Productivity in the steel
industry
Denise Groot
10551999
Economics and Business
Specialization: Finance and Organization
Field: Organization
In combination with an internship at Tata Steel IJmuiden
Supervisor: A. Kiss
Statement of originality
This document is written by student Denise Groot who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Import competition from China and firm-level productivity in
the European steel industry
Denise Groot
University of Amsterdam 17 June 2016
Abstract
This paper estimated the effect of Chinese steel prices on the productivity of European steel manufacturers in the period 2009 to 2014. First I estimated the Cobb-Douglas productivity function, using the OLS method. Using the OLS estimation of the coefficients, I calculated the firm-level Total Factor Productivity. Thereafter, I analyzed the effects of changing Chinese steel prices on the productivity of European steel manufactures, using the fixed effects method. Two key results emerged. First, the productivity of the average firm increased by 1.84 percent when Chinese steel prices dropped by 10 percent. Secondly, allowing for firm heterogeneity revealed that Chinese steel prices did not affect firms differently depending on their initial productivity. Interpreting these results should be done with care. I did not control for selection bias nor simultaneity bias, further studies should use the Olley-Pakes method to control for these facts. Moreover, they should use different and larger samples, and cover longer periods to verify my results.
I. Introduction
For the past ten years, the steel industry1 has been characterized by rising steel production in China. As shown in graph 1, in 2014 the Chinese crude steel production was almost seven times as great as the production in 1999. During this period, China has become the major steel exporting country of the world2. However, the Chinese steel industry faces overcapacity now. Excess steel supply led to decreasing Chinese steel prices, shown in graph 2. Due to these lower prices, importing steel products from China becomes more appealing for the rest of the world. Therefore, domestic steel producers outside China face fiercer import competition. Many of those steel makers blame the low price of Chinese material for the collapse of their steel market. A lot of governments confirmed this and increased the import duties on Chinese steel products to protect their domestic steel markets. The effect of trade protection and trade liberalization on firms has triggered a substantial amount of research, especially to the effect on the productivity of firms. For example, Pavcnik (2002) investigated the effects of liberalized trade on plant productivity in the case of Chile. She showed that when import competition became fiercer, due to a 10 percent reduction of import tariffs, the within-plant productivity increased by 2.8 percent.
1
I am currently doing an internship at Tata Steel IJmuiden, a company that is active in the steel industry. In this paper I therefore focused on the steel industry.
2
The world steel association had ranked each country based on their net-export levels. In 2014 China’s net-export were the highest with 78 million tonnes, followed by Japan with 34.6.
0 5 10 15 20 1999 2001 2003 2005 2007 2009 2011 2013 To n n e s (x100. 000) Year
Crude steel production
European union (28) China world 0 100 200 300 400 500 600 700 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 € /to n n e Year
Chinese steel prices
Note: the prices are a year average of all
the different geographical submarkets in China in euros.
Graph 1: Crude steel production3 Graph 2: Chinese hot rolled coiled
steel prices4
The current literature about trade liberalization and firm-level productivity do not control for fluctuating product prices. Moreover, they assume that prices stayed flat during their investigated time period. However, the Chinese steel prices have substantially changed. Besides, the import tariffs in the European market for Chinese products did not significantly change in the period from 2009 to 20145. The purpose of this paper was to contribute to the literature about the effect of import competition on firm-level productivity. I have examined whether or not decreasing Chinese steel prices, and thus rising import competition, induced firm-level productivity gains in the European steel industry between 2009 and 2014.
My empirical strategy in part followed Amiti and Konings (2007). They estimated the productivity gains from reducing import tariffs on final goods and intermediate inputs. First they estimated the Cobb Douglas productivity function of a firm, using the Ordinary Least Squares (OLS) method as well as the Olley and Pakes method. Thereafter, they estimated the productivity gains from trade liberalization using the fixed effects approach.
The Olley-Pakes approach is based on dynamic optimization of firms. They assume that a firm maximizes its productivity and based on this productivity the firm chooses an optimal investment level (Olley and Pakes, 1996). Pakes (1994) specified the conditions under which the investment function is monotonically increasing in productivity, which makes it possible to invert the investment function and gives an expression of productivity as a function of capital and investment. This approach makes it possible to control for both selection bias and simultaneity bias. My data set was limited by the reporting standards of material costs. If firms that report materials costs differed from those firms that did not report, I would had a selection bias. The input demand of firms depend on their knowledge of their productivity,
3
The crude steel production comes from the statistic archive of the World Steel Association. 4
Chinese hot rolled coiled steel prices come from Bloomberg. 5
Tariff data is retrieved from World Trade Organization (WTO). The 6-digit code products are selected based on the descriptions of products which are included in the 4-digit NACE rev.2 code 2410.
which was unobserved by me, this could lead to the simultaneity bias. If firms with higher productivity buy more inputs, my estimations will be biased. However, due to the complexity of the Olley-Pakes method, I accepted the problems of selection bias and simultaneity bias and estimated the Cobb-Douglas function using the OLS method. Thereafter, I estimated the effects of Chinese steel prices on firm-level productivity using the fixed effects method, like Amiti and Konings (2007) did. My main data source was the database Amadeus. This database provided me an annual European steel manufacturing census of all firms with 20 or more employees for the period 2009 to 2014.
My main results can be summarized as follows. First, the productivity of the average firm increased by 1.84 percent when Chinese steel prices dropped by 10 percent. Secondly, allowing for firm heterogeneity revealed that Chinese steel prices did not affect firms differently depending on their initial productivity. The last result is inconsistent with the current literature, like Konings and Vandenbussche (2008). They showed that firms reacted differently to trade protection depending on their initial productivity. They found that firms with low initial productivity experienced productivity gains during trade protection while firms with high initial productivity faced productivity losses.
The remainder of this paper is organized in seven sections. Section 2 provides a brief review of the most relevant literature about import competition and firm productivity. Section 3 describes the dataset. Section 4 specifies the method used to determine the Total Factor Productivity and presents the results. Section 5 describes the approach used to analyse the effect of Chinese steel prices on firm-level productivity and reports the results. Section 6 will discuss the shortfalls of this study and section 7 concludes.
II. Background literature
This paragraph provides a review of existing literature about import competition and firm level productivity. Based on the papers described, the expected result of my analysis will be discussed.
There have been many studies that try to examine the effects of import competition on productivity. For example Auer et al. (2013), who investigated whether import competition from low wage countries (LWC) have a uniform impact on producer prices of labor intensive products in some European markets. They showed that European producer prices decreased by about 3% when LWC exporters captured 1% of the European market, the results were mainly driven by Chinese exports. After decomposing the mechanisms that underlie the LWC price effect, they showed that import competition had a small effect on the relative wages of production workers, no effect on firm’s margin and a large effect on average firm productivity. Their results are based on 110, 4-digit code (NACE), manufacturing industries in Europe for the period 1995 to 2008.
