from 33 percent in 1888 to 80 percent in 1913.¹⁴

RELATED LENDING AND THE MEXICAN BANKING SYSTEM

In the late 1870s Mexico's banking system was so small as to be practically nonexistent. Only two chartered banks existed in the entire country. One was a branch of a British bank that operated in Mexico City and focused primarily on financing foreign trade. The other was a small American-founded operation chartered by the government of the border state of Chihuahua. The reason for the tiny banking system is not hard to divine: Mexico's nineteenth-century governments, fighting for their survival against numerous rebellions, coups, secessions, and foreign invasions, preyed upon private wealth. Bankers feared that as soon as they made their capital visible, by obtaining a charter, the government would confiscate it via forced lending.¹⁵

Spurred on by legislation enacted by Porfirio Díaz, who ruled from 1876 to 1911, Mexico's banking system expanded rapidly. The key to Díaz's banking policies was that he provided bankers with a series of segmented monopolies and oligopolies that raised rates of return high enough to compensate them for the risk of expropriation. In 1897, when Díaz's regulatory system was first codified, the entire banking system was comprised of just ten banks with total assets equal to only 12 percent of GDP. (See Table 3). By 1910, (Díaz's last full year in office before he was overthrown), there were 32 banks with total assets equal to 32 percent of GDP. Not only was this a sizable banking system by the standards of developing countries at the turn of the century, it was large by Mexico's current standards: the ratio of total commercial bank assets to GDP in Mexico in 2004 was only 33 percent.

The Banking Act of 1897 divided the banking system into three sectors: banks of issue, which emitted bank notes, discounted bills, and made commercial loans; mortgage banks, which lent long term on agricultural and urban properties; and investment banks (bancos refaccionarios), which were supposed to make long-term loans to agricultural and industrial enterprises. Only one of these sectors, the banks of issue, prospered. As Paolo Riguzzi has shown, limitations on the number of charters the government was willing to grant to mortgage banks, along with difficulties in enforcing contract rights on real property, meant that there were never more than three mortgage banks in the entire country.¹⁶

From 1897 to 1911 total mortgage bank assets averaged only 6 percent of total banking system assets. The investment banks also faltered. They were at a distinct disadvantage against banks of issue, because they could not issue bank notes. They had to compete, however, in the same markets as banks of issue, because the latter were able to skirt the laws that restricted the term of their loans (to six months) by continually renewing credits as they expired.¹⁷ As a result, there were never more than six chartered investment banks, and the combined assets (earning and nonearning) of these banks, on average, accounted for only 10 percent of total banking system assets. Moreover, the largest bank of this type (the Banco Central Mexicano) did not actually make any long-term loans at all. Instead, it operated a clearinghouse for the notes emitted by smaller banks of issue. The investment banking charter was simply a way to get around regulatory restrictions on creating a clearinghouse. In short, when we speak of Mexican banking during this period, we are really speaking about the banks of issue.

Mexico's banking system had three salient features. First, the federal government tightly regulated the number of banks competing in any market. Second, the institutions governing banking gave bank directors strong incentives to monitor one another: there was no deposit insurance; banks were extremely well capitalized; significant amounts of this capital were owned by banks' own directors; and minority shareholders had mechanisms to monitor bank directors. Third, the vast majority of lending was related lending.

Mexico's banking regulations created binding constraints on entry and competition. Only the federal government could grant a bank charter. It allowed only two banks to branch nationally: the Banco de Londres y México (BLM) and the Banco Nacional de México (Banamex). All other banks were prohibited from branching outside their concession territories, generally contiguous with state lines. With very few exceptions, the government chartered only one bank in any territory, meaning that there were typically only three banks operating in any state: Banamex, BLM, and the federally chartered local bank for that state. Special tax concessions and high minimum capital requirements safeguarded these charters.¹⁸

In addition, the law prohibited nonchartered banks from issuing notes, meaning that they could not effectively compete against chartered banks.¹⁹

