To solve this problem Arellano and

To solve this problem, Arellano and Bover as well as Blundell and Bond (1998) suggest additional moment conditions by using lagged first-differenced instruments for the equation in levels. The system GMM method, therefore, combines lagged level as instruments in the first-differenced equation with lagged first differences variables as instruments in the level equation. Hence, it is expected that system GMM may cause a dramatic reduction of finite sample bias. However, our interest lies in the restricted parameters . Given the estimated coefficients in (18) and the system of equation originated by the autoregressive shocks process, we estimate the restricted parameters by minimum distance method suggested by Blundell et al. (1996). After obtaining these estimates, we test whether our empirical findings are in concordance with the theoretical model discussed previously.
Data Our data covers the learn this here now 1996–2011 and the 17 manufacturing sectors listed in Table 1. The manufacturing data is obtained from Monthly Manufacturing Survey from Brazilian Institute of Statistics and Geography – PIM-IBGE (index of the number of employees associated to production, index of real payroll per worker and manufacturing production index), Annual Manufacturing Survey from Brazilian Institute of Statistics and Geography PIA-IBGE (total number of employees associated to production and manufacturing value of production) and Aliceweb (Brazilian imports and exports by sector). Moreover, besides the information about Brazilian manufacturing sectors, our empirical analysis also requires information on Brazil\'s foreign trade partners. The sources of trade partner\'s information are: WDI – World Development Indicators of World Bank Database (consumer index price and exchange rate) as well as Penn World Table (GDP at chained PPPs and the number of employees). Based on these data sets, we construct four main variables present in our theoretical model: the import penetration coefficient, the real effective exchange rate, the Brazilian labor productivity and an index of foreign productivity. The import penetration coefficient was conventionally constructed as the ratio between imported value and apparent consumption, where the latter variable is defined as the sum of industrial production plus imports minus exports. All measures are expressed in terms of constant prices. The real effective exchange rate is calculated by weighting the real exchange rate (R$ / LCU) of the trading partners in relation to their participation on the total imports of all 17 manufacturing sectors. The construction of real exchange rate is based on consumer prices. Furthermore, we believe that the measure of labor productivity proposed above is the most appropriate in our context, given that the foreign labor productivity index used in our analysis also relies on total number of employees as input. Similar to the definition proposed for the effective real exchange rate, this variable is constructed by weighting the labor productivity of trading partners in relation to their import\'s share on the total imports of all sectors. Note that the construction of this index demands extensive information about the labor productivity of Brazilian import partners. Thus, in order to maximize the number of trading partners present in our sample, we decided to use the more conventional way of computing labor productivity, i.e., gross domestic product (GDP) per number of persons engaged. The number of partners with available information about annual hours worked by persons engaged in Penn World Table database is considerably smaller than the number of partners with available information for the number of persons engaged. A detailed description of these variables is given below, where manufacturing sectors are indexed by i and years by t. The descriptive statistics of these variables, reported in Table 2, reinforces the stylized facts about the relationship between import penetration, exchange rate and productivity. The average coefficient of import penetration in the period between 1996 and 2011 was 10.6%. Note, however, that its average growth rate was 7% per year reflecting a substantial increase of manufacturing imports participation on domestic consumption. It is a similar result when compared to the two large groups of sectors (consumption goods as well as intermediate and capital goods).