• Keine Ergebnisse gefunden

- Model Specifcaton

Im Dokument Trade and Uncertainty (Seite 49-56)

III - Exchange Rate Policy and Exports of Firms

III.II - Empirical Analysis

III.II.III - Model Specifcaton

Based on the theoretcal literature of heterogenous frms and internatonal trade, we distnguish between the extensive and intensive margin of exports and estmate its determinants using two separate model specifcatons. In order to estmate the determinants of the probability to export, that we defne as the extensive margin of trade, we estmate a probit model that includes a number of frm- and country-specifc variables and also industry (k), country (j) and tme (t) fxed-efects and is given by:

Pr(Exporterijkt=1)=Φ

(

β01lnSalesijkt2FirmSizeijkt3Qualityijkt4Eurojt+

β5ERM jt6Pegjt7lnExchangeRatejt8Volatilityjt9RTAjtjktijkt

)

, (III.2) where Exporterijkt is a dummy variable that takes the value one if frm i is an Exporter at tme t and zero otherwise, ln Salesijkt denotes the natural log of total sales, FirmSizeijkt is a measure of frm size in terms of employees30, Qualityijkt is a dummy variable that takes the value of one if the frm has an internatonally-recognized quality certfcaton and zero otherwise.

We have introduced three dummy variables to control for the exchange rate policy in country j:

First, Eurojt takes the value one if the Euro is the domestc currency at tme t in country j and zero otherwise. Second, ERMjt takes the value one if country j takes part in the European Exchange Rate Mechanism II and fnally, Pegjt takes the value one if the domestc currency is pegged to the Euro or to a basket of currencies that includes the Euro. Volatlityjt is our measure of choice for volatlity of the exchange rate of the domestc currency with the Euro for the twelve months of the past year specifed in equaton (III.1) and ln ExchangeRatejt is the natural log of the frst diference of the average exchange rate of country j with respect to the Euro in year t-1. As a control for the degree of trade liberalizaton in a given country, the model includes the variable RTAjt, which is the total

30 Firm size is a binary variable that takes the value one if the frm has less than 20 employees, two if the frm has between 20 and 99 and three if the frm has 100 or more employees.

42 III - Exchange Rate Policy and Exports of Firms

number of RTAs in which country j partcipates in year t. In additon to a standard probit regression, we estmate the model with random-efects and with and without industry, country and year dummies.

Furthermore, we estmate the determinants of the intensive margin of exports, defned as share of sales to foreign markets. The model is given by:

ExportIntensityijkt01lnSalesijkt2FirmSizeijkt3Qualityijkt4Eurojt+

β5ERMjt6Pegjt7lnExchangeRatejt8Volatilityjt9RTAjtjktijkt , (III.3) where ExportIntensityijkt is the share of total producton that is exported of frm i at tme t and varies between 0 and 100. All other variables are the same as described for equaton (III.2).

The models specifed in equatons (III.2) and (III.3) are estmated with and without industry, country and year dummies. In additon, each sector is estmated separately in order to analyse sectoral diferences in the responsiveness to exchange rate policy.

III.III - Results

In this secton we present the estmaton results of the determinants of the extensive and intensive margin of exports. Table III.III.1 shows the results of model (III.2), which estmates the determinants of the extensive margin. In the frst three columns a pooled probit model is estmated, that is augmented with year and industry efects in column two and country fxed-efects are added in column three. Our measure for exchange rate volatlity yields negatve and mostly signifcant estmates. This indicates that lower volatlity in the exchange rate with respect to the Euro increases a frms' probability to export.

The estmates for variables capturing frm specifc characteristcs yield the expected signs and are statstcally signifcant. The natural log of sales, the number of employees and the existence of internatonally recognized quality certfcates in the frm all have a positve impact on the probability and the intensity of exports. Higher sales and holding internatonal and quality certfcates increases the probability to export. For country specifc variables, estmates are rather mixed, especially for the exchange rate agreement dummies. Imperfectons in the fnancial sector, which are likely to be present in many countries of our sample, can disturb the link between exchange rate movements and exports (Berman & Berthou 2009). Nevertheless, our variable capturing movements of the exchange rate has the expected sign and is mostly signifcant. Euro

and ERM II membership increase the probability to export. Our measure for trade liberalizaton yields signifcant positve estmates for all regressions, indicatng that a higher number of signed regional trade agreements in a country increases the probability of a domestc frm to export.

Reliability of our probit estmates may for some variables sufer from unobserved heterogeneity and in this case it is not possible to interpret them as a causal relatonship, but only as correlatons between dependent and independent variables. Therefore, we estmate a linear probability and a probit model in a diferences-in-diferences (DID) design, which are reported in column four to six.

While the interacton term for the year 2009 and ERM II dummy yields always insignifcant estmates, it is always positve and signifcant for the Euro in the probit model. Interactons for diferent years and our dummy for fxed pegs yields mixed results. Thus, once we estmate our model in setng insusceptble for unobserved heterogeneity, the Euro is the only dummy variable measuring exchange rate policy that has signifcant positve impact on the probability to become an exporter. Movements and volatlity of the exchange rate we assume to be exogenous as they depend on the interacton of domestc and foreign factors via fnancial markets.