Bloom et al.(2016) analyzed Europe too. However, they specifically analyzed the effect of Chinese import on innovation during 1996 to 2007. Their results showed that Chinese import competition led to increased technical changes within firms and reallocated employment between firms towards more technological advanced firms. The technical improvements could be an explanation for the increase in firm-level productivity when fiercer import competition takes place. Bloom et al. (2016) as well as Auer et al. (2013) used amounts of import as a measure for import competition. Yet, firm-level data about imports was not available for my sample.
Another measure for increasing import competition is a reduction of import tariffs.
There have been many studies that used this method to analyze the effects on firm-level productivity. Like Pavcnik (2002), she showed productivity gains up to 10 percent higher for import competing industries than gains in the non-traded goods sector due to liberalized trade. Amiti and Konings (2007) confirmed the results of Pavcnik (2002). They showed that a 10 percentage point fall in output tariffs, led to a productivity gain of 2.1 percent. However, once they included input tariffs, their results showed that a 10 percentage point fall in input tariffs resulted in productivity gains of 12 percent for firms that import their inputs, which was at least twice as high as any gains from reducing output tariffs. In addition, Trefler (2004) showed that labor productivity had increased by 14 percent in the industries that experienced the largest tariff cuts in Canada and the United states.
Pierce (2011) investigated it the other way around. He analyzed the firm responses to rising import tariffs. He showed that the way of measuring productivity affected the results. Using units of quantities as productivity measure, his results showed that firm-level productivity decreased when import tariffs rise. This finding complements the results of Pavcnik (2002), Amiti and Konings (2007) and Trefler (2004). On the contrary, using revenue based productivity, Pierce (2011) showed that an increase in import tariffs was associated with productivity gains. Konings and Vandenbussche (2008) confirmed this fact. Their results showed that the productivity of the average firm moderately improved when import tariffs rose. Besides, they showed that the effect of higher import tariffs was not homogeneous among firms. Firms with relatively low initial productivity had productivity gains, while firms with high initial productivity firms experienced productivity losses.
As said before, current literature about trade liberalization do not include any controls for fluctuating product prices. Moreover, they assume that prices were constant during their analyzed period. However, the Chinese steel prices had substantially changed. Consequently, I have used Chinese steel prices as a measure for import competition. Based on the existing papers, I expected to find heterogeneous responses of firms to fluctuating Chinese steel prices. When Chinese steel prices fall, I expected that firms with relatively high initial productivity would experience productivity gains, while low initial productivity firms
would either exit the industry or had productivity losses. Furthermore, I expected that the productivity of the average firm would moderately improve when Chinese steel prices went down.
III. Data
This section describes the data sources used. Additionally, it presents some data descriptive statistics. The firm level balance sheet data came from Amadeus6, which is a database of comparable financial and business information on European largest public and private companies by total assets. In Europe, private companies are required by their regulatory bodies to publicly file financial reports. Amadeus collects this data from regulatory fillings of local governments and transforms it into a standardized format, which makes it possible to do cross-border searching and analysis.
The study of Mion and Zhu (2013) showed that the effect of Chinese imports on firm exit, employment growth and skill upgrading vary across industry levels of technology. They investigated 15.000 manufacturing firms in Belgium for the period 1996-2007. Based on these findings, the expectation was that the effect of Chinese import on productivity also varied across industries. This paper have therefore focused on one 4-digit NACE rev.2 (European Classification of Economic Activities) industry, namely the steel industry with the code, 24107. This code stands for the manufacture of basic iron, steel and ferro-alloysis (Eurostat 2008).
My data source was an annual manufacturing census of all firms with the NACE rev.2 primary code: 2410 and consisting of 20 or more employees between 2009 and 2014. This provided the needed information of 487 firms in total. Chinese steel prices have shown a slightly stable growth for the years 2009 to 2011. After 2011, the prices dropped dramatically until now. Unfortunately, the dataset at firm-level for 2015 was not available yet. So that was why a sample from 2009 up to 2014 was used. The Chinese steel prices came from Bloomberg database. The prices are a year average of all the different geographical submarkets in China in Euros.
To compare financials in real terms, I deflated the profit and loss data by, a 3-digit NACE industry code, domestic producer price index. The 3-digit NACE rev. 2 industry, c241, contains only one subindustry, namely c2410. It could be concluded that the 3 and 4-digit code industries were the same, which made it possible to deflate the 4-digit code revenues and materials costs with a 3-digit code producer price index (Eurostat 2008). The producer price index came from the Intra-European trade data, which is a harmonized and comparable
6
Amadeus is a commercial dataset that has increasingly been used in this field. Examples are Konings & vandenBussuche (2008) and (2005) as well as bloom et al. (2016)
7
statistical database for EU countries. The database is compiled by Eurostat using statistics from the member states.
Table 1 displays the composition of my sample. As shown in table 1a, the amount of observations was not constant throughout the years. Table 1b presents the number of firms that stayed in the panel for a total of 6, 5, (etc.) years. Concluding, I had an unbalanced sample. Pavcnik (2002) showed that the coefficients for the balanced and unbalanced samples did not differ much. Moreover, I did a robustness check and the results of the balanced sample did not change much8. See appendix A for the sources of the data and appendix C for more data descriptive statistics.
Table 1a - Panel information Table 1b - Panel information
Year Number of Firms Years in the panel Number of Firms 2009 390 6 218 2010 394 5 93 2011 410 4 74 2012 363 3 42 2013 378 2 30 2014 350 1 30 Total number of firm-years 2285 Total number of firms 487
Note: Table shows the number of firms in a given year. Note: The right hand side of the table gives the number
of firms that stay in the panel for a total of 6,5, (etc) years.
IV. Total factor productivity
The first part of this section describes the method used to determine the Total Factor Productivity. In the second part the estimations of the total factor productivity are presented.
A. Determining Total factor productivity
To determine the influence of the Chinese steel prices on the productivity of a firm I applied the same method as Amiti and Konings (2007). First, I considered a firm with a Cobb-Douglas function,
(1)
Where the production of a firm i at time t is a function of labor , capital , and
input materials . The interesting part was whether the productivity of firm i was a
8
function of Chinese steel prices, denoted by . In this section, I estimate firm-level Total Factor Productivity (TFP). In the following section, I specify how Chinese steel prices could affect the productivity. To estimate the production function of a firm I took the natural logarithms of equation (1),
(2)
The firm specific error term, , could be divided into two parts (3)
Where the unexpected productivity shock was, this was neither observed by the firm nor by me. And , was the firm-specific productivity level, this part was observed by the firm
but not observed by me. This asymmetric information problem, introduced a simultaneity bias in my estimations. The simultaneity bias arose because a firm’s private knowledge of its productivity affected its choice about hiring labor, purchasing materials and investing into new capital. If more productive firms were more likely to hire extra workers and invest in new capital due to higher current and anticipated future profitability, the OLS estimates of the labor and material input coefficients would be higher than their true values (Pavcnik 2002).
