Journal number 3 ∘ George Berulava ∘
SERVICES SECTOR REFORMATION AND ENHANCEMENT OF EXPORT ACTIVITY OF MANUFACTURING FIRMS: EVIDENCE FROM TRANSITION ECONOMIES[1]

Summary.

The objective of current paper is to explore the link between reforms in services sector and export performance of manufacturing firms in transition economies. The results of the study provide a new understanding of the consequences of trade liberalization in services sector. In particular, positive impact of services sector efficiency on export performance of manufacturers is revealed. Along with services impact, we find that firm specific characteristics such as introduction of new products, investments in research and development, employment of advanced technologies, and employee skills are key drivers of export performance in manufacturing sector in transition economies. Firm’s size and foreign investments do matter as well. The results of this study provide information for policymakers and stakeholders that will facilitate elaboration of policy interventions aimed at improvement of export performance of manufacturers in transition economies.

JEL Classification: F10, F14, C33

Keywords:services inputs,transition economies, export performance, manufacturing industry, panel data analysis.

1. Introduction.

Exporting is an important type of economic activity that many consider crucial to the growth of productivity, and living standards. The experience of the East Asian tigers provides   evidence that exporting is an important component of the growth strategy in emerging markets [The World Bank, 1996]. Ensuring a favorable environment for exporting thus represents one of the key challenges for transition economies on their path to economic development.

Discussions of factors that determine success of export performance have been ongoing for many years. Both the factors that are under the firm control and external factors have been studied extensively in the academic literature. However, the role of services sector as one of the external factors in promoting export performance of downstream sectors remained relatively unstudied. The existing research of the consequences of services sector liberalization is limited mainly to the analysis of the impact of services sectors on the on the productivity in downstream industries [Arnold, Javorcik and Mattoo, 2011; Arnold, Mattoo and Narciso, 2006; Esschenbach and Hoekman, 2006; Fernandes and Paunov, 2012]. In this paper[2] we extend the existing research by emphasizing the relationship between the services sectors and the export performance of downstream industries. In particular, the objective of the study is to explore the impact of services inputs on export performance of manufacturing firms in transition economies. The results of the study are intended to improve our understanding of the consequences of services sector policy, and thus they extend the existing theoretical framework. However, the findings of the current research are important not only for theoretical but also for practical considerations. They provide grounds for the recognizing key determinants of manufacturers’ export performance in transition economies. In that way, the research contributes to the ongoing political debate on economic development issues and provides insights for targeting of public policies.

The rest of the paper is organized as follows. Section 2 examines the relevant literature on the link between liberalization in services sector and manufacturers’ export performance. Further, in this section, based on the literature review, the research hypothesis of the study is formulated. Section 3 presents a discussion of the research methodology, including description of econometric model, variables and estimation techniques. In the fourth section the data description is provided. The fifth section reviews the study results on the impact of services inputs onboth the decision of manufacturers to participate at export markets and their export intensity. The final remarks are discussed in section 6.

 2. Literature Review.

This paper focuses on the role of the services sector in influencing export performance of manufactures in transition economies. The literature indicates that countries in transition can benefit from increased exports. An increase in exports might boost productivity through “learning by exporting” of individual companies; or it may allow additional imports of high tech products. Either avenue would stimulate economic growth.

The relationship between exporting and firm’s productivity has been attracting attention of academicians for a long period of time. The literature emphasizes two alternative but not mutually-exclusive views on the relationship between firm’s export activity and productivity [Wagner, 2007; Bernard and Jensen, 1999; Bernard and Wagner, 1997]. According to the first approach the more productive firms self-select themselves to export markets. The other view holds that participation at export markets makes firms more productive through learning effect. Recent period a fast growing number of studies have found evidence consistent with both alternative hypotheses on exporting-productivity relationship. Bernard and Jensen [Bernard and Jensen, 1995] in longitudinal study of productivity differences between US exporter and non-exporter firms find that exporters as compared to non-exporters: were substantially larger; had higher levels of capital intensity and investment per employee; were characterized by higher wage payments and showed higher labor productivity. Bernard and Wagner [Bernard and Wagner, 2001] on the base of study of German firms indicate that future entry at export markets is significantly and positively associated with higher productivity. Similarly, Alvarez and Lopez [Alvarez and Lopez, 2005] using dataset of firms in Chile show that export participants have higher both labor productivity and total factor productivity than non-exporters. Moreover, the authors conclude that firms make conscious efforts to enhance productivity before entering export markets. In support of self-selection hypothesis, the positive and significant impact of labor productivity on exporting was discovered also in number of other studies [Greenaway and Kneller, 2004; Greenaway and Yu, 2004; Bernard and Jensen, 2004].

The strong support of learning-by-exporting hypothesis was found in Baldwin and Gu [Baldwin and Gu, 2004] study of Canadian manufacturing plants. The authors suggest that exporting had a positive influence on plant’s productivity growth through the following mechanisms: learning by exporting; exposure to international competition; and increases in product specialization that allowed for exploitation of scale economies. The study shows that trade liberalization had a significant impact on the strong export growth and that exporting is associated with increased investments in R&D and training, and innovations. The fact that labor productivity improves after firms enter export markets was proved also in other studies [Clerides et al., 1998; Bernard and Wagner, 1997; Blalock and Gertler, 2004].

