*2.1. Policy Background*

Inaugurated in 1990, the Collor de Mello administration implemented a series of economic reforms aiming to reintegrate Brazil into world markets. Such reforms eliminated hurdles in the foreign currency market and implemented changes in the trade policy to reduce protection levels (Baumann 2001). In fact, the new president suddenly and drastically reduced the non-tariff measures of protection (NTMs) and also scheduled nominal tariff cuts that were heterogeneous across industries, to be implemented between 1990 and 1994 (Kume et al. 2003, 2008). The protection of the manufacturing sector declined substantially from a 40 percent average tariff in 1989 to a 17 percent average tariff in 2000. Accordingly, manufacturing imports grew by more than 200 percent between 1990 and 2000, and the import penetration in manufacturing almost tripled, growing from its initial level of 5.7 percent in 1990 to 14 percent in 2000.

The trade protection reforms implemented in the 2000s by the da Silva and the Roussef administrations were considerably different than the reforms of the 1990s (Paz 2018). In 2004, the da Silva administration granted market economy status to China in November of 2004. This decision came in the aftermath of China's accession to the WTO in 2001 (Chandra 2014). It conceded most favored nation tariffs to Chinese imports and reduced the ability of the Brazilian government to impose safeguards countervailing duties and anti-dumping against Chinese exporters. These reforms resulted in a greater openness of the Brazilian economy that is illustrated by an increase in the overall manufacturing import penetration from 14% in 2000 to 18% in 2012.

#### *2.2. Data Description*

The database assembled for this study comprises information on international trade flows, on Brazilian national accounts, and on household surveys. The bilateral international trade data are downloaded from the Comtrade system (United Nations 2003) for the period between 2000 and 2012 using the six-digit 1996 version of the harmonized system. These are used to build industry-level Brazilian imports from China and from the remaining countries of the world (hereafter called ROW) series, and also the excluded instruments, as discussed in the next section.

The Brazilian national accounts data (IBGE 2015, 2016) provide information on total output level, on employment level, imports and exports at IBGE's level 56 industry classification. The worker-level data come from the PNAD-Pesquisa Nacional por Amostra de Domicilios (Brazilian household survey) and from the Brazilian demographic censuses of 2000 and 2010, as the PNAD household surveys do not take place in census years. These

surveys ask similar questions about workers' observable characteristics such as earnings, hours worked in a week, job formality status, industry affiliation, education, gender, age, marital status, race, and Brazilian state of residence. The period under analysis ends in 2012 because in 2013 a major change in the social security contribution incidence was enacted by Federal Law 12546. PNAD's methodology substantially also changed in the 2015.

This study only considers employed workers. Employers, self-employed, and unemployed people are excluded from the analysis. Moreover, an informal job is defined as the employment relationship in which the employer does not comply with the social security contributions, as in Paz (2014). The different industry classifications used in the original data were harmonized by means of correspondence tables from the CONCLA-IBGE website (https://concla.ibge.gov.br/, accessed on 19 April 2021). The classification used by the National Accounts data is the most cursory in this study. Hence, it dictated the industry classification used here, which consists of a modified version of the Nível 56 classification with 26 manufacturing industries.

### *2.3. Raw Data Patterns*

This subsection starts with an overview of the trade relationship between China and Brazil and of the import competition experienced by the Brazilian manufacturing sector. In 2000, Chinese imports made up 2.7 percent of Brazilian imports, which ranked it as the tenth largest exporter to Brazil. Yet, these figures were radically different in 2012, when Chinese imports amounted to approximately 20 percent of Brazilian imports, of which more than 90 percent are manufactured goods. Moreover, other labor abundant countries like India, Indonesia, and Vietnam accounted for less than one percent of Brazilian imports in this period (Paz 2019a). Figure 1 shows that imports from China and from high-income countries account for most Brazilian imports. Thus, ROW imports are mainly driven by imports from high income countries. Figure 1 also displays a growing Chinese share of Brazilian imports and a shrinkage of the high-income countries' share. Indeed, both the Chinese and the ROW volume of imports grew over time, albeit at a different pace. Hence, the Brazilian experience in the 2000s cannot be summarized merely into a case of substitution of suppliers.

**Figure 1.** Brazilian manufacturing imports by source. Notes: L. America–Latin America. Other LDC–developing countries other than China and those in Latin America.

The import competition measure of Brazilian firms used in this study is the industrylevel import penetration. This is the ratio between imports and the apparent consumption (production plus imports minus exports). In contrast with tariffs, the import penetration

also captures the effects of NTMs, such as import licenses, quotas, and anti-dumping duties. Moreover, the import tariff between 2000 and 2012 shows little variability across industries and over time, despite the large variations in import volumes and in import penetration (Paz 2018). As a result, import tariffs are not recommended for the analysis carried out in this study.

Turning to the descriptive statistics at the industry level, Table 1 reports the 2000 and the 2012-level, the average, and the standard deviation of ROW and Chinese import penetrations. We can see that 19 out of 26 industries exhibited an increase in import penetration, and in most of these cases the increase was in excess of 20 percent. Most importantly, these industries employ more than half of the workers in manufacturing. Additionally, the Chinese import penetration grew in 24 out of 26 industries. Although the average Chinese import penetration is smaller than the ROW import penetration, the former has a significantly larger coefficient of variation due to the increase in Chinese participation in the Brazilian imports in this period. Together with Figure 1, these statistics indicate that this growth in Chinese import penetration was not simply a case of substitution of ROW imports. Table 1 also displays descriptive statistics for industry-level informality share. They show shares below 5 percent in industries such as automobiles, steel, and biofuels, while industries such as apparel and wood products exhibited an informality share close to 30 percent.

**Table 1.** Industry-level trade exposure and labor market characteristics of manufacturing industries in Brazil during 2000–2012.


Notes: Informal workers are those without social security coverage. Number of observations is 312. Household survey weights used for informal share.

Table 2 presents industry-level average characteristics of manufacturing workers according to their formality status. The hourly wage consists of the monthly wage divided by 4.3 times the number of hours worked in a week. The inflation adjustment is conducted according to Corseuil and Foguel (2002) using inflating factors from IPEADATA (2017). The natural logarithm of the real hourly wage also exhibits heterogeneity across industries, being larger in skilled-labor intensive industries. These figures also indicate that formal workers earn substantially higher hourly wages and are more likely to be male. Formal workers are slightly older than informal workers. The industry-level average education level shows substantial cross-industry variation, which tend to be higher in skilled-laborintensive industries such as pharmaceutical products. Even though formal workers are more educated on average, the difference in the average of years of schooling range from half year in footwear to three years in other chemical products. This study now turns to the presentation of the theoretical framework that motivates the description of the empirical methodology used here.


**Table 2.** Formal and Informal workers' average characteristics at industry level.

Notes: Number of observations is 669,966. Informal workers are those without social security coverage. Household survey weights used.