The Ordinary Least Squares (OLS) regression using fixed effects9 would have partially solved the problem of simultaneity bias, while introducing a new problem (Pavcnik 2002). The fixed effects method assumes that firm’s productivity was constant over time, which made this approach un-useful for the purpose of this paper. I was interested in how firm efficiency evolved over time in response to the dynamics of the Chinese steel imports.
Cornwell et al. (1990) and Liu (1993) relaxed the assumption that the firm-level productivity was time-invariant, by introducing a parametric function of time into the production function (2) to replace the coefficient of firm-specific productivity . The
functional form was
(4)
Cornwell et al. (1990) first estimated the production function (2) by fixed effects to obtain the input coefficients. Then, they calculated the residuals by subtracting the predicted from the actual value of output. Thereafter, they regressed for each plant i this residuals measure on a constant, time and time squared. Using these estimated coefficients, they calculated the firm-specific productivity level, equation (4). Finally, they estimated equation (2) again including the calculated firm-specific productivity level. In the presence of simultaneity bias this procedure still use fixed effects estimation in the first step with the assumption of time-invariant productivity level. Although the measure was time varying, it was still likely to be
based on biased coefficients. Moreover, the fitted values for firm-specific productivity level
9
I also did a Hausman test to see whether this test confirms that I need to use the fixed effects method. The Hausman test shows me that I need to use fixed effects method with a p-value of 0.000. See appendix C for the test results
provided an estimate of that would only be consistent as T goes to infinity. Yet, my sample covered only 6 years, therefore my estimations would be inconsistent, making this approach not appropriate for my analysis. Therefore, I estimated equation (2) using the OLS method and accepted the problem of simultaneity.
Now, I will discuss each variable in equation (2) in more detail, starting with production (Y). Ingene (1982) stated that when a single type of output is produced, number of units is an ideal output measure. However, firms in the 4-digit industry did not produce a single product, the qualifications of the steel produced may have varied across firms and years. Moreover, data about quantities were not available for many companies. An alternative approach is to use operating revenue. Yet, revenues are likely to reflect differences in prices. Deflating the firm-level revenue with an industry wide price deflator would only be appropriate if all firms were producing a single and homogeneous product and all faced the same price for their products. As mentioned before, the assumption of homogeneous products was not realistic for the 4-digit steel industry. However, Mairesse and Jaumandrue (2005) showed that using un-deflated value added, or deflated value added by an industry price index or individual firm-price index, had almost no effect on the estimates of the production coefficients. Their results suggested that customary practice of deflating the revenue by the industry output-price index is an acceptable approach when estimating production functions.
Secondly, I used numbers of employees because of the unavailability of information about hours worked. Ingene (1982) discussed some weaknesses of the approach of using numbers of employees 1) fulltime and part-time employees may have been counted equally (2) average hours worked may have varied, partly due to different fulltime to part time ratio’s, across geographic area and (3) owners and partners may have worked harder and longer than regular employees. Despite of the weaknesses, Ingene (1982) mentioned that it is the most accessible and used measure of labor input. Therefore, I used this approach.
Thirdly, like Jefferson et al. (1992), I used the year-end value of net fixed assets to measure the capital intensity of a firm. The costs for production installations or machinery for steel production are high, so firms usually capitalize these costs. The economic tenure of this production equipment is often long and thus will this capital be categorized as tangible fixed assets. Unfortunately, the dispersion between intangible and tangible fixed assets was not available for many firms, clarifying why total fixed assets were used. The shortcoming of using fixed assets is that the value is affected by depreciation methods, which differs between firms. Importantly, I was interested in how firm efficiency evolved over time and not between firms. By making the realistic assumption that the depreciation method of a firm did not change over time, this shortcoming of using fixed assets would not influence my results.
Like Grossman (1985), I assumed that steel is produced with five inputs: Labor (L), capital (K), energy (E), iron ore (I) and scrap (S). The first two were discussed above. Unfortunately, no firm-level information was available about input units nor costs of energy, iron ore, or scrap. This was the reason why I used the total material costs instead of each item separately. The costs were also deflated by a 3-digit industry-level producer price indices, to express the costs in real terms.
To calculate the log of Total Factor Productivity of firm i at time t, I used the OLS
estimates of the production coefficients of equation (2),
(5)
A. Estimations of Total Factor Productivity
The results of the input coefficients from estimating the production function specified in equation 2 are presented in table 2. The OLS10 with fixed effects estimates, in column 2, were included for comparison with the OLS method. To control for the problem of heteroscedasticity and non-normality of the random error term11, I applied the OLS method using robust standard errors.
First, I will analyze the dissimilarities between OLS estimates and fixed effects estimates. Pavcnik (2002) stated that the OLS estimation of a production function may led to overestimated input coefficients due to the simultaneity problem, as mentioned before. As said before, if more productive firms are more likely to hire extra workers and invest in new capital due to higher current and anticipated future profitability, OLS estimates of the labor and material input coefficients would be higher than their true values (Pavcnik 2002). Whereas the bias of the capital coefficient estimate is ambiguous according to the theory. The results in table 2 did not completely confirm the theory. The coefficient of labor (0.139) estimated by OLS was indeed higher than the estimation by the fixed effects method (0.039). Yet, it was the other way around with the coefficient of materials. The fixed effect estimation of the material coefficient (0.728) was higher than the estimation by OLS method (0.715). According to the theory, the bias of capital is ambiguous so I could not conclude anything based on table 2.
10
A multi-collinearity test was performed in Stata: VIF, variance inflation factor. As a rule of thumb, a variable whose VIF values are greater than 10 may merit further investigation. The tests in Stata shows a maximum VIF of 3.30, so I could conclude that there was no multi-collinearity. See appendix B for more details.
11
Based on the Breusch-Pagan test in Stata I could conclude that the variance of the residuals was heteroscedastic. This test is very sensitive for model assumptions, like normality. That’s why I combined Breusch-Pagan test with the Shapiro-Wilk W test. The Shapiro-Wilk W tests showed that the assumption of normal distribution should be rejected. I used the robust option in Stata to control for heteroscedasticity and the non-normal distribution. See appendix B for the detailed results.
Secondly, I compared the OLS estimates to the results of current literature. Amiti and Konings (2007) investigated the iron and steel industry too. They estimated the production function with both the OLS and the Olley-Pakes method. My OLS estimates of the material and capital coefficients are comparable to their results. Their estimate of the capital coefficient was 0.015 and my estimation was 0.010, for material coefficient their result was a coefficient of 0.787 and my prediction was 0.715. The results of the labor coefficients were quite different. My estimation was 0.139 while their estimation was 0.259. Amiti and Konings (2007) investigated firms in the US steel industry, rather than firms in the European steel industry as I did. A possible explanation for the different labor coefficients could be the divergence labor market characteristics. The results of Inklaar and Timmer (2006) confirmed that the US and European markets have different labor markets. They revealed that the US, with a reputation for having one of the least restrictive labor markets, showed a larger effect of output growth on labor input compared to the European Union.