Though the productivity-export link has been studied very extensively in recent years, some aspects of this relationship remain relatively unexplored.  For instance, the now large heterogeneous firms’ literature initiated by Melitz [Melitz, 2003] suggests that the more productive firms are the ones that export.  Melitz assumes that there is a fixed cost in selling in export markets and only the more productive firms will choose to export, while less productive firms will decide to serve the domestic market.  In this stream of research, high-productivity of firms that self-select into export markets is considered as an outcome of firm’s deliberate strategy. However, the productivity of firms can be caused by factors external to the firm and which are not under its control. The recent empirical research of the relationship between export activity and external factors influencing productivity is focused mainly on the study of the effects of business climate variables. For instance, Clarke [Clarke, 2005] in a study of African exporters finds that in addition to enterprise characteristics, policy-related variables also affect export performance. In particular, the author suggests that restrictive trade and customs regulations as well as poor customs administration can discourage manufacturing enterprises from exporting. Balchin and Edwards [Balchin and Edwards, 2008] find that the business climate is closely associated with firm-level manufacturing export performance in Africa. The empirical evidence on the effects of business climate and infrastructure on manufacturers’ export supply capacity is also documented in [Dollar et al., 2006; Iwanow and Kirkpatrick, 2008]. Similarly, liberalization of the services sectors can be considered as one of the external factors that positively influences costs and productivity of downstream firms’ and thus promotes their export activities. Services can be viewed as a factor of production along with labor, capital and other inputs. The enhancement of services inputs can reduce production costs, increase the marginal productivity of other inputs and raise output. The impact of services sector liberalization on the productivity in downstream sector is well documented in academic literature [Arnold, Javorcik and Mattoo, 2011; Arnold, Mattoo and Narciso, 2006; Esschenbach and Hoekman, 2006; Fernandes and Paunov, 2012]. These studies indicate that the availability of high-quality and low cost services contributes to the reduction of costs and increase of productivity of downstream manufacturing firms. Taking into account the fact that services sector efficiency is an important determinant of manufacturing firm productivity and productivity is a crucial factor of exporting, one may hypothesize that services sector liberalization through the improvement of productivity of the firms in downstream industries can increase their exports. Again relying on Melitz [Melitz, 2003], services sector liberalization can positively influence not only the export intensity of manufacturers but also their decisions to participate in export markets and the number of export markets that they serve. To say distinctly, theory suggests that a more efficient services sector through increasing the productivity of firms and reducing the fixed costs of exporting can boost the number of firms in downstream industries that “self-select” into export markets. Thus, services sector liberalization, by increasing the efficiency, variety and quality of services markets, can then increase exports.

Though theory indicates that better services should increase exports (both intensively and extensively), the empirical links are not well studied. Further, those studies that do exist are based on African or Latin American data, so there is a lack of literature based on transition country data.  In this research we try to fill this gap by examining the relationship between performance of services sector and export performance of manufacturing firms in transition economies.  Based on the literature review, the main research hypothesis of the study can be formulated as follows: the enhancement of services sector positively and significantly influences both the decision of manufacturers to participate in  export markets (“extensive margin”) and their export intensity in any market (“intensive margin”).

3. Research Methodology.

This section discusses the following issues: formulation of econometric model, description of measures and estimation issues.

Econometric model and variables. In order to estimate the impact of services inputs on export performance of manufacturers we use the following panel data regression model for firm i at time t:

EI_it = Υ^'SI_it + Ψ^'C_it + ε_it                   (1)

with

ε_it = a_i + v_it

where, Υ^' and Ψ^' are vectors of parameters to be estimated

EI - export intensity, represented by the share of exports in total sales (exports/total sales).

SI- vector of services input variables. In this study we use two groups of services input variables. The first group reflects performance of three services sectors – telecommunications, electricity, and finance. These variables are represented by subjective measures from the BEEPS dataset that are based on firm’s valuation on a scale from 1 to 5 as to how much of a constraint they consider telecommunications, electricity and finance for their business. The second group of variables are BRD[3] indices of policy reforms [EBRD, 2009], which reflect the overall liberalization of services sector. In particular, in this study we employ: EBRD overall index of infrastructure reform, which reflects reforms in telecommunications, electric power, railway transport, road transport, and water distribution sectors; and EBRD index of banking sector reform.

C - set of control variables:

S - firms size (small-5-19 employees; medium-20-99 employees; large-more than 100 employees).

Tech- employment of advanced technologies. This is a dummy variable, which shows whether the firm has a high-speed, broadband internet connection on its premises.

RD – dummy variable, which reflects whether the firm in the last three years invested in research and development.

Innov – dummy variable, which reflects whether the firm in the last three years introduced new products or services.

FO– dummy variable for foreign ownership.

Ind – industry type (food, textiles, garments, chemicals & rubber, non-metallic mineral products, basic metals, fabricated metal products, machinery and equipment, electronics, other manufacturing);

ES - employee skills, measured by percent of employee with tertiary education.

