3. CAPITULO III ANÁLISIS DE DATOS Y RESULTADOS
3.3 ANÁLISIS DE LAS ENTREVISTAS
3.3.3 Logros obtenidos por los participantes de los procesos de Formación Teatral
In the empirical analysis I use different measures of industry characteris- tics. The most important data are on financial constraints. The primary data source is the Worldscope Database (2008) by Thomson Financial. This database provides detailed balance sheets and income statements of 8915 listed US companies over the period 1995-2004. This database is used to calculate measures of financial vulnerability following the methodology de- veloped by Rajan and Zingales (1998) and Braun (2003).30
Rajan and Zingales (1998) identify a particular channel through which credit affects economic activity, i.e. via the dependence on external finance. They argue that not every firm (industry) benefits from a better supply of credit. Only those who are “naturally” more dependent on external finance will be able to gain. However, there is a typical identification problem in cap- turing empirically the (natural) demand for credit because only equilibrium quantities are actually observed. Therefore, country-specific firm-level mea- sures of the credit usage may constitute poor measures of financing needs. For example, in economies with poor credit supply firms are likely to re- spond endogenously with less external financing. The economic effects of an
30These measures have been extensively used by many researchers, e.g. Manova (2006)
for export activity, Klapper et al. (2006) or Aghion et al. (2007) for the market entry of new firms, Kroszner et al. (2007) or Dell’Ariccia et al. (2005) for the effect of banking crises on growth of industry value added, etc., when looking at the effects of credit constraints.
increased credit supply would therefore be underestimated. A solution to this problem arises only if the supply of credit is nearly perfectly elastic, so that the demand for credit determines the actually observed amount. Rajan and Zingales argue that publicly listed US companies are good candidates because they face such conditions for two reasons. First, among all countries the US has one of the most developed financial systems. Second, within the US publicly quoted companies face the least barriers to external funding. In order to use this measure for all other countries the assumption made by Rajan and Zingales is that the natural demand for credit of a company within an industry depends mainly on the technological characteristics of the production process of that particular industry.31 This assumption does not mean that the financial dependence of industries does not vary across coun- tries. It requires only that the relative ranking of industries with respect to this measure does not change across countries.
The second alternative measure of tight credit constraints is introduced by Braun (2003). He argues that in the presence of incomplete contract- ing (captured by the level of financial development) higher asset tangibility serves as collateral (protection) to banks (outside investors). Only those firms that have (naturally) more tangible assets that can serve as collateral will be able to access credit markets and gain from a higher credit supply. The same problem of identification arises when one uses country-specific firm-level measures. In economies with poor credit supply companies are likely to compensate endogenously for this country’s disadvantage by hold- ing more tangible assets on their balance sheets. This would underestimate the economic effects of the availability of credit. To overcome this problem
31For Western European countries this is clearly a good assumption. I argue that this
is mainly the case for Eastern European countries as well for the following reasons. After the fall of the “iron curtain” the old production technologies were replaced with new ones because firms were not competitive compared with the Western European imported goods. Moreover, the transition process to a market economy in Eastern European countries was guided by technology transfer from Western Europe through inward FDI. This was particularly important after 1995, which is the start of my sample period.
again publicly listed US companies are good candidates for “estimating” the natural financial “vulnerability” of a particular sector.
Rajan and Zingales (1998) and Braun (2003) calculate the variables only for the manufacturing sector, which incorporates up to 36 three-digit indus- tries. Moreover, these measures are based on data for the period before my starting year of 1995. Since I am considering up to 179 manufacturing and
non-manufacturing industries and the financial conditions in the 1995-2004 period are likely to be different from those in previous periods, I construct these two measures with new data.32 Rajan and Zingales’s measure is called
external financial dependence of industry j (F inDepj). It represents the
fraction of a firm’s investment (capital expenditures, CapExp) that is not financed by the internal net cash flow (CashF low) from operations. The financial dependence of industry j is calculated by the following formula:
F inDepj =M edianj P t
(CapExpijt−CashF lowijt)
P t CapExpijt
First, I obtain a firm-specific, time-invariant ratio of the capital expendi- tures not covered by internal cash flow as a ratio of total capital expenditure. This is achieved by summing for each firm i the use of external finance over time and dividing it by the sum of CapExp over the same period. This strategy reduces the effects of the business cycle on these flow variables. In order to obtain an industry-specific measure, Rajan and Zingales consider the median value of the distribution of all ifirm-specific measures within an industryj. The median has an advantage over the sample mean because it is not sensitive to the presence of big outliers in the data. The higher F inDep
is, the more credit is demanded by the firm/industry.
32Rajan and Zingales (1998) and Braun (2003) use the Compustat database of Stan-
dard and Poors for variable calculation. However, as reported by Ulbricht and Weiner (2005), after the year 1998 Worldscope’s firm coverage is significantly higher than that of Compustat, which leads to better representativity.
The measure of asset tangibility (“asset hardness”) is calculated following Braun (2003): T angj =M edianj ( 1 t X t
(N et P roperty, P lant and Equipment)ijt
T otal Assetsijt
)
Again, for each firm i an asset tangibility measure is constructed by di- viding the “Net Property, Plant and Equipment” balance sheet item, which represents the tangible assets of the firm, by the total assets and averaging it over the sample period. The industry tangibility measure is the median value of the firm-specific ratios. The higher this measure is, the more collateral can be pledged to outside investors (banks) and the more likely it is that credit can be granted.
The second and third columns of Table 3.15 in Appendix B include the calculated measures of credit constraints. For example, a highly financially dependent industry is engines and turbines (SIC 351), while the cigarettes industry (SIC 211) is less dependent on external finance. Further, the mis- cellaneous personal services industry (SIC 729) has a relatively low level of asset tangibility, while hotels and motels (SIC 701) have a very high ratio of tangible to total assets.