Bank directors had strong incentives to monitor one another, because the banks were very highly capitalized—and the directors owned much of that capital. One of the most striking characteristics of Mexico's banks during this period was their high capital-adequacy ratios, which is to say that their stockholders had significant amounts of capital at risk. From 1897 to 1910 the ratio of equity to total assets never fell below 24 percent. (See Table 3). Even the banks of issue, which had lower capital-asset ratios because of their ability to create bank notes, had extremely high capital-asset ratios: from 1897 to 1910, the average ratio of equity to total assets never fell below 21 percent. In part, these capital ratios were driven by the legal requirement that note issues not exceed two (sometimes three) times a bank's cash on hand, or three times its paid-in capital.²⁰

Banks usually did not, however, issue notes up to their legal maximum, indicating that these high capital ratios were also driven by risk aversion on the part of bankers and their creditors (depositors and noteholders).

Independent directors appointed by outside shareholders also monitored bank directors. The 1884 Commercial Code, required the recipients of a bank charter (who became the directors) to subscribe to the first tranche of the bank's capital. Banks could later sell additional tranches to outsiders. In addition, the founding directors of a bank could (and often did) sell parts of their original stakes, often to other banks. These outside shareholders (who owned a majority of bank stock) then insisted on the appointment of independent directors who monitored the founding board members.²¹

The result was interlocking boards of directors who monitored each other.²² The presence of outside directors increased the incentives of the insiders to monitor one another: not only their capital was at risk, their reputations were as well. Outside monitoring was not a distant abstraction for the insiders. In March 1908, for example, the outside shareholders of the Banco de Jalisco, displeased with the discovery of “severe irregularities” in the bank's books, replaced the entire board of directors save Vice-president Eugenio Cuzin.²³

Mexico's bankers started out by making arm's length loans, but quickly shifted to related lending. Banamex, the largest bank in the country, received one of the first federal charters in 1884. It began by making arm's length loans. The problem was that it could not easily assess the quality of borrowers or the collateral they offered. It therefore responded by placing onerous requirements on borrowers, but the requirement led to problems of adverse selection. The history of its largest early manufacturing loans is instructive in this regard. In 1884 Banamex opened a 200,000 peso credit line (roughly US$200,000) to the Hercules textile factory for the purpose of purchasing new plant and equipment. It charged an interest rate of 8 percent and required that the loan be collateralized with 250,000 pesos worth of the factory's inventory, with the warehousing costs to be borne by the factory. Given that the factory had to finance the cost of the inventory, this implied an effective interest rate of 18 percent. Terms like these, of course, tend to attract low-quality borrowers—and this case was not an exception to that general rule. The Hercules mill was unable to make its payments. Eventually, Banamex sold a portion of the loan to a New York trading house (for only 65 percent of its face value) and recouped the rest by converting the loan into an equity interest in the mill.

In the wake of the Hercules fiasco, Banamex attempted to create mechanisms that would allow it to monitor outside borrowers directly. For example, Banamex lent 79,000 pesos to a cashmere factory in Guanajuato on the condition that it accept one of its own correspondents as manager.²⁴

The new manager soon discovered the factory to be in such bad condition that Banamex was eventually forced to take it over as part of a consortium with other creditors.²⁵ In another case, a mine and metal refinery in Guanajuato failed to make its monthly loan payment of Mx$10,600. Banamex immediately sent an auditor, and discovered the market value of the inventory was only 60 percent of what the borrowers had claimed. Banamex sold off a portion of the debt at a 20 percent discount and absorbed the rest as a loss.²⁶

Banamex's experiences with these loans induced it to shift strategy: from 1886 to 1901 all of Banamex's private (nongovernment) loans went to its own directors. After 1901 Banamex extended credit to nonrelated borrowers, but only if they satisfied one of two criteria: the borrower had a loan guarantee from the federal government (as with some railroad companies); or was either the Banco Oriental or one of that bank's directors. The reason the Banamex board gave here is instructive: most of the loans made by the Banco Oriental went to its own directors, all well-known textile magnates. Loans to them, and to their bank, were a means of investing in their manufacturing enterprises. Thus Banco Oriental loans were deemed low risk precisely because the Banco Oriental itself practiced related lending.²⁷