Comp – degree of competition, measured on the basis of firm’s subjective responses on a scale from 1 to 5 as to how much pressure from domestic and foreign competitors, effects firm’s decision to develop a new product and reduce costs. This variable is constructed using principal component factor analysis.

RQ – regulatory quality. This is one of the six dimensions of the Worldwide Governance Indicators[4] which reflects the ability of the government to stimulate free trade and promote private sector development.

CapLoc – dummy variable which reflects whether the firm is located in the capital;

EU - European Union membership.

ε_it  - is an error term, which consists from two error components: a_i  - the unobservable individual (time-invariant) effect which may be correlated with the observed variables SI_it and C_it; and v_it- the remainder disturbance, which varies with individuals and time and can be thought of as the usual disturbance in the regression. a_i and v_it are assumed to be i.i.d.  (0,δ_a) and (0,δ_v)  i.i.d. correspondingly.

In this model productivity doesn’t enter in the equation directly. We proxy the productivity by firm-specific characteristics, like size, foreign ownership, employee skills and services inputs variables.

Estimation. The econometric model discussed above is intended to estimate the respective impacts of services input (telecommunications, finance, and electricity) and specific firm characteristics on export intensity of manufacturers. However, some issues can arise while estimating the model. First, since the export intensity is truncated variable the sample selection bias issue can arise while estimating the model. The second problem is related to potential endogeneity of service input variables as well as some other independent variables.

To deal with selection bias problems the two-stage estimation process is employed in this study [Heckman, 1979; Helpman et al., 2008; Shepotylo, 2009].  First, we formulate a model for the probability of exporting. The specification for this relationship is of the following form:

 D^•_it = V^'_itη - e_{it;      Dit = 1[D^•it ≥ 0]         (2)}

where, i and t denote firm and time, respectively. The variable EI_it in equation (1) is observable only if dummy variable D_it=1. The vector of independent variables V_it, in probit equation, is a superset of the vector of independent variables in equation (1). Along with services input and control variables it includes a selection variable.

Based on results of previous studies, at this stage we employ the following selection variable: EF – export facilitation index constructed using principal component factor analysis from the Doing Business database[5]. The index consists of the following elements: number of all documents required to export goods; time necessary to comply with all procedures required to export goods; cost associated with all the procedures required to exporting.

Like in panel regression equation (1), the error term e_it comprises two elements: k_i - unobservable and time invariant individual effect, which may be correlated with V_it, and ν_it - unobserved disturbance. The error term is normally distributed with zero mean and variance equal to (δ²_k + δ²v). Following Shepotylo [Shepotylo, 2009] we assume that  (δ²_k + δ²v) = 1 and estimate the equation using standard probit regression:

Pr(D^•_it = 1|V'_it) =Φ(V'_itη)            (3)

 

Based on these estimations we calculate the inverse Mills ratio  ''λ_it'' [Heckman, 1979]:

λ_it = Ø(V'_itÝ)/Φ(V'_itÝ)     (4)

where, Ø and Φ are the density and distribution function for a standard normal variable correspondingly; Ý-vector of estimated parameters;V^'_itÝ - prediction of the probit equation.

At the second stage, we correct for self-selection by incorporating a transformation of the predicted individual probabilities or the inverse Mills ratio  as an additional explanatory variable to regression equation (1). In particular, the export intensity equation (1) may be specified as follows:

 EI_it = α'SI_it + β'C_it + ξ_tλ_it + ε^•_it                 (5)

where, ε^•_it- is error term corrected after including inverse Mills ratio in the equation.

To address the problem of endogeneity of the services input variables, which are very likely to be correlated with individual specific effect (a_i), the export intensity equation is estimated by applying Hausman-Taylor IV estimation procedure [Hausman and Taylor, 1981]. Formally, the Hausman-Taylor model can be represented in its most general form as follows:

 y_it = X'_1itβ₁ +X'_2itβ₂ + W'_1itα₁+ W'_2itα₂ + u_i + ω_{it         (6)}

where,

X'_1it–are time varying exogenous variables;

X'_2it– are time varying endogenous variables;

W'_it – are time invariant exogenous variables;

W'_2it – are time invariant endogenous variables;

U_i– is unobserved individual specific effect;

ω_it- the remainder error term.

In our study, services inputs variables (SI) and innovation (Innov) are time varying endogenous variables (x_2it); industry type (Ind) and membership in European Union (EU) aretime invariant exogenous variables (w_1it); and all other variables are time varying exogenous variables (x_1it).

This estimation procedure uses only the information within the model and employs both the between and within variation of the strictly exogenous variables as instruments. More specifically, the deviation from mean    serves as an instrument for time varying endogenous variables x_2it; and the individual means of the strictly exogenous regressors are used as instruments for the time invariant endogenous regressors w₂ that are correlated with the individual effects [Hausman and Taylor, 1981]. The advantage of Hausman-Taylor estimation approach is that it addresses the limitations that would arise with alternative estimation models such as fixed and random effects estimators. The major drawback of the fixed effects estimator is that the coefficients of time-invariant explanatory variables are not identified. Thus in the current research it is not suited to estimate the effects of time constant variables, such as industry, membership in EU. These variables while not central to the main research hypothesis of the study are nevertheless of some interest. Random-effects models on the other hand assume that all independent variables are uncorrelated with the individual specific error , which, if not correct, leads to inconsistent estimates. However, the main limitation of Hausman-Taylor estimation method is that it doesn’t solve the simultaneity problem, which can arise from the correlation of independent variables with the remainder disturbance term.  To control for the robustness of the Hausman-Taylor estimation results with regard to simultaneity and sample selection bias issues we employ several robustness checks in this study.