Table 2 - Coefficients of the production function Dependent variable:
Ln(Y)
OLS Fixed effects
Ln K (Capital)it 0.010*** (0.012) 0.039* (0.022)a Ln L (Labor)it 0.139*** (0.010) 0.038 (0.077)a Ln (M) it Materials 0.715*** (0.013) 0.728*** (0.049)a Constant 1.749*** (0.088) 3.084*** (0.530)a
Fixed effects No Yes
R2 0.9642 0.9906
Observations 2285 2285
Firms 487 487
Notes:
1) Robust standard errors in the parentheses
2) ***denotes 1% significance; **denotes 5% significance; and * denotes 10% significance 3) a Robust standard errors corrected for clustering at the firm level in the parentheses
4) Robustness check outliers. I have also estimated the fixed effects including time dummies regression including outliers. The coefficients do not change much and all significant variables remain significant at the same level. See appendix C.
Graph 3 presents the average Total Factor Productivity per year from 2009 to 2012 for the steel industry 12. I used the input coefficients based on the OLS estimation from column 1 in table 2 to construct a measure of firm-level productivity.13 As shown in the graph, the Total Factor Productivity (grey bars) had decreased in the period 2009 to 2011 and had increased from 2011 to 2014. While the Chinese steel prices (black line) had increased during 2009 to 2011 and dropped thereafter. The productivity pattern in graph 3 confirmed the theory of Pavcnik (2002), Trefler (2004) and Nataraj (2011). They all found evidence for productivity improvements that can be attributed to fiercer import competition, measured by decreasing import tariffs.
Graph 3 – Average total factor productivity and Chinese steel prices
Note:
1) Chinese steel prices are deflated by a Chinese producer price index for the manufacturing of metals. 2) The yearly average total factor productivity is calculated as follows:
V. Chinese steel price effects
In this section I first specify how the Total Factor Productivity could be affected by Chinese steel prices. Thereafter I show the estimated effects of Chinese steel prices on firm-level productivity.
A. Determining Chinese steel price effects
In the second stage, using the firm-level measure of TFP from equation 5, I assessed whether Chinese steel prices affected the productivity of firms,
4 + (6)
12
First, the total factor productivity is calculated using the OLS estimates of the production coefficients. Thereafter, the yearly average total factor productivity is calculated as follows:
13
To control for certain types of omitted variables called unobserved firm-level heterogeneity, I estimated equation 6 using OLS with fixed effects, . Treating as a fixed effect means that I allowed the cross-sectional differences in productivity between firms to be freely correlated with all the variables in equation 6.
The Chinese steel prices at time t, were the Chinese hot rolled coiled prices, this is an often used steel price indicator. Every steel product needs to be hot rolled and will be processed further depending on the final product qualifications. To compare the steel prices in real terms I deflated the prices with the Chinese producer price index14. Based on the findings of Pavcnik (2002) and Trefler (2004), I hypothesized that a fall in Chinese steel prices would increase firm-level productivity ( , as the increase in import competition is likely to force firms to search for ways to improve their efficiency.
The level of import competition was probably also affected by the decrease of oil prices. The transportation costs of imports had become lower and made it more appealing to import steel products. Besides the effects of import competition on steel products, oil prices had also influence on the inputs for the production of steel. The lower oil prices made it more appealing for steel firms it selves to import their inputs. Amiti and Konings (2007) showed that the productivity increase was mainly due to an import tariff reduction on inputs (-0.441), rather than a decrease in output tariffs (0.070). To control these two effects of oil prices, I included oil prices. The oil prices were also deflated by a 3-digit industry-level producer price indices to get the prices in real terms. A negative and significant coefficient of oil prices, , would imply that firms became more productive due to higher import competition on steel products or better technology embodied in foreign inputs (Amiti and Konings, 2007).
The age of a firm is a standard control variable in the empirical industrial organization literature (Mion and Zhu, 2013). The structure of the company is dependent on the life cycle stage: starting-up, growing, mature, declining, this will be reflected by the age of the firm. I used age per year and not one observation per company, by using one observation per company and OLS method with fixed effects, multi-collinearity would arise. Then, there would be no age effects which could explain the different productivity levels of the firm throughout the period.
I also included a control variable for firm size. Griliches and Mairesse (1983) investigated five manufacturing industries in France and the US for the period 1973 to 1978. The sample in France contained 185 firms and in the US somewhat more, namely 343 firms. They showed that larger firms had a slight tendency for faster growth of total factor productivity. Based on their findings, I hypothesized that would be slightly positive, indicating that an increase in firm size would led to faster growth of total factor productivity.
14
B. Estimations of Chinese steel price effects
The regression results of equation 6 are presented in table 3. First, I estimated the effect of Chinese steel prices on TFP as a benchmark. This procedure is common in the literature about productivity changes and trade liberalization. Although the effect of Chinese steel prices on productivity was not significant (0.444), the negative relation (-0.050), shown in column 1 in table 3, is consistent with my expectations and the current literature. However, my results showed a much lower effect of Chinese steel prices than the effects of output tariff reduction found by Pavcnik (2002) or by Amiti and Konings (2007). They showed that fall in output tariffs of 10 percentage points increased productivity by respectively, 2.8 and 2.1 percent.
In regression 2, column 2 in table 3, I included another environmental variable, namely oil prices. While including oil prices in the regression, the coefficient on Chinese steel prices had more than tripled. The results showed that a 10 percent fall in Chinese steel prices would increase the firm-level productivity by 1.53 percent. The coefficient on oil prices was higher, indicating that a 10 percent fall in oil prices increased productivity by 1.95 percent. As already mentioned, this result indicate two effects 1) higher productivity due to fiercer import competition on steel products and 2) benefits from importing inputs like higher quality inputs, more varieties of inputs, or learning effects (Amiti and Konings 2007). To be able to assign the effects of oil prices to the different causes, I should have included an import dummy and interacted it with oil prices. The coefficient of the interaction term would then indicate the effect of importing inputs on firm level productivity. The coefficient on oil price would then present the effect of fiercer import competition on productivity. Unfortunately, the firm-level imports were not available. This made it hard, if not impossible, to separate the different effects.
Once I included firm characteristics, in column 3, the coefficient on Chinese steel prices stayed almost the same. In that case it suggested that a 10 percent fall in Chinese steel prices increased productivity by 1.84 percent. In contrast, the coefficient on oil price had slightly decreased. The results showed that the productivity gains from lower Chinese steel prices were lower than those from decreasing oil prices. My results on the coefficient of firm size is inconsistent with the theory of Griliches and Mairesse (1983). My results showed that firms size had no effect on the firm-level productivity.
The results of falling Chinese steel prices and increasing within-firm productivity is consistent with Bloom et al. (2016), they revealed that firms facing higher levels of Chinese import competition create more patents, raised their IT intensity and increased their overall level of productivity. Their coefficient suggested that a 1 percent increase in Chinese import penetration was associated with a 0.257 percent increase in Total Factor Productivity growth.
