foreign languages and trade technical university košice, herl’any, october 14-15, 2010 jarko...
TRANSCRIPT
Foreign Languages and Trade
Technical University Košice, Herl’any, October 14-15, 2010
Jarko Fidrmuc OeNB Vienna, Comenius University Bratislava,
CESifo Munich, OEI Regensburg
Jan Fidrmuc Brunel University London, CEPR, and CESifo
The research was largely completed during Jarko Fidrmuc’s stay at the University of Munich. The opinions are those of the author and do not necessarily reflect the official viewpoint of the Oesterreichische
Nationalbank or of the Eurosystem. We acknowledge CESIUK support from the Operational Program of Research and Development (OP VaV) in the framework of the European Regional Development Fund (ERDF).
- 2 -
Literature Review – I
• Gravity models usually include also dummies for
common languages as a control variables.
• Helpman, Melitz and Rubinstein (2008) derive the
gravity equation by in a model with heterogenous firms
which stresses the link between productivity and
export performance of firms.
• Their empirical results indicate that common
languages are an important part of fixed costs related
to market entry.
- 3 -
Literature Review – II
Mélitz (2008)
• Official and non-official indigenous languages
• Language impact measured using dummy
variables
(if official or spoken by more than 20%) or
communicative probability
• Only indigenous languages (Ethnologue
database)
- 4 -
Literature Review – III
Rauch (1999, 2001),
Rauch and Trindade (2002),
Bandyopadhyay, Coughlin and Wall (2008)
• Ethnic-networks increase trade
• Rauch and Trindade (2002): ethnic Chinese
networks in SE Asia increase trade by at least
60%
- 5 -
Our Contribution
• First to study effect of native and foreign
(learned) languages alike
- Trade often relies on communication in non-
native languages
• Unique extensive dataset on language
proficiency
in the EU
- 6 -
Data–Foreign Languages
• Special Eurobarometer 255: Europeans and their
Languages, November - December 05.
• Nationally representative surveys; only EU nationals
included.
• Respondents were asked on their language skills:
- Native language(s),
- up to 3 other languages that they speak well enough
to have a conversation,
- Self-assessed proficiency of foreign languages:
basic (not used here), good, and very good.
- 7 -
English (good/very good skills) French (good/very good skills)
Foreign Languages in Europe – I
- 8 -
Foreign Languages in Europe – II
German (good/very good skills) Russian (good/very good skills)
- 9 -
Communicative Probability
• Probability that two randomly selected individuals from two different countries can speak sufficiently well the same language
- English• Languages spoken by at least 10% of population
in at least 3 countries- German - French - Russian (only in Eastern Europe)
• We compute the overall communication probability based of possible multiple knowledge of English, French, and German.
- 10 -
Communicative Probability
EU15 NMS/ACs EU29
English 22 6 13
German 7 1 3
French 5 0 1
Russian 0 4 1
Cumulative 30 6 16
- 11 -
Average Cumulative Com. Probabilities–EU15 (English, German, and French)
0
5
10
15
20
25
30
35
40
45
50
- 12 -
Average Cumulative Com. Probabilities–EUROPA29(English, German, and French)
0
5
10
15
20
25
30
35
- 13 -
Data–Further Variables
• Trade and GDP data are taken from the IMF (IFS and DOT)
• All variables are converted to euro.
• PISA test results in 2006 (because of country availability),
• Public and private expenditures on education in 2000.
• We cover EU15, new member states, and the candidate
countries.
- 14 -
Gravity Model–Core Variables
• Trade (in logs) between countries i and j, Tijt,
• Log of output of i and j, yit and yjt,, measured in
nominal EUR,
• Distance between i and j, dij
• Common border dummy, bij.
• A dummy for former federations in CESEE, fij.
ijijijjtitijt
fbdyyT4321
- 15 -
Gravity Model–Languages
• Communication probabilities: Pfij - English, - English, French, German- Cumulative cummulative probability for English,
French and German
• In this version we do not include dummies for
common languages.
F
fijffijijijjtitijt
PfbdyyT,4321
- 16 -
Gravity Model–Panel Structure
ijtjtit
F
fijffijijijjtitijt
PfbdyyT ,4321
• Time-varying country dummies following Baldwin
and Taglioni (2006): - Country-specific time-invariant and
time-varying omitted variables- Country-specific measurement problems- This lowers the possible endogeneity problems.