Related lending, in fact, appears to have been standard business practice for Mexico's banks. Data we have retrieved on the loan portfolio of the Banco Mercantil de Veracruz indicate that 86 percent of its loans to individuals from 1898 to 1906 went to the bank's own directors.²⁸

Banamex's largest competitor, the BLM (which controlled, on average, 17 percent of total bank assets), also made sizable loans to its own board members to finance manufacturing start-ups.²⁹ A cross-section of loans we have drawn for 1908 for four other banks indicate similar lending strategies. Twenty-nine percent of the Banco de Nuevo León's loans went to a single firm, owned by one of its directors. Thirty-one percent of the Banco Mercantil de Monterrey's loans also went to a single firm owned by one of its directors. Fifty-one percent of the Banco de Durango's loans went to enterprises owned by the family members of one of its directors. An astounding 72 percent of the Banco de Coahuila's loans went to a single firm owned by family members of a director.³⁰ Qualitative evidence from case studies by historians, on the Banco de la Laguna, the Banco Occidental, and the Banco de Durango concur with our quantitative analysis.³¹

Precisely because bank insiders had much at stake, the loans that they granted to one another tended to be made on fairly conservative terms. First, they often lent to directors as individuals, rather than to their enterprises. Second, they typically required that credit lines be secured by liquid assets, such as cash, government securities, or corporate securities, which were physically held by the bank. When lines of credit were secured by a cash deposit, they were obviously not 100 percent secured. Nevertheless, the existence of a security deposit substantially raised the cost of defaulting and lowered the cost of collateral repossession: the bank simply kept the security that it already held in its vault.³²

The way in which downstream firms established a relational tie to a bank demonstrates that related lending resulted from the bankers' inability to enforce arm's-length contracts. Mexico's bankers did not choose to lend to a particular textile company, and then demand a seat on that company's board of directors. Rather, the textile mill owner would obtain a bank charter, sell shares in the bank to outside investors, issue bank notes, and then lend the notes to textile mills that he already owned (or, in some cases, found an entirely new mill). Of the 34 textile mills that switched from being nonrelated to being bank related between 1888 and 1912, 33 were owned by textile entrepreneurs who later became bankers. In short, bankers did not look at their banks as independent credit intermediaries in the textbook sense of the term. Instead, they looked at them as the investment arms of their widespread commercial and industrial interests.

RELATED LENDING AND THE PERFORMANCE OF THE BANKING SYSTEM

Did Mexican bankers use related loans to loot their own banks? One would imagine that they had strong incentives to do so. Mexico was hit by an external shock in 1908 that drove down the prices of its major export commodities between 14 and 56 percent (depending on the product). The decline in prices caused mineral and agricultural producers to drastically curtail production by between 20 and 64 percent (depending on the product), which in turn caused the demand for manufactured goods to fall by 9 to 20 percent (depending on the product). The decline in Mexico's export and manufacturing sectors soon threatened the banking system. Interest rates on commercial paper rose from 8 percent to 10 percent, net new lending dropped to zero, and bank rates of return fell considerably as borrowers began to default.³³

In response to the crisis, the government quickly organized a rescue. In September 1908 the federal government chartered the Caja de Préstamos para Obras de Irrigación y Fomento de la Agricultura, or “Lending Institution for Irrigation Works and Agricultural Development.” Despite its name, the Caja made no direct loans to farmers or irrigation companies; nor did its creators within the government intend it to. Rather, the Caja purchased existing agricultural loans and bank-issued mortgage bonds from the banks, with the explicit intention of injecting liquidity into the banking system.³⁴

Financing for the Caja came from the issuance of 44.5 million pesos of government-guaranteed bonds in Europe. In addition, the government required Mexico's four largest banks to purchase 10 million pesos of Caja shares, 25 percent of which they were not permitted to resell.³⁵

The banking crisis, in conjunction with a government-led bailout, gave bank directors incentives to loot their own banks. First, the crisis gave the directors of weak banks incentives to tunnel as much as they could before the banks failed anyway. Second, the bailout gave the directors of stronger banks incentives to tunnel, as they would be protected from some of the consequences of their malfeasance.