Yet, they also showed that there is a U-shape response of prices, and markups to competition. Initially, both prices and markups fall. However, as import competition increased further, the scope for quality differentiation rose and so did average quality. The higher quality products would increase markups. The higher quality products, and thus higher prices, could be reflected in the revenue base productivity. Sadly, it was not possible to identify differences in quality separately from differences in measured productivity. Whether any improvements in product quality were reflected in Total Factor Productivity depended on how inputs were priced. If quality of inputs were fully reflected in changes in input prices, then it would not showed up in TFP.
Table 3 – Basic results Dependent variable: Ln( ) (1) (2) (3) (4) Ln (Chinese prices)it -0.050 -0.153* -0.185* -0.184* (0.065) (0.081) (0.108) (0.105) Ln (Oil prices)it -0.195*** (0.069) -0.194*** (0.072) -0.167** (0.072) Ln (Firm size)it 0.008 (0.045) 0.006 (0.046) Firm ageit -0.002 (0.007) -0.001 (0.007)
Exit it=1 if firm exits in t+1 -0.028*
(0.019) Constant 2.062*** (0.408) 2.742*** (0.516) 2.879** (1.210) 3.455*** (1.197)
Firm fixed effects Yes Yes Yes Yes
R2 0.7302 0.7316 0.7317 0.7317
Observations 2285 2285 2285 2285
Firms 487 487 487 487
Notes:
1) Robust standard errors corrected for clustering at the firm level in the parentheses15 2) ***denotes 1% significance; **denotes 5% significance; and * denotes 10% significance
15
To control for potentially auto correlated and potentially heteroskedastic error term, I will use
clustered standard errors. These standard errors allow the regression error to have an arbitrary correlation within a firm, but assume that the regression errors are uncorrelated across firms.
In column 4, I controlled for the exit of firms by including a dummy variable which was equal to 1 if the firm exited the industry in the following period. The existing literature showed that the productivity level of exiting firms was lower than the surviving firms, like Liu (1993). As well as Amiti and Konings (2007), they showed that the productivity level of exiting firms was 4.2 percent lower than surviving firms. My findings confirmed the theory that exiting firms have lower productivity. The results in column 4 showed that the productivity of exiting firms was 2.8 percent lower than surviving firms.
Obviously, my results showed that a fall in Chinese steel prices induced higher firm-level productivity. However, my results did not indicate how firms increased their productivity. Increased import competition, measured with lower Chinese steel prices, may have directed firms to switch their product mix from low- to high- productivity products (Bernard et al., 2010). Pavcnik (2002) investigated the possible option from Bernard et al. (2010) at a broader level, instead of analyzing product switches, she investigated industry switches. She found that firms that switch their industry category were on average 2.9 percent more productive than other firms. Another reason could be productivity gains due to technological improvements induced by higher imports (Bloom et al., 2016). Alternatively, as said before, productivity increases could be induced by benefits from importing inputs like higher quality inputs, more varieties of inputs, or learning effects (Amiti and Konings 2007).
C. Productivity groups and Chinese steel prices
Current literature suggested that the effect of trade liberalization on firm-level productivity is depended on firm characteristics, like Konings and Vandenbussche (2008). They showed that the effect of anti-dumping protection on firm-level productivity was not homogeneous. They revealed that firms with a lower initial firm-level productivity experienced productivity gains during trade protection, while high initial productivity firms experienced productivity losses during trade protection. Moreover, Antoniades (2015) showed that the most productive firms escaped competition by raising quality, prices and markups, while the least productive either exited or responded in the exact opposite manner.
To examine the possibility that the effect of Chinese steel prices was not homogenous among firms, I divided my sample in two parts. Group 1 consisted of firms with low initial productivity level and group 2 consisted of firms with high initial productivity level. I estimated equation 6 for each group separately, table 4 presents the estimated coefficients. As shown, the Chinese steel prices did not have a significant effect on productivity for either of the groups. This could be due to the relatively low number of observations per group. Moreover,
the coefficients did not significantly16 differ from each other, indicating that the effect of Chinese steel prices on productivity was homogeneous among firms with varying initial productivity levels. This result is thus inconsistent with the current literature. Additionally, the coefficients of firm size and firm age did not differ between the two groups too.
Column 1 shows that a 10 percent decrease of oil prices increased the productivity of the low initial productivity group with 3.1 percent. However, oil prices had no effect on the productivity of high initial productivity firms. The productivity levels of firms that exited the industry was also different between the two groups. In the high initial productivity group, firms that exited the industry have a 9 percent lower productivity level than firm that survive. While in the low initial productivity group, the productivity level of firms that exited did not differ from the firms that survive. The productivity in group 1 was already quite low so the firms that exited had already a relatively low productivity level.
16
I used the t-test to compare the coefficients. This is possible because the samples are almost of the
same size.
Table 4 – Total Factor Productivity per group Dependent variable: Ln( ) Low initial productivity High initial Productivity Ln (Chinese prices)it -0.170 (0.163) -0.158 (0.147) Ln (Oil prices)it -0.310* (0.213) -0.059 (0.091) Ln (Firm size)it 0.040 (0.063) -0.051 (0.069) Firm ageit -0.002 (0.009) -0.003 (0.011) Exit it=1 if firm exits in t+1 0.017
(0.055) -0.090* (0.052) Constant 1.923 (1.833) 3.956** (1.594)
Firm fixed effects Yes Yes
R2 0.4935 0.6712
Observations 1142 1143
Firms 252 235
Notes:
1) Robust standard errors corrected for clustering at the firm level in the parentheses17 2) ***denotes 1% significance; **denotes 5% significance; and * denotes 10% significance
.
VI. Discussion
This section discusses the shortfalls of this study. First of all, the OLS estimations of the production function (1) did not control for the problem of simultaneity nor selection bias. As said before, my data set was limited by the reporting standards of material costs. This data limitation gave rise to a potential selection bias if firms that report material costs differed from those that did not. Olley and Pakes (1996) stated that one cannot ignore selection nor simultaneity issues in the estimation of a production function. Their results showed that the failure to control for selection and simultaneity problems results in biased estimates and can be very large. Their estimate of capital had more than doubled, using the Olley and Pakes method instead of OLS. While the coefficient of labor is 30 percent lower, using the Olley and Pakes method. These differences are quite larger, therefore, my results should be
17
To control for potentially auto correlated and potentially heteroskedastic error term, I will use
clustered standard errors. These standard errors allow the regression error to have an arbitrary correlation within a firm, but assume that the regression errors are uncorrelated across firms.
interpreted with care. Subsequent studies should use the Olley-Pakes method to estimate production function (1).