- 17 -
OLS-Results for EU15
Variable OLS1 OLS2 OLS3
Intercept 15.705 *** 15.568 *** 15.659 *** (63.338) (59.522) (60.502)
GDP 0.915 *** 0.920 *** 0.923 *** (65.436) (65.426) (64.564)
Distance -0.767 *** -0.757 *** -0.759 *** (-31.489) (-29.496) (-29.934)
Contiguity 0.541 *** 0.519 *** 0.466 *** (15.831) (14.848) (13.152)
EMU 0.460 *** 0.454 *** 0.408 *** (11.991) (11.814) (10.541)
English 1.078 *** 1.101 ***
(11.231) (11.377)
French 0.048
(0.467)
German 0.241 ***
(3.466)
Cumulative 0.573 *** (8.760)
N 1470 1470 1470
Adjusted R2 0.965 0.966 0.964
- 18 -
OLS-Results for EUROPA29Variable OLS1 OLS2 OLS3
Intercept 18.027 *** 17.981 *** 18.016 *** (101.297) (100.013) (100.553)
GDP 0.916 *** 0.924 *** 0.916 *** (79.985) (80.472) (79.903)
Distance -1.047 *** -1.051 *** -1.043 *** (-50.708) (-50.409) (-50.162)
Contiguity 0.384 *** 0.405 *** 0.361 *** (9.902) (10.231) (9.190)
Federations 2.401 *** 1.880 *** 2.426 *** (26.099) (15.301) (26.252)
EU 0.038 0.031 0.058
(0.813) (0.664) (1.236)
EMU 0.188 *** 0.193 *** 0.165 *** (4.925) (4.935) (4.292)
English 0.653 *** 0.622 ***
(5.523) (5.280)
French -0.479 **
(-2.759)
German 0.051
(0.411)
Russian 1.642 ***
(6.159)
Cumulative 0.336 *** (3.739)
N 5634 5634 5634
Adjusted R2 0.923 0.924 0.923
- 19 -
Controlling for Endogeneity
• Proficiency in foreign languages and trade may be
endogenous:
- People learn foreign languages of their main
economic partners (e.g. the rise of interest in
Chinese courses in the last decade)
- People forget languages which are not frequently
used
(e.g. Russian in CEECs)
• OLS results may be biased.
- 20 -
Instrumental Strategy
We compare several sets of instrumental variables:
• Language groups are our preferred instruments
- We use dummies for countries with Germanic and
Romanic languages.
- Unlike common languages, language groups are not
correlated with free trade areas (e.g. Germany and
Norway, France and Spain).
• Pisa tests results are valid but weak instruments.
• Public expenditures on education are invalid instruments.
• All instruments work worse for the CESEE.
- 21 -
IV-Results for EU15: Language Groups
Variable GRP1 GRP2 GRP3
Intercept 15.647 *** 14.132 *** 15.151 *** (58.735) (24.909) (46.572)
GDP 0.916 *** 0.944 *** 0.932 *** (67.546) (45.501) (64.578)
Distance -0.763 *** -0.610 *** -0.716 *** (-30.266) (-11.349) (-23.783)
Contiguity 0.542 *** 0.430 *** 0.425 *** (16.368) (5.164) (11.099)
EMU 0.464 *** 0.403 *** 0.408 *** (12.215) (6.987) (10.760)
English 1.165 *** 1.423 ***
(5.926) (6.392)
French 1.950 ***
(3.346)
German 0.241
(0.409)
Cumulative 0.953 *** (5.762)
N 1470 1470 1470
Adjusted R2 0.965 0.957 0.963
Sargan statistics 1.429 2.440 1.322
[0.232] [0.486] [0.250]
- 22 -
English English French German Cumul.
GRP1 GRP2 GRP2 GRP2 GRP3
Germanic (country 1) 0.179 0.062 *** 0.015 *** 0.011 0.201 ***
(20.572) *** (7.591) (4.356) (1.032) (21.840)
Germanic (country 2) 0.179 0.062 *** 0.015 *** 0.011 0.201 ***
(20.533) *** (7.589) (4.310) (1.017) (21.780)
Germanic (country 1 & 2) 0.142 *** -0.018 *** 0.069 ***
(21.245) (-6.094) (7.968)
Romance (country 1) -0.135 *** 0.027 *** -0.013
(-17.529) (8.035) (-1.335)
Romance (country 2) -0.135 *** 0.027 *** -0.013
(-17.530) (8.038) (-1.334)
Romance (country 1 & 2) 0.095 *** 0.007 ** -0.010
(14.010) (2.239) (-1.171)
F-test of excluded instruments 233.17 301.77 23.63 18.95 262.57
[0.000] [0.000] [0.000] [0.000] [0.000]
First Stage Equation for EU15: Lang. Groups
- 23 -
IV-Results for EU15: PISA Variables
Variable PISA1 PISA2 PISA3
Intercept 14.569 *** 8.864 *** 11.255 *** (43.416) (3.714) (11.129)
GDP 0.924 *** 1.031 *** 0.999 *** (61.338) (19.890) (35.436)
Distance -0.680 *** -0.132 -0.390 *** (-22.279) (-0.576) (-4.472)
Contiguity 0.562 *** 0.137 0.114
(15.281) (0.724) (1.217)
EMU 0.540 *** 0.362 *** 0.404 *** (12.401) (3.342) (6.412)
English 2.774 *** 4.420 ***
(8.590) (5.206)
French 6.101 **
(2.401)
German 1.832 *
(1.749)
Cumulative 3.861 *** (5.559)
N 1470 1470 1470
Adjusted R2 0.958 0.860 0.899
Sargan statistics 0.100 5.533 0.017
[0.752] [0.137] [0.895]
- 24 -
First Stage Eq. for EU15: PISA Instruments
English English French German Cumul.