4. Data Description.

The main source of the data for the research is the micro-level unbalanced panel data from the Enterprise Surveys database (Business Environment and Enterprise Performance Survey (BEEPS) Panel)[6]. The surveys were conducted by the European Bank for Reconstruction and Development (EBRD) and the World Bank Group (the World Bank) in 2002, 2005, 2007, and 2008/09 for firms in 29 countries in the European and Central Asian region. In all countries where a reliable sample frame was available (except Albania), the sample was selected using stratified random sampling. Three levels of stratification were used in all countries: industry, establishment size and region. The more detailed description of the sampling methodology can be found in the Sampling Manual.[7]

The panel provides totally 29,386 observations. Since the objective of our study is the export performance of manufacturing firms, we limit the sample only to manufacturing sector. This gives us the final sample size of 11,293 observations at the firm level, which corresponds to 10,263 firms. On average there are 1.1 years of data per firm available.

Table 1 shows descriptive statistics of the variables used in the study. According to the data from this table, out of 11,293 observations 26.9 percent belong to exporting firms. This corresponds to 2,760 firms in the sample that participate at export markets. On average the share of export in total sales is 11.83 percent. Telecommunications sector represents the lowest, while finance sector creates the biggest obstacles for businesses. In the sample, according to EBRD assessments, more progress was achieved in banking sector reform as compared to infrastructure reform.

 Table 1. Descriptive statistics

Fifty-two percent of firms introduced a new product or services and thirty-three percent of firms invested in research and development during last three years. More than sixty percent of companies have high-speed broadband internet connection and only 11.3 percent of firms are owned by foreigners. On average 21.7 percent of firm’s employees in the sample have university degree. Less than 10 percent of companies are located in the capital. The sample includes firms of all sizes. Generally, firms are equally distributed by size with little advantage of medium and small firms.

The comparison of exporting and non-exporting firms is presented in Table 2. In conformity with international experience, on average the share of firms that innovate and use advanced technologies are substantially higher among exporters as compared to non-exporters. Similarly, the share of firms that invest in R&D is almost twice higher among exporting firms in our sample. Exporting firms are larger and have higher shares of foreign ownership. However, contrary to international evidence non-exporting firms have better skilled personnel.  Location in capital doesn’t play important role in differentiation between exporting and non-exporting firms.

Table 2. Descriptive characteristics of exporters and non-exporters

Variables	Export status
	Exporter		Non-Exporter
	Percent	Number of observations	Percent	Number of observations
Innovation during last 3 years	61.3	3,031	49.4	7,113
R&D during last 3 years	50	2,554	25.5	5,004
Foreign ownership	20.5	2,978	7.9	8,126
High-speed internet connection	81.4	1,542	54.2	3,817
Percent of employees with university degree	19.6	2,964	22.6	6,930
Location in capital	8.6	3,037	9.4	8,247
Size (Large firm)	51	3,037	22.8	8,247
Size (Medium firm)	34	3,037	35.8	8,247
Size (Small firm)	15	3,037	41.4	8,247

5. Study Results.

Selection equation. Five different specifications of the model (eq.1-4) are estimated. Each of five services input measures is entered into the model one by one. Table 3 reports the parameter estimates and goodness-of-fit indicators for probit regressions. All five models reflect a good fit with the data (Wald Chi-squares are significant at p< 0.01 or at p < 0.05 levels; Likelihood ratio tests of rho = 0 are significant at one percent level). As it was expected, the selection variable – export facilitation (which reflects costs and time necessary to export goods) - has significant and negative impact on probability of exporting. This result provides support to the stylized fact that restrictive trade regulation discourages manufacturers from exporting.

Surprisingly, only two from five services input variables have a significant impact on dependent variable (at 5 percent level). Marginal improvement in the performance of telecommunications sector and marginal increases in EBRD index of infrastructure reforms raise probability of positive exports correspondingly by 4.1 and 21.3 percent. Thus the hypothesis that efficient services inputs encourage manufacturers to participate at export markets is supported only partially. The results of the probit regression suggest that membership in European Union and regulatory quality (with the exception of the fourth specification) don’t have any significant effect on export participation. 