Another concern was that the methodology I used did not control for any dynamics in price mark-ups. The increase in mark-ups would be reflected in higher measured productivity. Pierce (2011) showed indeed that traditional revenue productivity measures are misleading. He concluded that increases in prices and mark-ups artificially inflate the effect of trade liberalization on revenue productivity, while the physical productivity actually falls. This suggested that a rise in Chinese steel prices could increase mark-ups and artificially improve productivity. Deflating revenue by producer price indexes reduces this risk. However, future research can do more by including Herfindahl indices or market shares of firms to control for mark-up effects. The Herfindahl index presents an industry width level of competition while the market shares reflect the market power of a specific firm. A higher Herfindahl index indicates a more concentrated industry, where firms have more market power (Amiti and Konings 2007). Firms with higher market power could set higher prices, the revenue productivity measure would reflect this by increased productivity. However, the productivity did not increase, only the mark-ups did increase, which may have biased my results.
The results of homogenous responses of firms to Chinese steel prices should be verified by new research. The results indicated that firms may respond homogeneous to Chinese steel prices, however the results were not significant. This could be due to too low amount of observations per level of initial productivity. To confirm that firms did react homogeneous to Chinese price changes, further studies should use a different and larger sample covering a longer time period.
VII. Conclusion
This paper studied the effects of Chinese steel prices on the evolution of firm productivity. My motivation was that the rise of China’s steel production which constitutes perhaps the most important exogenous trade shock from low-wage countries to hit the European steel market.
Current literature is focused on the effects of trade liberalization on productivity. These studies did not include product prices, moreover they assumed that prices stayed constant during their investigated time period. This was not a realistic assumption for the steel industry. Prices have changed substantially while tariffs stayed constant. In my analysis I paid attention to the effects of price fluctuations instead of tariff changes.
First, using a Cobb-Douglas production function, I estimated the Total Factor Productivity (TFP). Thereafter, I tested whether Chinese steel prices had influence on the TFP. My results should be interpreted with care, my estimations could be biased due to the failure to control for selection and simultaneity bias. However, my results indicated that the productivity of the average firm increased significantly when Chinese steel prices dropped. As showed, a 10 percent reduction in Chinese steel prices may have improved productivity by 1.84 percent. My finding regarding within-firm productivity improvements is consistent with the literature, like Pavcnik (2002) or Amiti and Konings (2007). Allowing for firm heterogeneity revealed that the effect of the Chinese steel price was probalby homogeneous among firms with various initial productivity. These results are inconsistent with the current literature about trade liberalization and productivity changes, like Konings and Vandenbussche (2008). I had a low number of observations per group. My results should therefore be confirmed on different, larger samples, covering longer periods.
I. Reference list
Amiti, M., & Konings, J., 2007. Trade liberalization, intermediate inputs, and productivity: evidence from Indonesia. The American Economic Review, vol. 97(5), 1611-1638 Antoniades, A., 2015. Heterogeneous Firms, Quality, and Trade. Journal of International
Economics, vol.95(2), 263-273
Auer, R.A., Degen, K., & Fischer, A.M., 2013. Low-wage import competition, inflationary pressure, and industry dynamics in Europe. European Economic Review, vol. 59, 141-166
Bernard, A.B., Redding, S.J., & Schott, P.K., 2010. Multiple-product Firms and Product Switching. American Economic Review, vol.100(1), 70-97
Bloom, N., Draca, M., & Reenen van, J., 2016. Trade Induced Technical Change? The Impact of Chinese Imports on Innovation, IT, and Productivity. Review of Economic Studies, vol.83, 87-117
Cornwell, C., Schmidt. P., & Sickles, P., 1990. Production frontiers with cross-sectional and time-series variation in efficiency levels. Journal of Econometrics, vol.46(1-2), 185 200
Griliches, Z., & Mairesse, J., 1983. Comparing productivity growth: an exploration of French and U.S. industrial and firm data. European Economic Review, vol.21(1), 89-119 Grossman, M., 1986. Imports as a cause of injury: the case of the U.S. steel industry. Journal
of International Economics, vol. 20(3-4), 201-223.
Ingene, C.A., 1982. Labor productivity. Journal of Marketing, vol. 46(4), 75-90
Inklaar, R., & Timmer M.P., 2006. Resurgence of employment growth in the European Union: the role of cycles and labour market reforms. Economics letters, vol.91(1), 61-66 Jefferson, G.H., Rawski, T.G., & Zheng, Y.,1992. Growth, efficiency, and convergence in
China’s State and Collective Industry. Economic Development and Cultural Change, vol.40(2), 239-266
Konings, J., & Vandenbusche, H., 2008. Heterogeneous responses of firms to trade protection. Journal of International Economics, vol.76(2), 371-383
Liu, L., 1993. Entry-exit, learning, and productivity change: Evidence from Chile. Journal of Development Economics, vol.42 217-242
Mairesse, J., & Jaumandreu, J., 2005. Panel-data Estimates of the Production
Function and the Revenue Function: What Difference Does It Make? Scandinavian Journal of Economic, vol.107(4), 651-672
Mion, G., & Zhu, L., 2013. Import competition from and offshoring to China: A curse or blessing for firms?. Journal of International Economic, vol. 89(1), 202-2015. Nataraj, S., 2011. The impact of trade liberalization on productivity: evidence from India’s
formal and informal manufacturing sectors. Journal of International Economics, vol.85(2), 292-301
Olley, S.G., & Pakes, A., 1996. The Dynamics of Productivity in the Telecommunications Equipment Industry. Econometrica, vol. 64(6), 1263-1297
Pakkes, A., 1994. Dynamic Structural Models, Problems and Prospects: Mixed Continuous Discrete Controls and Market Interactions. Advances in Econometrics: Sixth World Congress, vol.(2), 171-259
Pavcnik, N., 2002. Trade liberalization, exit and productivity improvements: evidence from Chilean plants. Review of Economic Studies, 69(1), 245-276
Pierce, J.R., 2011. Plant-level responses to anti-dumping duties: Evidence from U.S. manufacturers. Journal of International Economics, vol. 85(2), 222-233
Trefler, D., 2004. The long and short of the Canada-U.S. free trade agreement. American Economic Review, vol. 94(4), 870-895
Appendix Appendix A Data collection
This appendix presents a roadmap to retrieve the data used in the article. Data from Amadeus
The steps to get a query at Amadeus should be filled in as follows to get the same dataset as used in this paper.
Step 1: Choose your data range fill in: 2009 - 2014 Step 2: Apply your company code BvDEP ID number
Select an option for entering company codessearch the entire database Select a datasetV+L+M+S: plus small firms
Screening variables, countryall countries Conditional Statements:
- Variable: NACE Rev.2, primary code(s) (NACE_PRIM_CODE) - Select: =
- State: 2410 Step 3: Query variables
Select the following variables: - BvDEP ID number
- Company name
- Country
- Year of DATEINC
- NACE Rev.2, primary code(s) - Account date
- Exchange rate from local currency to EUR - Fixed assets
- Total assets
- Number of employees
- Operating revenue (Turnover) - Material costs
Step 4: Select query output Choose STATA file (*.dta) Data from Eurostat
The domestic producer price indexes are retrieved using the following database code: sts_inppd_a
In the option screen select the followings: Time period: 2009 – 2014
Classification of economic activities – NACE Rev.2.: C241 GEO: European union (28 countries)
Seasonal adjustment: Unadjusted data
Business trend indicator: domestic output price index Unit of measure: index, 2010=100
Data from National Bureau of statistics of China
The Chinese producer price indices are on annual basis. In the list of producer price indices for industrial products by sector, I selected the sector: manufacturing of metal products. The given price indices have as base year the preceding year. I rebased the price indices to the year 2009, to make them appropriate for my purpose.