The AMADEUS database is used to compute other variables. For ro- bustness and comparability checks I construct two firm-level asset tangibility measures. Again, whereas the industry-specific US measure is exogenous, the firm-specific one is likely to be endogenous. The first one, Tang1, is defined as the ratio of tangible fixed assets to total assets averaged over time. The second one, Tang2, is the ratio of tangible and financial fixed assets to to- tal assets again averaged over the years 1995-2004. Unfortunately, I cannot build a firm-level financial dependence measure since AMADEUS does not report capital expenditure figures.
For robustness checks I construct two industry-specific asset tangibility measures using UK data as a benchmark from AMADEUS. The reason lies in the fact that the United Kingdom is considered to have one of the most developed financial systems in Europe, which is comparable to that of the US. Hence, the Rajan and Zingales’s (1998) and Braun’s (2003) assumption of nearly perfectly elastic credit supply is likely to hold for the UK. The first measure is the ratio of tangible fixed assets to total assets averaged over time for the median UK firm. The second one is the median value of the ratio of tangible and financial fixed assets to total assets for UK enterprises. The correlation of the US and UK asset tangibility measures is high at around 0.67, which indicates that indeed technological characteristics of the partic- ular industry common across countries determine the measures.
Table 3.2: Firm- vs. industry-level asset tangibility measures
Measure Eastern Europe Western Europe
Tang (US) 0.26 0.24 Tang1 (UK) 0.23 0.22 Tang2 (UK) 0.27 0.25 Tang1 (firm-level) 0.37 0.20 Tang2 (firm-level) 0.39 0.24 Observations 130,726 1,410,871
Notes: The table shows by country group the means of the variables using the estimation sample from section 3.5.
In Table 3.2 I compare the firm-level asset tangibility measures with the exogenous industry-specific measures using the US or the UK as reference countries. The expectation is that in countries with unsound financial condi- tions like the Eastern European countries enterprises will endogenously react to these conditions by holding more collateralizable assets than are needed for technological reasons. Obviously, the measures for Western European
countries are very similar. Tang (US), Tang2 (UK) and Tang2 (firm-level) are almost identical and equal to around 0.24. However, for Eastern Europe the firm-level tangibility measures are much higher than the “estimated” ones. Tang (US) and Tang2 (UK) are on average equal to 0.26 compared with 0.39 for Tang2 (firm-level).33 Therefore, this evidence indeed supports the argument that firm-level financial constraint measures are not suitable for consistent estimation of their effects on TFP or export participation.
In the next step two industry-level measures of human capital inten- sity (HumInt) and physical capital intensity (CapInt) are calculated using AMADEUS data on UK firms as a benchmark. These measures are used in the empirical analysis in order to make sure that the estimated effects of financial constraints do not capture other sources of comparative advantage. Following Braun (2003), Hur et al. (2006) and Manova (2006) among others, human capital intensity is proxied by the average wage (in million USD per employee) of the median firm in each UK industry. The intuition is that since high-skilled workers are more productive than low-skilled ones then the relative wage compensation for skilled employees will be higher. Physical capital intensity (otherwise called the capital to labor ratio) is defined as the value of fixed assets (in millions USD) per employee for the median UK firm. These measures are, like the others, time-invariant and industry-specific. The United Kingdom is used as a reference country because it has one of the most liberalized (flexible) labor markets in Europe. Since firms face nearly per- fectly elastic factor supply, the relative factor prices (wages) would therefore capture the true “technological” demand for skilled and unskilled labor and physical capital.34 Again, like for the financial constraint measures, even
33A further comparison of the industry-level measures across Eastern and Western
Europe suggests that in the sample there are on average more tangible industries selected in Eastern Europe. This can be explained by the survival of more financially healthy industries in countries with unsound financial conditions. This hypothesis is in addition supported by the observation that in the same sample Eastern European industries are
also less financially dependent than Western European ones (0.004 vs. 0.05 on average).
though these two quantities are not required to be exact across countries, the assumption needed is that the ranking of industries remains unchanged across the European countries in the sample.35 The last two columns of Table 3.15 present the calculated measures of physical and human capital intensi- ties. For example, a low-skill-intensive industry is knitting mills (SIC 225), while firms from the communications equipment (SIC 366) industry employ more skilled workers. Further, personnel supply services (SIC 736) have a relatively low physical capital intensity, while air transportation (SIC 451) companies have a very high physical capital to labor ratio.
Table 3.16 in Appendix B calculates the pairwise correlation between the four industry-specific variables. Interestingly, asset tangibility is only slightly negatively correlated with financial dependence, which is not significantly dif- ferent from zero. This suggests that they capture two different characteristics of financial constraints, which can be jointly exploited as potential determi- nants in a regression framework. The physical and human capital intensities are not significantly correlated with each other, either.
Finally, the World Integrated Trade Solution (2009) database is used to obtain tariff rates, which are shown to be important determinants of trade patterns and productivity differences. It combines several data sources in an integrated solution. The tariff data come originally from the UNCTAD TRAINS (Trade Analysis and Information System) database. For each re- porter country and in each three-digit manufacturing and construction in- dustry within this country I obtain a simple average or a weighted average of the import tariffs. Tariffs are effectively applied tariff rates (the lowest avail- able). The partner countries (exporting countries) for which these tariffs are calculated are all other countries (the rest of the world). These tariffs then of exposure to international trade and rigid wages for unskilled labor would artificially drive the demand for skilled labor and its relative wage. Hence, countries with regulated labor markets are not suitable for “estimating” the “natural” factor intensity of an industry.
35For robustness checks I recalculate the HumInt measure using French, Spanish and
capture the intensity of import competition in each of the country-industry pairs.