Widespread tunneling would have three testable implications for the banking system. First, tunneling would cause bank failures resulting in losses for depositors and noteholders, not just for shareholders. Second, tunneling directors would divert profits into loans for themselves and their other enterprises, reducing the value of bank profits paid to shareholders. Third, shareholders would judge that bank stock was risky and discount its value accordingly.

The evidence does not indicate that any of these outcomes obtained. Mexico had, in fact, a remarkably stable banking system. As shown in Table 3, the number of reporting banks and total bank assets increased steadily throughout the period under study. The only downturn was in 1909, when, as a result of the crisis of 1908, seven small banks of issue failed (two were later rechartered as investment banks, and the others were purchased by larger, more solvent banks). Nevertheless, depositors do not seem to have believed that their wealth was not at risk: bank deposits grew in both absolute and relative terms from 1908 to 1909. (See Table 3). In fact, in 1897, deposits (exclusive of those securing credit lines) accounted for only 2 percent of total bank assets. By 1910 they accounted for 16 percent.

One might argue that although the banking system was stable, the directors were still able to extract resources from outside shareholders. That hypothesis, however, is not consistent with the fact that Mexican banks were extremely profitable enterprises for their shareholders, contrary to what we would expect to see in the presence of tunneling. The real return on the book value of equity in 1901–1912 for the entire banking system was 12 percent. These returns were not driven by the profits earned by a few large banks: the unweighted average real return-on-equity for all banks was 10 percent per year. Moreover, the evidence does not indicate that the 1908 financial crisis had a long-term impact. The rate of return on equity fell in 1908, but it returned to pre-1908 levels in 1909.

Mexican banks returned these high profits to shareholders by paying out high and regular dividends. In fact, over the 1901–1910 decade, the banks paid out almost all of their profits to shareholders in the form of dividends.³⁶

Steady dividends translated into high returns from the ownership of banking stock. As Table 4 shows, someone who purchased an index of banking stock weighted by market capitalization would have earned an average real return of 9 percent per year. Our estimate of market returns is not driven by the high returns available from owning the stock of the largest banks: an investment strategy based on purchasing equally sized stakes in all the banks would have yielded a slightly higher average annual real return of 10 percent. The returns available to investors in Mexican banking stock were, in fact, more than twice those available from investing in the Dow Jones Industrials. (See Table 4). They were also no lower, on average, than those available from investing in a portfolio of Mexican manufacturing, mining, or land development firms.³⁷

Were large dividends (and hence high returns) a sign that directors were in fact tunneling? It is possible that directors tunneled by paying out all of the bank's cash flow to themselves as dividends. This interpretation is not, however, consistent with the evidence about the identity of bank shareholders. Banks did not grow by taking deposits (as Table 3 shows, deposits never accounted for more than 16 percent of total assets), but rather by issuing additional shares. Lists of the purchasers of these additional shares, which we have for the Banco de Jalisco, the Banco Mercantil de Veracruz, and the Banco del Estado de México, indicate that the majority of these additional shares (57, 85, and 93 percent, respectively) were bought by outsiders.³⁸

Thus, paying out high dividends would have been a very inefficient way to tunnel: most of the dividend payments would have gone to outside shareholders.

Did bankers tunnel through some other mechanism? It might be that high dividends paid to outsiders were simply compensation for the risk of tunneling. This argument is not, however, consistent with the evidence regarding the price of bank shares. If stockholders feared that the directors were tunneling, they would have discounted the price of shares. We therefore estimated two measures of the degree to which investors discounted shares: market-to-book ratios and the average yield on banking stock. Neither measure is consistent with the hypotheses that banking stocks were heavily discounted.