Table 3. Probit selection equation

Variables	Coefficients
	I	II	III	IV	V
Dependent Variable: Export Choice
Electricity as an obstacle	-.0526  (.0433)	-	-	-	-
Telecommunications as an obstacle	-	-.1562*** (.0595)	-	-	-
Finance as an obstacle	-	-	-.0334  (.0374)	-	-
EBRD index of infrastructure reform	-	-	-	.8206*** (.2113)	-
EBRD index of banking sector reform	-	-	-	-	.1152   (.1655)
Innovation during last 3 years	.2258**  (.094)	.2525**  (.1013)	.2049**  (.0934)	.2258**  (.0983)	.2031** (.0953)
R&D during last 3 years	.4641***    (.1247)	.5065***   (.1377)	.4382***   (.1230)	.4895***   (.1312)	.4589***   (.1287)
Technological level of company (High-speed internet connection)	.6634***   (.1457)	.6742*** (.1562)	.6638*** (.1480)	.7965*** (.1617)	.6687***   (.1489)
Employee skills	.0063***   (.0022)	.0064***   (.0023)	.006***   (.0022)	.0053***   (.0022)	.0066***   (.0022)
Foreign ownership	.8782***    (.193)	.8933***    (.2078)	.8796***    (.1966)	.9249***    (.2042)	.8814***    (.1983)
Size (Small firm)	-1.589***    (.2946)	-1.653***    (.3276)	-1.570***    (.2970)	-1.842***    (.3315)	-1.685***    (.3166)
Size (Medium firm)	-.5849***    (.1453)	-.6074***    (.1570)	-.5694***    (.1450)	-.7106***    (.1631)	-.6362***    (.1554)
Location in Capital	-1.246**  (.5281)	-1.174**  (.5439)	-1.313**  (.5297)	-1.354**  (.562)	-1.172**  (.5507)
European Union country	.0542 (.1468)	.0375  (.1541)	.0487  (.1484)	-	-
Competition	.4489***  (.0828)	.4704***  (.0924)	.4385***  (.083)	.4086**  (.1846)	.4297***  (.0816)
Export facilitation	-1.655*** (.053)	-.1788*** (.0582)	-.1705*** (.0538)	-.2412*** (.0613)	-.181*** (.057)
Regulatory quality	.1043   (.1016)	.1192   (.1076)	.1038   (.1017)	.5909***   (.1608)	.0421   (.1663)
Number of obsevations	2598	2579	2511	2507	2507
Wald chi-sq	35.78*** (22)	30.9* (22)	34.39** (22)	36.88** (21)	34.4** (21)
Likelihood-ratio test of rho=0	6.45***	7.26***	5.94***	9.69***	6.66***

Standard errors are in parentheses.*** — statistically significant at p < 0.01 level; ** — statistically significant at p < 0.05 level; * — statistically significant at p < 0.1 level.

All other coefficients are significant and have expected sign.  Indeed, introduction of new products, investments in R&D, size, foreign ownership, employment of advanced technologies, percentage of employees having tertiary education have significant (at p< 0.05 level) and positive effect on the propensity to export.

Export intensity equation. Table 4 presents results of the estimation of the impact of services input variables on the export intensity of manufacturers.  Hausman-Taylor estimation procedure employed at this stage allows controlling for endogeneity of services input variables caused by their correlation with unobserved individual level heterogeneity. All the five equations have Wald chi-square significant at one percent level. The inverse Mills ratio is significant at p< 0.01 level and positive, which reflects the significance of the first-stage selection equation. In conformity with the main research hypothesis electricity and telecommunications sector have significant (p < 0.01) impact on export intensity of manufacturers. According to data from Table 4, the obstacles created by these two service sectors for the business activities of individual manufacturers have negative impact on their export performance. The obstacles formed by finance sector also have negative impact on manufacturer’s exporting. However, this impact is not statistically significant.

Similarly, the overall liberalization of service sector (EBRD index of infrastructure reform) and reforms in banking sector (EBRD index of banking sector reform) have significant (at 5 percent level) and positive effect on export performance of downstream firms. Thus deep reforms and liberalization in such service sectors as electric power supply, railways, roads, telecommunications and water supply as well as in banking sector substantially improves export activity of manufacturing firms. These findings, in general, provide support for the main research hypothesis of the study that the enhancement of services sector positively and significantly influences export performance of downstream industries.

The effects of firm specific characteristics – innovations, research and development, employment of advance technologies, employee skills, size and foreign ownership - are generally significant (in most specifications at 5 percent level) and have expected signs. Introduction of new product and services, investment in research and development as well as employment of advanced technologies (high-speed, broadband internet connection) increase competitiveness of the manufacturing firms at global markets and thus encourages export intensity.

Firm size and foreign ownership also have positive and significant impact on the expansion of export activities. Large firms have more advantages in accessing to finance, necessary for establishing distribution networks at global markets. Generally they have more resources for investments necessary for attaining of competitive advantage globally.  This is especially true for transition economies.  Foreign ownership, in turn, facilitates transfer of advanced managerial expertise, skills and technologies that makes firm more competitive at international markets.  Employee skills variable is also expected to have a positive impact on export performance. This variable measured as a percentage of employees with tertiary education, is supposed to enhance firm’s productivity and thus to improve its competitiveness at export markets. The results in table 4 show that employee skills have positive effect on export intensity; however, not like in selection equation this effect is not significant in all specifications.