2 2 2 2 2 2 4 4 4 4 4 4 5 5 5 5 5 6 6 6 6 6 6 7 7 7 7 7 7 8 8 8 8 8 8 9 9 9 9 9 9 10 10 10 10 10 10 11 11 11 11 11 11 13 13 13 14 14 14 14 14 14 16 16 16 16 16 17 17 18 1718 1718 1718 20 2023 2023 202123 202123 2021 24 24 24 24 26 26 26 26 27 27 28 28 28 28 28 28 31 32 32 32 32 32 32 33 3336 33 33 33 33 38 38 38 38 38 39 39 39 39 39 39 40 40 40 40 40 40 41 41 41 41 48 48 48 48 48 48 50 50 50 50 50 50 53 53 53 53 5355 5355 56 56 56 56 56 56 57 5761 5761 5761 5761 5761 63 63 63 63 63 63 64 64 64 64 65 65 65 65 65 65 66 66 66 66 66 66 67 67 67 67 67 67 68 68 68 68 68 68 69 69 69 69 69 69 70 70 70 70 70 70 71 71 71 71 71 72 72 72 72 72 72 73 73 73 73 73 73 74 74 74 74 74 75 75 75 75 75 76 76 76 76 76 76 77 77 77 77 77 77 78 78 78 78 78 78 79 79 79 79 80 80 80 80 80 81 81 81 81 81 81 82 82 82 82 82 84 84 84 84 84 84 85 85 85 85 86 86 86 86 86 86 87 87 87 87 87 87 88 88 88 88 89 89 89 89 89 89 90 90 90 93 93 93 93 93 94 94 94 94 94 94 95 95 95 95 95 95 100 100 100 100 100 100 101 101 101 101 101 101 102 102 102 102 105 105 105 105 105 106 106 106 106 106 106 107 107 107 107 107 107 108 108 108 108 108 110 110 110 110 111 111 111 111 111 111 113 113 113 113 113 114 114 114 114 114 115 115 115 116 116 116 116 116 117 117 117 117 118 118 118 118 118 118 121 121 121 121 121122 121122 126 126 126 126 126 126 127 127 127 127 127 127 128 128 128 128 128 129 129 129 129 136 136 136 136 136 136 138 138 138 138 138 139 139 139 139 139 139 141 141 141 141 141 141 144 144 144 144 144 144 145 145 145 145 145 149 149 149 149 149 149 150 150 150 150 153 153 153164 153160164 164 160164 166 166 166 166 167 167 167168 168 168 173 173 173 173 173 173 181 181 181 181 181 181 182 182 182 182 211 211 211 211 211 213 214 214 214 214 214 215 215 215 217 221 221 221 221 221 221 222 222 222 223 223 223 223 223 223 226 226 226 226 226 226 228 228 228 228 228 228 230 230 230 230 230 230 232 232 232 232 232 245 245 245 245 246 246 246 247 247 247 247 247 251 251 251 251 251 254 256 256 256 256 256 256 260 260 260 260 260 260 261 261 261 261 261 262 262 262 262 262 262 266 266 266 266 266 266 267 267 267 267 267 267 272 272 272 272 276 276 276 276 276 277 277 277 277 277 279 279 279 279 279 279 280 280 280 280 280 281 281 281 281 281 282 282 282 282 282 283 283 283 283 283 284 284 284 285 285 285 285 285 285 286 286 286 286 286 287 287 287 287 287 288 288 288 288 288 288 289 289 289 289 292 292 292 292 292 292 293 293 293 293 293 293 294 294 294 294 294 294 295 295 295 295 295 295 298 298 298 298 301 301 301 301 301 302 302 302 302 302 302 305 305 305 305 305 306 306 306 306 306 307 307 307308 308 307308 307308 310 310 310 310 310 310 311 311 311 311 311 311 313 313 317 317 317321 317 325 325 325 325 325 325 326 326 326 326 326 326 330 332 333 333 333 333 333 334 334 334337 334337 334337 334337 340 341 341 341 341 341 341 342 342 342 342 342 342 346 346 346 346 346 347 347 347 347 348 348 348 348 348 348 349 349 349 349 349 358 358 358 358 358 358 360 360 360 360 360 360 362 362 362 362 362 362 363 363 363 363 363 363 364 364 364 364 364 364 365 365 365 366 366 366 366 366 366 370 370 370 370 370 370 374 374 374 374 376 376 376 376 376 376 377 378 378 378 378 378 379 379381 379381 381 381 385 385 385 385 385 385 388 388 388 388 388 388 391 391 391 391 391 391 393 393 393 393 393 393 396 396 396 396 396 396 397 397 397 397 397 397 401 401 401 401 401 401 402 402 402 402 402 408 409 409 409 409 409 409 411 411 411 411 411 411 416 416 416 416 416 416 417 417 417 417 417 417 418 418420 418420 418420 418420 418420 421 421 421 423 423 423 423 423 423 424 424 424 424 424 425 425 425 425 425 427 427 427 427429 430 430 430 430 430 430 431 431 432 433 433 434 434438 434438 434 434 434 443 443 444 444 444 444 444 444 445 445 445 445450 445 445 472 472 472 472 472 472 474 474 474 474 474 474 475 475 475 475 475 475 478 479 479 484 484 484 484 484 484 485 485 485 485 485 485 490 490 490 490 490 490 492 493 493494 493494 493 493494 493494 496 496 496497 496 496 496 500 500501 500501 501 501 501 503 503 504 504 504 504 504 505 505 506 506 506 506 506 506 510 510 510 510 510 510 518 518 521 521 521 521 521 521 522 522 522 522 522 522 523 523 523 523 523 523 527 527 527 527 527 527 528 528 528 528 528 528 529 529 529 529 529 530 530 530 530 530 530 531 531 531534 531534 531534 531534 535 535 535 535 536 539 539 539 539 539 539 541 541 541 541 541550 541550 551 551 551 551 551 551 552 552 552554 552554 552554 552554 555 555 555 555 555 556 556 556 556 556 556 557 557 557 557 557 557 558 558 558 558 558 558 559 559 559 559 559 559 560 560 560 563 563 563 564 564566 564566569 564566 564566 564566 570 570 570 570 570 570 571 571 572 572 572 572 572 573 573578 573578 573 573 573578 579 579 581 581 581 581 581582 581582 588 588 588 588 588 588 589 589 589 592 592 592 592 592 592 595 595 595 598 598 598 598 598 598 599 604 604 610 610 610 610 610 610 612 612 612614 612614 612614 621 621 621 621 621 621 630 630 630 630 630 630 644 644 644 644 644 644 670 670 670 670 670 670 675 675 675 675 675 677 677 677 677 677 677 678 678 678 678 678 678 679 681 681 681 681 681 681 683 683 683686 683686 686 688 688689 688689 688689 688689 688 690 690 690 690 690 690 694 694 694 694 694 695 695 704 704 704 704 704 706 706 706 706 706 706 707 707 707 709 707709 707 710 710 710 710716 710716 710716 719 719 719 719 720 725 725730 725730 725730 730 730 734 734 734 734 734 734 737 737 737 737 737 739 739 739 