Table 5 shows the average (weighted and unweighted) market-value-to-book-value ratio for Mexican banks in 1900–1911. Bank stock traded at an average premium of 33 percent over its book value. Moreover, the crisis of 1908 does not appear to have had a major effect on how investors valued their assets. In fact, only one bank, the Banco de Michoacán, was valued at less than its book value in the years 1909 and 1910.³⁹

The data on banking yields are also inconsistent with the hypothesis that investors heavily discounted banking stock. As Table 6 shows, between 1901 and 1912 the difference between the yield on Mexican banking stock and Mexican government bonds dropped from 4.4 percentage points to 3.0 percentage points. Moreover, the 1908 financial crisis does not appear to have had a major effect on this overall trend.

The evidence, in sum, does not indicate that Mexican bankers tunneled either before or after the crisis of 1908. In fact, there is evidence that during the crisis, bankers propped up banks with which they shared interlocking directorates. For example, the Banco de Jalisco rescued the Banco de Aguascalientes, and the Banco Oriental purchased and merged with the troubled Banco de Oaxaca and the Banco de Chiapas. Banamex assumed responsibility for the note issues of the Banco de Michoacán and Banco de Campeche when they faced runs and could no longer support their note issues. Banamex also aided two troubled banks on the Yucatán peninsula in merging.⁴⁰

Did the banks succeed in weathering the crisis through the expedient of passing off their bad related loans to the Caja de Préstamos? The evidence indicates that in 1908 the banks did transfer related loans to the Caja, but these loans were chosen by the government precisely because they were high-quality loans. In point of fact, the Caja de Préstamos may be the only banking bailout in world history to have made money. The assets held by the Caja (loans and mortgage bonds transferred from banks) consistently generated a positive cash flow, allowing the Caja to not only pay the interest due on its bonds, but to pay a 10 percent annual dividend on its outstanding share issues. We calculate that the Caja generated a nominal return to all claimants of its assets (bondholders and shareholders) of 4 percent in 1909, 7 percent in 1910, and 8 percent in both 1911 and 1912.⁴¹

DID RELATED LENDING MISALLOCATE CAPITAL?

One might argue that even though bankers did not do enough looting to jeopardize the health of the banking system, they may nonetheless have diverted resources to their own, relatively inefficient, enterprises. In order to test this hypothesis we turn to our panel of textile mills. If bankers were using their textile mills to channel resources from the banks to themselves, then we would not expect bank-related mills to grow. Textile mills would simply be mechanisms to extract the wealth of the bank.

When we look at the growth in the size of mills, however, we find precisely the opposite: not only did bank-related mills grow, they grew faster than their nonrelated competitors. In Table 7 we calculate the growth rates of mills that existed (and did not switch between the bank-related and nonrelated categories) across various census periods. In each period, we find that mills that were bank-related outgrew mills that were not bank-related.

A somewhat weaker hypothesis about tunneling would suggest that bankers may have used their banks to support their own, relatively inefficient firms. In this view, bank-related mills may have been productive enterprises (rather than zombie firms whose purpose was to extract bank resources), but they would be less productive than their competitors. If this hypothesis holds, it implies that related lending misallocated capital.

As a first step in testing this hypothesis, we estimate a series of probit regressions, where the dependent variable is whether a mill was bank-related, and the independent variables are the characteristics of mills. If credit was misallocated, then we should be able to predict which firms were bank-related based on their performance characteristics. We measure these characteristics as profitability and technical efficiency. We begin our analysis with the 1893 manufacturing census, because it provides extremely detailed data on the costs and volumes of all inputs and outputs. We employ two measures of mill performance: operating margins, and labor productivity. We add controls for mill size and age.