Among the environmental variables competition, regulatory quality and membership in European Union are important predictors of export performance of manufacturers.  Study results show that competition measured as a pressure on companies to develop a new product and reduce costs encourages export intensity of manufacturers (significant at p< 0.01). Regulatory quality has positive and significant at one percent level effect (non-significant in equations 4 and 5) on export intensity. Better business environment reduces costs of doing business, improves competitiveness and thus makes it easier to expand business activities at export markets.

Table 4. Export intensity equation

Variables	Coefficients
	I	II	III	IV	V
Dependent Variable: Export Intensity (EI)
Electricity as an obstacle	-2.281*** (.6238)	-	-	-	-
Telecommunications as an obstacle	-	-6.093*** (.7216)	-	-	-
Finance as an obstacle	-	-	-.1498 (.5666)	-	-
EBRD index of infrastructure reform	-	-	-	9.297**  (3.885)	-
EBRD index of banking sector reform	-	-	-	-	9.3904*** (2. 837)
Innovation during last 3 years	11.894***  (1.563)	11.929***  (1.472)	12.724***  (1.575)	4.116***  (1.069)	11.358***  (1.513)
R&D during last 3 years	5.878***    (1.492)	5.348***    (1.452)	2.817*    (1.479)	2.281    (1.434)	5.630***   (1.608)
Technological level of company (High-speed internet connection)	8.989***   (2.185)	7.674***   (2.082)	6.053***   (2.277)	7.823***   (2.511)	5.737***   (2.183)
Employee skills	.0527**   (.0265)	.048*   (.0259)	.034*   (.0267)	.014   (.0274)	.0669**   (.0303)
Foreign ownership	24.513***    (2.205)	23.741***    (2.127)	23.604***    (2.300)	11.606***    (2.035)	23.725***    (2.310)
Size (Small firm)	-44.511***    (3.88)	-44.052***    (3.747)	-43.043***    (3.933)	-16.047***    (3.626)	-45.776***    (4.409)
Size (Medium firm)	-16.066***    (1.67)	-15.981***    (1.665)	-16.370***    (1.729)	-9.592***    (1.724)	-17.406***    (1.832)
Location in Capital	-8.914**  (3.856)	-2.335 (3.439)	-9.131** (3.875)	11.451*** (3.94)	-10.029** (4.012)
European Union country	6.336*** (1.518)	6.078*** (1.561)	6.733*** (1.607)	-	-
Competition	10.851***  (1.136)	10.689***  (1.093)	9.902***  (1.135)	4.173***  (.4676)	9.457***  (1.094)
Regulatory quality	4.556*** (1.212)	5.235*** (1.264)	4.520*** (1.242)	-4.515 (4.907)	-5.564 (5.772)
Inverse Mills ratio	21.651*** (3.525)	19.990***     (3.235)	19.054***     (3.598)	6.510**   (2.844)	19.499***  (3.619)
sigma_u	90.479	99.150	100.946	87.660	96.306
sigma_i	13.732	13.355	13.804	13.731	13.892
rho	.9774	.9821	.9816	.9760	.9796
Number of obsevations	2598	2598	2511	2507	2507
Wald chi-sq (df)	769.02*** (22)	825.44*** (22)	760.75*** (22)	698.87*** (21)	727.90*** (21)

Standard errors are in parentheses.*** — statistically significant at p < 0.01 level; ** — statistically significant at p < 0.05 level; * — statistically significant at p < 0.1 level.

The non-significant coefficients of this variable in equations 4 and 5 can be explained by their correlations with respective EBRD indexes used in these specifications. The same is true for the variable which reflects membership in European Union. Industry effect is controlled but not reported in Tables 3 and 4.

Robustness checks. The empirical strategy of this study involves two-stage Heckman procedure to deal with selection bias problem, and the Hausman-Taylor IV estimation method to control potential endogeneity of services inputs variable. Thus, series of robustness checks of selection bias problem and endogeneity issue are performed in this study. We restrict our robustness checks to effects of services input variables on firm’s export intensity (see Table 5).

Table 5. Robustness checks of the link between services input variables and export intensity (standard errors in brackets).

№	Services Input Variables	Estimation Methods
		Baseline Hausman-Taylor IV estimation	Hausman-Taylor IV estimation with different selection variable (time necessary to comply with all procedures required to export goods)	Hausman-Taylor IV estimation with different selection variable (cost associated with all the procedures required to exporting)	Tobit regression with country and industry fixed effects	Hausman-Taylor IV estimation with services input variables replaced by their industry-country-year averages	2-SLS Fixed Effects estimation
1	2	3	4	5	6	7	8
1	Electricity as an obstacle	-2.28*** (.6238)	-2.25***   (.6051)	-2.27***   (.6210)	-1.46***  (.5373)	-3.04***   (1.119)	-3.84**   (1.654)
2	Telecommunications as an obstacle	-6.09*** (.7216)	-6.02***   (.7041)	-6.00***   (.7076)	-2.32*   (1.353)	-3.83***   (1.365)	-31.67  (26.362)
3	Finance as an obstacle	-.149 (.5666)	-.255   (.5629)	-.021  (.5619)	-2.42**    (1.110)	-4.328**      (1.719)	-7.981*   (4.220)
4	EBRD index of infrastructure reform	9.29**  (3.885)	9.98** (3.915)	8.678**   (3.841)	4.853**   (2.162)	-	-32.0***   (12.188)
5	EBRD index of banking sector reform	9.39*** (2. 837)	11.97***   (2.914)	9.54***   (2.767)	9.97***   (2.125)	-	-23.1***  (9.008)

Standard errors are in parentheses.*** — statistically significant at p < 0.01 level; ** — statistically significant at p < 0.05 level; * — statistically significant at p < 0.1 level.