PISA1 PISA2 PISA2 PISA2 PISA3
Reading performance 0.257 *** 0.499 *** 0.012 -0.468 *** 0.184 ***
(country 1) (11.771) (8.910) (0.216) (-5.670) (5.419)
Reading performance 0.256 *** 0.498 *** 0.018 -0.457 *** 0.185 ***
(country 2) (11.739) (8.895) (0.321) (-5.535) (5.454)
Mathematic performance -0.034 0.087 0.210 ***
(country 1) (-0.651) (1.619) (2.693)
Mathematic performance -0.036 0.085 0.210 ***
(country 2) (-0.681) (1.587) (2.701)
Science scale -0.223 *** -0.170 ** 0.274 ***
(country 1) (-3.385) (-2.536) (2.817)
Science scale -0.221 *** -0.174 *** 0.262 ***
(country 2) (-3.348) (-2.600) (2.697)
F-test of excluded instruments 77.40 30.27 2.85 6.78 16.55
[0.000] [0.000] [0.009] [0.000] [0.000]
- 25 -
IV-Results for EUROPA29: Language GroupsVariable GRP1 GRP2 GRP3
Intercept 17.830 *** 18.526 *** 17.171 *** (87.713) (62.796) (45.300)
GDP 0.910 *** 0.920 *** 0.901 *** (76.705) (51.200) (64.632)
Distance -1.040 *** -1.131 *** -0.961 *** (-48.162) (-27.348) (-24.371)
Contiguity 0.371 *** 0.736 *** 0.131
(9.428) (4.903) (1.532)
Former Federations 2.416 *** 2.053 *** 2.702 *** (25.955) (2.851) (19.193)
EU -0.101 -0.023 -0.192 * (-1.412) (-0.335) (-1.850)
EMU 0.220 *** 0.535 *** 0.019
(5.506) (3.947) (0.269)
English 2.702 *** 2.921 ***
(3.326) (3.401)
French -1.996
(-1.599)
German -7.137 ***
(-3.126)
Russian -0.121
(-0.059)
Cumulative 3.796 *** (3.011)
N 5634 5634 5634
Adjusted R2 0.919 0.871 0.902
Sargan statistics 3.747 49.251 3.200
[0.053] [0.000] [0.074]
- 26 -
First Stage Eq. for EUROPA29: Lang. Groups
English English French Germ. Russian Cumul.
GRP1 GRP2 GRP2 GRP2 GRP2 GRP3
Germanic (country 1) 0.097 *** 0.007 *** 0.019 *** -0.012 ***
(30.475) (7.090) (7.057) (-5.415)
Germanic (country 2) 0.097 *** 0.007 *** 0.019 *** -0.012 ***
(30.417) (7.178) (7.063) (-5.597)
Germanic (country 1 & 2) 0.166 *** 0.002 ** 0.037 *** 0.012 ***
(50.895) (2.240) (13.067) (5.394)
Romance (country 1) -0.051 *** -0.039 *** 0.014 *** -0.008 *** -0.009 *** -0.038 *** (-10.91) (-11.02) (12.90) (-2.66) (-3.860) (-5.54)
Romance (country 2) -0.052 *** -0.039 *** 0.014 *** -0.008 *** -0.009 *** -0.037 *** (-10.90) (-11.00) (13.03) (-2.71) (-3.93) (-5.52)
Romance (country 1 & 2) 0.029 *** 0.023 *** -0.007 ** 0.009 ***
(7.089) (17.788) (-2.045) (3.133)
Slavonic (country 1) 0.020 *** 0.001 0.006 * 0.000
(4.876) (0.528) (1.828) (0.019)
Slavonic (country 2) 0.020 *** 0.001 0.007 ** -0.001
(4.983) (0.532) (2.094) (-0.280)
Slavonic (country 1 & 2) -0.015 *** -0.001 -0.021 *** -0.042 ***
(-2.96) (-0.59) (-4.63) (-11.95)
F-test of excl. instr. 60.42 603.63 87.12 59.81 21.98 17.18
[0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
- 27 -
IV-Results for EUROP29: PISA TestsVariable PISA1 PISA2 PISA3
Intercept 18.427 *** 15.148 18.369
(109.943) (0.688) (89.785)
GDP 0.918 *** 1.157 0.918
(76.026) (0.704) (74.720)
Distance -1.106 *** -1.003 -1.099
(-51.901) (-1.497) (-40.568)
Contiguity 0.294 *** 0.056 0.264
(8.802) (0.162) (6.690)
Former Federations 2.371 *** -7.612 2.404
(32.038) (-0.104) (28.169)
EU 0.056 -0.321 0.055
(0.985) (-0.117) (0.960)
EMU 0.236 *** -0.397 0.211
(7.248) (-0.124) (5.654)
English 0.466 0.549
(0.891) (0.065)
French 1.006
(0.214)
German 6.891
(0.617)
Russian 34.414
(0.140)
Cumulative 0.443
(0.879)
N 4870 4870 4870
Adjusted R2 0.936 0.570 0.935
Sargan statistics 0.911 1.744 0.921
[0.340] [0.418] [0.337]
- 28 -
First Stage Eq. for EUROPA29: PISA Tests
English English French Germ. Russian Cumul.