Regardless of the specification, our qualitative results are the same: there were no statistically significant differences between bank-related and nonrelated mills. (See Table 8.) We estimate similar probits—whose results are not reported here—for 1888, 1891, 1895, 1896, 1912, and 1913. The only difference is that these probits do not include a variable for operating margins because of data constraints. Not a single one of these probits, regardless of the specification employed, found any relationship between technical efficiency and bank relation. The probits imply, in short, that there was a dead-heat between bank-related and nonrelated mills in terms of their technical efficiency.⁴²

It is conceivable that the probit results on individual cross sections are too blunt an instrument to pick up small, but consistent, differences in productivity across censuses and mill types. We therefore estimate a time series, cross-sectional regression on labor productivity, and report the results in Table 9.⁴³

We control for mill age, location, bank-relation, and whether the mill was publicly traded.⁴⁴ Our results do not support the hypothesis that bank-related mills had lower labor productivity than their nonrelated competitors. None of the coefficients on bank-relation are significant.

Does the dead heat in labor productivity between bank-related and nonrelated firms mask capital misallocation? Bank-related firms could have used more capital per worker, and thus similar output per worker even though their output per unit of capital was lower. We therefore re-estimated our regressions on labor productivity, controlling for capital intensity and mill size. That regression produced the same qualitative results as our earlier regression: none of the coefficients on bank-relation come up significant.⁴⁵

One may worry that technical efficiency is the wrong metric by which to judge the productivity of bank-related and nonrelated textile mills. What matters is economic efficiency: efficient mills thrive and grow; inefficient mills go out of business. If so, and if bank-related mills were less efficient, then bank-related mills would fail more frequently than nonrelated mills. In order to test this hypothesis, we employ a Cox maximum-likelihood proportional hazards model to estimate the effect of bank-relation on the probability of mill failure. Mills are defined as “failed” when they disappear from the subsequent census and never reappear. All coefficients (and standard errors) are transformed into hazard rates.

Our findings, presented in Table 10, are not consistent with the hypothesis that bank related mills were less economically efficient. In fact, we find exactly the opposite: bank-related mills were only 23 percent as likely to fail as their nonrelated competitors. This result is robust to the addition of conditioning variables for mill size, labor productivity, and age.

The Cox hazard model also suggests that being big was endogenous to being bank related. Bank-related firms lived longer, and therefore grew larger. This is consistent with our finding that bank-related mills grew much faster than their competitors, reported in Table 7.

MARKET STRUCTURE

If bank-related firms grew at a much faster rate than their nonrelated competitors, then it follows that there should have been big size differences between bank related and nonrelated mills. Table 11 is unambiguous on this point: in 1888, bank related mills were, on average, almost twice the size of unrelated mills; by 1913, they were nearly four times as large.

It also follows that the market structure of the textile industry should become more concentrated as the proportion of bank-related mills grew. In order to measure concentration we aggregate mills into firms, and estimate four-firm concentration ratios and the Herfindahl index.

In order to determine how low concentration would have been in the absence of related lending, we specify three counterfactuals. The first compares Mexico to itself over time. Cotton textile manufacturing was an industry characterized by constant returns to scale technologies and the absence of entry barriers. We should expect that, in the absence of related lending, as the industry grew, concentration should have fallen. The second compares Mexico to other countries that had large textile industries, but which did not have Mexico's banking system. We focus on three countries: the United States, Brazil, and India.⁴⁶

The third, following John Sutton, compares the Mexican textile industry's actual market structure to a hypothetical, fully competitive industry, in which the market structure was a function solely of industry size and a stochastic growth process.⁴⁷

The results of all three experiments, reported in Table 12, indicate that the Mexican cotton textile industry was “too concentrated.” First, concentration in Mexico actually increased over time, even though the industry was growing quickly. (In the United States, Brazil, and India, concentration fell or remained stable as the textile industry grew.) Second, the Mexican cotton textile industry was much more concentrated than the U.S., Brazilian, or Indian cotton textile industry. Third, the Mexican cotton textile industry showed much higher four-firm ratios compared to the ratio that would be expected in a perfectly competitive market, given the number of firms in the industry.