First, we check whether the study results are sensitive to alternative exclusion restrictions in the probit model. In particular, instead of employing a composite factor score of trade barriers we use separate variable that influence fixed costs of exporting (time necessary to comply with all procedures required to export goods) and variable that impacts variable costs (cost associated with all the procedures required to exporting). Reported in the fourth and fifth columns of Table 5, the coefficients of the all five services input indicators, received after the estimation of the models with the application of the alternative exclusion restrictions, are very similar to the baseline estimates (the third column). Second, we employed Tobit regression to deal with sample selection bias, which is an alternative approach to the Heckman’s two-stage method.  The results of Tobit estimation presented in the sixth column of Table 5, suggest that the study results are quite robust to this check. Third, following Dollar et al. (2006) we replace subjective measures of services inputs by their country-industry-year averages. This allows to copy with the endogeneity problem caused by reverse causality of this variables. The results of Hausman-Taylor estimation of average values of services input variables, showed in the seventh column of Table 5, generally are similar to baseline estimations. The only difference is that the average value of finance indicator is statistically significant now. Therefore, we can conclude that our results are robust.

Finally, as a second robustness check against endogeneity of services input variables, we employ alternative to Hausman-Taylor approach: fixed effects panel IV estimator[8].  The results of these estimations are generally consistent with the results of baseline model (the eighth column of Table 5). However, there are some important differences from earlier results. In particular, the EBRD indexes are significant and negative in fixed-effects 2-SLS model, which contradicts to our expectations. Thus, there is weak support of the assertion that EBRD indexes estimates are robust. Also the financial institution’s variable found to be statistically significant, while telecommunication indicator non-significant predictor of export intensity in this specification.

6. Conclusions

In this paper, we examine the impact of services sector inputs on the export performance of downstream industries in transition economies. To our best knowledge this is the first attempt to explore the empirical link between the efficiency of service industry and export activity of manufacturing sector for transition economies.

To measure services sector performance, the two kinds of indicators were employed in this paper. First one includes the subjective measures of individual manufacturers as to how much of an obstacle for their business they consider the performance of the following three service sectors: electricity, telecommunications and finance. Another type of variables employed in this study was EBRD policy reform indices [EBRD 2009] which reflect the overall liberalization of services sector.

Generally, the key finding of the study is that improvement in the services sectors would enhance the export performance of manufacturers in transition economies. In particular, the study results suggest that reducing constraints and obstacles originating from inefficiencies in electricity, telecommunication, infrastructure and banking will encourage export performance of downstream industries. Thus, advancing liberalization reforms in telecommunications, electric power, railway transport, road transport, and water distribution sectors as well as in banking sector will stimulate expansion of export activities of manufacturers. Our results also suggest that services reform impacts more strongly on the intensity of existing exporters than it does in encouraging new exporters or new export markets.

This paper looks at firm specific factors that affect the export performance of manufacturers in transition economies as well. Consistent with the results in existing research, we find that firm specific characteristics such as the introduction of new products, investment in research and development, employment of advanced technologies, and employee skills are key drivers of export performance in the manufacturing sectors of transition economies. Introduction of new products and services, investment in research and development as well as employment of advanced technologies (high-speed, broadband internet connection) increase the competitiveness of the manufacturing firms in global markets and thus improve export performance.

We find that the firm’s size and foreign investment do matter as well. These factors significantly and positively affect not only the decision to export, but also export intensity of manufacturers. Large firms have more advantages in accessing to finance, which is necessary to establish distribution networks in foreign markets. Generally, they have more resources for the investment necessary for attaining of competitive advantage globally.  This is especially true for transition economies.  Foreign ownership, in turn, facilitates transfer of advanced managerial expertise, skills and technologies that makes the firm more competitive in international markets. We also find that other factors such as trade facilitation, regulatory quality, the degree of competition, membership in European Union also positively affect exports.

The results of the study have several policy implications. The first insight is that an efficient service sector infrastructure represents a strategic and underexploited resource of export enhancement that can be influenced by policy makers. To stimulate export performance of manufacturing industries policy makers must emphasize further reforms and liberalization of their services sectors. These reforms must be focused on providing adequate access to services for downstream industries and thus on reducing their costs of doing business.  Moreover, government should create favorable conditions for attracting foreign direct investments and encourage investments in innovation, research and development, employment of advanced technologies. A final policy point is that reducing trade related costs, through trade and customs procedures facilitation, would also increase exports. Private entrepreneurs should also expect that that their investments in innovation, research and development, employee skills and advanced technologies will be beneficial for their export activity. In summary, the results of this study provide information for policymakers and stakeholders that will facilitate elaboration of policy interventions aimed at improvement of export performance of manufactures in transition economies.