PISA1 PISA2 PISA2 PISA2 PISA2 PISA3
Reading performance 0.098 *** 0.259 *** 0.030 ** -0.064 *** -0.007 0.103 ***
(country 1) (13.782) (16.399) (2.314) (-3.606) (-0.790) (10.179)
Reading performance 0.099 *** 0.255 *** 0.028 ** -0.064 *** -0.006 0.103 ***
(country 2) (13.851) (16.171) (2.209) (-3.618) (-0.730) (10.160)
Mathematic perf. -0.125 *** 0.070 *** 0.123 *** -0.002
(country 1) (-5.629) (3.896) (4.949) (-0.210)
Mathematic perf. -0.130 *** 0.071 *** 0.118 *** -0.001
(country 2) (-5.860) (3.950) (4.767) (-0.100)
Science scale -0.045 * -0.115 *** -0.035 0.005
(country 1) (-1.942) (-6.131) (-1.334) (0.420)
Science scale -0.035 -0.115 *** -0.030 0.003
(country 2) (-1.498) (-6.133) (-1.164) (0.270)
F-test of excl. instr. 106.910 62.640 8.180 9.200 0.270 57.950
[0.000] [0.000] [0.000] [0.000] [0.951] [0.000]
- 29 -
Economic Significance (EU15)
Consider the OLS and IV (Group) results:
• Coefficients for English communication probability is 1.1-
1.4.
• Average effect in EU15: 27%-36% increase due to English
proficiency (22% average communicative probability).
• UK-IRL trade is increased 3 to four times because of
English proficiency (97% communicative probability).
• Furthermore, NL-UK trade is higher by factor 2.3 to 2.9
(76% com. prob.).
• But also NL-S trade is increased 1.7-2.1 times (52% com.
prob.).
- 30 -
Extensions and Sensitivity Analysis
• The foreign language effects may be non-linear, but
the effects are not robust.
• Hump-shaped effect of English diminishing
returns to language skills.
• Median regression confirms the robustness to
outliers.
• Quantile regression shows that the effects are
highest for the lowest and highest quantiles (5th &
90th).
• The results are largely confirmed also East Europe
and EU27 despite a different language education
history.
- 31 -
Non-linear Effect for EU15
Variable NL1 NL2 NL3
Intercept 14.860 *** 14.749 *** 14.264 ***
GDP 0.900 *** 0.917 *** 1.011 ***
Distance -0.781 *** -0.798 *** -0.787 ***
Contiguity 0.474 *** 0.383 *** 0.405 ***
EMU 0.469 *** 0.419 *** 0.533 ***
English (official language) 1.237 *** 1.792 *** 1.077 ***
French (official language) 0.813 *** 0.116
German (official language) -0.012 0.522 ***
English 3.156 *** 5.431 ***
French -0.206
German -2.472 ***
English squared -2.173 *** -4.264 ***
French squared -0.795 *
German squared 3.116 ***
Cumulative 0.580
Cumulative squared -0.418
N 1470 1470 1470
Adjusted R2 0.975 0.979 0.976
- 32 -
Conclusions
• Language proficiency has a strong effect on trade.
• Large trade gains are possible through better foreign
language education.
• The gains are comparable to those of a monetary
integration.
• Improving English proficiency does not require
abandoning national languages.
• Large gains are possible at little cost.