739 739 739 740 740 740 740 741 741 741 741 741 741 742 742 742 742 742 742 743 743 743 743 743 743 746 747 747 747748 747748 747748 747748749 751 751 751 751 752 752 752 752 752 752 754 754 754 754 754 754 755 755 755 755 755 755 758 758 758 758 759 759 759 759 759 760 760 760 760 760 760 761 761 763 763 763 763 763 763 764 766 768 766768 766 766767768 766767768 769 769 769 769 769 769 770 770 770 770 770 770 771 771 771 772 772 772 772 772 772 773 773 773 773 773 773 774 774 774 774 774 775 775 775 775 775 775 777 777 777 777778779 778779 777778779 780 780 780 780 780 780 781 781 783 783 783 786 786 786 786 786 788 788 788 788 788 788 790 790 790 790 790 790 792 792 792 792 792 794 794 794795 794795 794795 794795 796 796 796 796 796 796 799 799 799801 799801 799801 799801 802 802 802 802 802 806 806 806 806 812 812 812 812 812 813 813814 813814 813814 813814 813814 816 816 816 816 816 816 819 820 820 820 820 820 820 822 822 822 822 822 822 825 825 825 825 826 826 826 826 828 828 828 828 828 828 834 834 834 834 834 834 836 836 836 836 836 836 838 838 838 838 838 843 843 843 844 844 844 844845 844845 844845 847 847 847 847 847 847 850 850 850 850 850 850 851 851 851 851 854 854 854 854 854 854 856 856 856 856 856 856 858 858 858 858 858 858 860 860 860 860 860863 860861863 864 864865 864865866 864865866 869 870 869870 869870 869870 869 871 871 871 871 871 871 872 872 872 872 872 872 873 873 873 873 873 874 874 874 874 874 875 875 875 877 877 877 877 877 877 880 880 880 880 880 880 881 881 881 881 881 881 882 882 882 882 886 886 886 886 886 887 887 887 895 895 895 895 900 900 900 900 904 904 904 907 907 907 907 907 907 908 908 908 908 908 908 909 909 909 909 909 909 910 911 911915 911915 911915 911915 911915 918 918 918 919 919 919 920 920 920 920 920 920 921 921 921 921 921 921 922 922 922 922 923 923 923 923924 924 924 925 925 925 925 928 928 929 929 929 930 930 930 930 931 931 931 931 931 931 933 933 933 933 933 935 935 935 935 937 937 937 937 938 938939 938939 939 938939 940 940 940 940 940 942 942 942943 942943 943 943 945 945 945 945 945 946 946 946 946 947 947 947 947 950 950 950 950 950 950 951 951 951 953 953 953 953 953 953 955 955 955 955 973 973 973 973 973 973 995 995 995 1004 1004 1004 1004 1004 10041019 1020 1020 1020 1020 1021 1021 1021 1021 10211026 10211026 1027 1027 1027 1027 1027 1029 1029 1029 1029 1044 1044 1044 1044 1044 1044 1045 1045 1045 1045 1045 1045 1046 10461047 10461047 10461047 10461047 1049 1049 1049 1049 1049 1051 1051 1051 1051 1051 1051 1052 1052 1052 1052 1052 1054 1054 1054 1054 1054 1054 1055 1055 1057 10571059 10571059 10571059 1057 10571059 1064 1064 1064 1064 1064 1064 1067 1067 1067 1067 1067 1067 1070 1070 1071 1071 1071 1071 1071 1074 1074 1074 1074 1074 1074 1075 1075 1075 1075 1075 1075 1076 1076 1076 1076 1076 1084 10841090 10841090 1084 1084 10841090 1091 1091 1091 1091 1091 1094 1094 1094 1094 1094 1094 1095 1095 1095 1095 1095 1095 1097 1097 1097 1100 10971100 10971100 1101 1101 1101 1101 1101 1102 1102 1102 1102 1102 1106 1106 1106 1106 1106 1107 1107 1107 1107 1108 1108 1108 1108 1108 1109 1109 1109 1109 1109 1109 1110 1110 1110 1110 1110 1111 1111 1111 1111 1111 1112 1113 1113 1113 1113 1114 1114 1114 1114 1115 1115 1115 1115 1115 1115 1117 11171120 11171120 11171120 1117 1120 1121 1121 1121 1121 1121 1121 1122 1122 1123 1123 1123 11231124 11231124 1123 1125 1125 1125 1125 1125 1127 1127 1127 1127 1127 1128 1128 1128 1128 1128 1131 1131 1131 1132 1132 1132 1132 1132 1133 1133 1133 1133 1133 1133 1134 1134 1134 0 2 .0 e + 0 7 4 .0 e + 0 7 6 .0 e + 0 7 O p e ra ti n g re ve n u e 2009 2010 2011 2012 2013 2014 Year
Appendix B checking the data
The collected data from Amadeus doesn’t provides me with a complete set of firms of which all the relevant data is available. That’s why I checked the data first.
The following steps were done:
The balance sheet data is transformed from local currency to Euros using the exchange rate from Amadeus.
Observations of firms with less than 20 employees are removed.
Observations of firms with fixed assets lower than €100.000 are removed. Observations of firms with material costs lower than €900 are removed Observations of firms with operating revenue lower than €900 are removed
These boundaries are based on my experience at Tata Steel. Tata Steel has a small
subsidiary in Belgium. It is a really small plant with only 24 employees but they still produce a lot of steel. I assume that a plant with lower than 20 employees aren’t able to produce a significant amount of steel.
Some firms have two observations per year, mostly due to the fact of consolidation.
If possible, these double observations were separated based on their consolidation codes. I removed the unconsolidated observations. Expected productivity increase is often a reason for mergers and acquisitions, that’s why I will use the consolidated version of the firm balance-sheet data.
Besides setting the boundaries for firm data, I used also scatter plots to identify outlying firms. The numbers stands for the firm ID.
Graph 1- Operating revenue and Year
Based on the graph we find 1 big outlier, firm 479 in 2011, and some firms which have stable high operating revenue, firms 14, 64, 63 & 1110. I decided to remove only firm 479. I have two observations for this firm, namely 2009 and 2011. The operating revenue increased in this period from €36 million to €9600 million, this doesn’t seem reliable.