Referecnces 

1. Alvarez, R. and. Lopez, R. A. 2005. Exporting and performance: evidence from Chilean plants. Canadian Journal of Economics, vol. 38(4), pp. 1384–1400.
2. Arnold, J.M., Mattoo, A. and Narciso, G. 2006. Services inputs and firm productivity in Sub-Saharan Africa: evidence from firm-level data. The World Bank, Policy Research Working Paper № 4048.
3. Arnold, J.M., Javorcik, B. and Mattoo, A. 2011. Does services liberalization benefit manufacturing firms: evidence from the Czech Republic. Journal of International Economics, vol. 85 (1), pp. 136-146.
4. Balchin, N. and Edwards, L. 2008. Trade-related business climate and manufacturing export performance in Africa: a firm-level analysis. Paper presented at the 2008 Conference of the African Econometric Society (AES), Pretoria, South Africa. Available at: http://www.africametrics.org/documents/conference08/day2/session5/balchin_edwards.pdf
5. Baldwin, J.R. and Gu, W. 2004. Trade liberalization: export-market participation, productivity growth and innovation. Oxford Review of Economic Policy, vol. 20(3), pp. 372-392.
6. Bernard, A. B. and Jensen, J. B. 1995. Exporters, jobs, and wages in U.S. manufacturing: 1976–1987. Brookings Papers on Economic Activity: Microeconomics, pp. 67–119.
7. Bernard, A. B. and Jensen, J. B. 1999. Exceptional exporter performance: cause, effect, or both? Journal of International Economics, vol. 47(1), pp. 1–25.
8. Bernard, A. B. and Jensen, J. B. 2004. Exporting and productivity in the USA. Oxford Review of Economic Policy, vol. 20(3), pp. 343–357.
9. Bernard, A. B. and. Wagner, J. 1997. Exports and success in German manufacturing. Review of World Economics, vol. 133(1), pp. 134–157.
10. Bernard, A. B. and Wagner, J. 2001. Export entry and exit by German firms. Review of World Economics, vol. 137(1), pp. 105–123.
11. Berulava G. 2012. The Impact of Services Sector on Export Performance of Manufacturing Firms in Transition Economies. Moambe, Bulletin of the Georgian National Academy of Sciences, vol.6(3), pp. 154-162.
12. Blalock, G. and Gertler, P. J. 2004. Learning from exporting revisited in a less developed setting.  Journal of Development Economics, vol. 75(2), pp. 397–416.
13. Clarke, G. 2005. Beyond tariff and quotas: why don’t African manufacturing enterprises export more? The World Bank, Policy Research Working Paper WPS3617.
14. Clerides, S. K., Lach, S. and Tybout, J. R. 1998. Is learning by exporting important? Microdynamic evidence from Colombia, Mexico, and Morocco. Quarterly Journal of Economics, vol. CXIII (3), pp. 903–947.
15. Dollar, D., Hallward-Driemeier, M. and Mengistae, T. 2006. Investment climate and international integration. World Development, vol. 34(9), pp. 1498-1516.
16. EBRD. 2009. Transition report 2009: transition in crisis? European Bank for Reconstruction and Development. Available at: http://www.ebrd.com/downloads/research/transition/TR09.pdf
17. Eschenbach, F. and Hoekman, B. 2006. Services policy reform and economic growth in transition economies. Review of World Economics, vol.142(4), pp. 746-764.
18. Fernandes A. M and Paunov, C. 2012. Foreign direct investment in services and manufacturing productivity growth: Evidence for Chile. Journal of Development Economics, vol. 97 (2), pp. 305-321.
19. Greenaway, D. and Kneller, R. 2004. Exporting and productivity in the United Kingdom. Oxford Review of Economic Policy, vol.  20(3), pp. 358–371.
20. Greenaway, D. and. Yu, Z. 2004. Firm-level interactions between exporting and productivity: industry-specific evidence.  Review of World Economics, vol. 140(3), pp. 376–392.
21. Hausman, J. A., and Taylor, W. E. 1981. Panel data and unobservable individual effect. Econometrica, vol.49 (6), pp. 1377-1398.
22. Heckman J. J. 1979. Sample selection bias as a specification error. Econometrica, vol.47(1), pp. 153–161.
23. Helpman, E., Melitz, M., and Rubinstein, Y. 2008. Estimating trade flows: trading partners and trading volumes. Quarterly Journal of Economics, vol. 123(2), pp. 441-487.
24. Iwanow, T. and Kirkpatrick, C. 2008. Trade facilitation and manufactured exports: is Africa different? World Development, vol. 37(6), pp. 1039-1050.
25. Melitz, M.J. 2003. The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica, vol. 71 (6), pp. 1695-1725.
26. Shepotylo O. 2009. Export diversification across industries and space: Do CIS countries diversify enough?  Discussion Papers 20, Kyiv School of Economics.
27. Wagner J. 2007. Exports and productivity: a survey of the evidence from firm-level data. The World Economy, vol. 30(1), pp. 60-83.
28. World Bank. 1996.  The East Asian Miracle: Economic Growth and Public Policy.World Bank Policy Research Reports (Oxford University Press).

---