44 III - Exchange Rate Policy and Exports of Firms Table III.III.1: Determinants of the Extensive Margin

Results of model (III.3), that estmates the efect of exchange rate policy on export intensity or the intensive margin, are presented in Table III.III.2 columns (1) to (3). The estmates of the efect of our measure of exchange rate volatlity on export intensity are always negatve and signifcant.

When estmatng with year and industry dummies, Euro membership increases the export intensity

Diferences in Diferences

(1) (2) (3) (4) (5) (6)

Probit Probit Probit LPM Probit Probit

0.0625*** 0.130*** 0.167*** 0.0302*** 0.121*** 0.129***

(0.00529) (0.00655) (0.00988) (0.00173) (0.00691) (0.00722) Quality 0.470*** 0.424*** 0.356*** 0.123*** 0.381*** 0.373***

(0.0261) (0.0286) (0.0299) (0.00889) (0.0286) (0.0288) Size 0.293*** 0.183*** 0.163*** 0.0585*** 0.199*** 0.183***

(0.0169) (0.0188) (0.0230) (0.00533) (0.0195) (0.0199)

ERM#2009 - - - -0.0370 0.0633 -0.101

(0.0250) (0.0832) (0.0910) Volatlity -0.716* -0.590 -0.875** -0.235* -0.150 -0.266**

(0.432) (0.392) (0.370) (0.142) (0.117) (0.109) RTA 0.00559*** 0.00960*** 0.00556** 0.00184*** 0.00295*** 0.00225***

(0.000756) (0.000983) (0.00251) (0.000248) (0.000298) (0.000629)

Year Dummies No Yes Yes Yes Yes Yes

Industry Dummies No Yes Yes Yes Yes Yes

Country Dummies No No Yes No No No

Observatons 17,354 17,354 17,354 17,354 17,354 17,354

(Pseudo) R^2 0.104 0.216 0.260 0.214 -

-Notes: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 ln Sales

ln ExchangeRate

by slightly over fve percentage points. The impact of ERM II membership is slightly lower with slightly less than fve percentage points. Much lower is the efect of exchange rate pegs with around two percentage points.

When considering the unobserved heterogeneity and estmatng the model with a DID design similar to estmatons on the extensive margin, estmates for the interactons measuring the efect of exchange rate policy, as shown in columns four and fve, are no longer signifcant. Also pegged exchange rates do no longer have a signifcant impact on export intensity. While exchange rate volatlity stll has signifcant negatve impact, it turns insignifcant for exchange rate movements.

46 III - Exchange Rate Policy and Exports of Firms Table III.III.2: Determinants of the Intensive Margin

In order to investgate sectoral diferences, the regressions in column (2) in Tables III.III.1 and III.III.2 are run for each industry in the sample. The sign of the estmates are presented in Table III.III.3 and Tables C.7 and C.8 in the Appendix show the complete regression results. We fnd

Year Dummies No Yes Yes Yes Yes

Industry Dummies No Yes Yes Yes Yes

Country Dummies No No Yes No No

Observatons 17.354 17.354 17.354 17,354 17,354

R^2 0.087 0.191 0.217 0.192 0.194

Notes: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 ln Sales

ln ExchangeRate

which are signifcant only for the following six sectors: other manufacturing, fabricated metal products, retail, hotels and restaurants and IT. Export intensity is afected negatvely in 12 of 18 sectors and the efect is signifcant for fve sectors. An appreciaton of the domestc currency leads to a lower probability to export in 15 of 18 sectors with efects being signifcant for eight sectors, and to a lower export intensity in 14 of 18 sectors, only in fve cases. Surprisingly, an appreciaton increases export intensity of garments and IT sector31.

Euro and ERM II membership increase the probability to export in 12 of 18 sectors and the export intensity in 14 of 18 sectors. Direct pegs yield very mixed estmates with 12 of 18 sectors facing negatve efects on the probability to export and 8 on the export intensity. Trade liberalizaton measured as the number of signed RTAs a has positve and signifcant impact on the probability to export and export intensity for all non-service industries besides food.

Table III.III.3: Directon of Coefcients for Sectoral Regressions

31 Results for the IT sector should be interpreted with cauton, as more then half of the frms in this sector are from Bulgaria.

Probability to Export Export Intensity

Industry Euro ERM Peg ExRate Volatlity Euro ERM Peg ExRate Volatlity

2 Other manufacturing + - - - - + + - +

-15 Food + + - - + - + - -

-17 Textles - + - - - - + - -

-18 Garments - + + - - - + + +

-23-24 Chemicals - - + - + + + + - +

25 Plastcs & rubber + + + - - + + + - +

26 Non-met. mineral products + - - + - + + + + +

27 Basic metals + - - - - + - - -

-28 Fabricated metal products - + - - - + + - -

-29 Machinery and equipment + + + - + + + + - +

31-32 Electronics - + - - - + + - -

-45 Constructon + + + - - + + + -

-50 Other services + + + - + + + + - +

51 Wholesale + + - - + + + + - +

52 Retail + + - - - + - - -

-55 Hotels and restaurants - - - + - + - - -

-60-64 Transport + + - - - + + + -

-72 IT + - - + - - - + +

-Notes: Shaded felds indicate signifcance at 10% level.

48 III - Exchange Rate Policy and Exports of Firms

Im Dokument Trade and Uncertainty (Seite 49-56)