technical barriers to trade and multi-destination rms. - technical barrie… · technical barriers...
TRANSCRIPT
Technical Barriers to Trade and Multi-destination firms.
Preliminary version∗
Lionel Fontagne † Gianluca Orefice ‡
February 19, 2016
Abstract
This paper analyzes the asymmetric trade effect of restrictive Technical Barriers to
Trade (TBT) measures on heterogeneous exporters with a focus on multi-destination
firms. By matching a database on TBT measures that have been raised as of concern
at the WTO (Specific Trade Concerns – STCs), with a detailed panel of French firm ex-
ports, we test the effect of restrictive TBT measures on the different margins of trade:
(i) the probability to export and to exit the export market, (ii) the value exported, and
(iii) export prices. We find that TBT concerns discourage the presence of exporters
in (encourage the exit from) TBT-imposing foreign markets. This negative effect of
TBT is attenuated for higher productivity firms. Interestingly, multi-destination firms
divert their exports to towards TBT-free destinations (extensive margin channel).
Key Words: Non-tariff measures, TBT, Multi-destination Firms, Trade Margins.
JEL Codes: F13, F14.
∗This paper has received funding from the FP7 of the European Commission (EC), under the PRONTOProject (Productivity, Non-Tariff Measures and Openness), grant agreement 613504. Views expressed heredo not engage the EC. We are grateful to participants [...].†PSE - Universite Paris 1 & CEPII. Correspondence: Centre d’Economie de la Sorbonne, 106-112 Bd de
l’Hopital, F-75647 Paris Cedex 13. Email: [email protected].‡CEPII. Correspondence: CEPII, 113 rue de Grenelle 75007 Paris. Email: [email protected].
1
1 Introduction
Technical Barriers to Trade (TBTs), concerning quality, labeling and technical standards,
represent a fixed cost of exporting discouraging small and less productive firms to serve
markets in which TBTs are imposed. With heterogeneous firms, the productivity cut-off
will differ by destination for a given exporting country and there will be selection of firms on
more difficult destinations Chaney (2008). Imposition of TBTs also raises the variable cost
of producing the exported goods: technical standards impose an upgrading or at least an
adaptation of the product or packaging, while different standards across destinations take a
toll on economies of scale. An adjustment at the intensive margin of trade is thus expected
as well. These are the effects of a NTM usually expected in the literature.
There is an additional effect to consider however. Interestingly, the imposition of a
such barrier to trade may lead exporters to divert trade towards other destinations that do
not impose TBT measures. Any exporter indeed compares the (fixed and variable) cost of
satisfying the new standard with the cost of diverting her shipments. Diversion towards a
new destination (at the extensive margin) will impose a fixed cost of entry, while diversion
towards an existing destination (at the intensive margin) will impose an incremental cost of
reaching marginal consumers (Arkolakis 2010). The higher the cost of complying with the
TBT, the higher the probability that exporters will divert trade towards other destinations.
However, given that diversion has a cost, incumbent exporters close to the exporting cut-
off will neither be able to comply with the standard nor to divert their exports (given their
insufficient productivity). The punch line is finally that not all firms will divert their exports
when the trade restrictiveness of a TBT is high. Trade restrictive TBTs will lead to trade
diversion, but differently for firms with different productivity. And small players might well
be simply driven out of the market.
So, the research question of this paper is the asymmetric effect of trade-hampering TBTs
on the export margins of heterogenous firms, but with a focus on the reorientation of exports
of multi-destination exporters.
To proceed, we need to focus on trade restrictive TBTs. The Specific Trade Concerns
raised by affected exporting countries at the TBT committee of the World Trade Organi-
2
zation (WTO) are a good tracker of such situations.1 Exporting countries will concentrate
their claims on the most restrictive TBTs, because there is a cost of complaining: trade
representatives have a certain amount of time and energy to allocate among a potentially
very large number of idiosyncratic regulations. As a consequence, focusing on those TBT-
market pairs spotted by the trade representatives in the committee, we capture the most
trade restrictive measures. This is the sub-sample of technical measures-destinations, for
which trade diversion should be observed.
We combine this information with custom data at firm level for the universe of French
exporters in order to uncover the usual adjustments of exports at the extensive and intensive
margins. In order to be as close as possible from our theoretical motivation, we will define an
”exporter” as a legal unit (here identified by her administrative identifier) exporting within
an HS4 category of products. In other words, a legal unit exporting in two different HS4
categories will be assumed to have paid twice the cost of launching a new variety (once in
each HS4 category). The level of product aggregation (HS4) has been chosen for coherence
with the TBT data (TBT concerns are also recorded at the HS4 level of the Harmonized
System).
Our first hypothesis is that small and less productive firms will be unable to cope with
the additional fixed/variable cost of the restrictive TBT, while larger multi-destination firms
will reorient their exports. Accordingly, stringent TBT will drive small players out of the
market – with an induced effect on market structure and competition – and encourage multi-
destination players reorienting their shipments towards other destination countries in which
they are already present (for which the fixed cost of entry is already paid) with less stringent
TBTs. As we do not have a continuous measure of stringency (either a TBT imposed by an
importer is challenged in the Committee or not), we will simply oppose destinations with
TBT concerns to destination without concerns, and this will be considered within the (firm
specific) range of destinations contemplated by each individual exporter.
Our second hypothesis is that big players will also look for new destinations, and expand
1According to multilateral rules (TBT agreement Art. 2.5): “ A Member preparing, adopting or applyinga technical regulation which may have a significant effect on trade of other Members shall, upon the requestof another Member, explain the justification for that technical regulation. The TBT committee is the placewhere such justification shall be provided.”
3
their geographical scope, as a result of stringent TBTs. They will accordingly bear the cost
of entering new markets, provided that these markets have less stringent TBTs. Here again,
we will simply oppose destinations with and without concerns. These two assumptions will
be confronted to the data subsequently.
Our findings nicely support the above theoretical predictions. We firstly show that the
mean effect of the presence of a stringent TBT within an HS4 product category on a given
destination is to induce exits of exporters. This selection is driven by exporters having
alternative destinations in their portfolio of markets – free of TBT for the considered HS4
category. This diversion effect is robust to the inclusion of a control for firm size (a proxy
of firm productivity). In line with theory again, the presence of a TBT concern is hitting
less the most productive firms. A side effect of this selection is to increase the market share
of incumbents (but less so for incumbents having alternative destinations in their portfolio
and diverting their trade to those TBT free markets). And prices charged by incumbent
exporters mirror this anti-competitive effects of stringent TBTs: they increase, but less
so for diverting firms. Our second series of results also support the theoretical prediction
regarding the search for new – TBT free – destinations. The presence of a stringent TBT
within an HS4 product category on a given destination incites exporters to bear the fixed
cost of entering into new markets. This effect is magnified for exporters having initially a
larger set of destination countries without TBTs, suggesting the presence of economies of
scope in market access – an element absent from our theoretical benchmark.
The rest of the paper is organized as follows. Section 2 summarizes the existing literature
on the trade effect of Technical Barriers to Trade. Section 3 describes the data and some
stylized facts. Section 4 discusses our empirical strategy. Section 5 presents the results. The
final section concludes.
2 Literature Review
TBTs are covered by a WTO agreement: in a nutshell, technical regulations, standards,
and conformity assessment procedures must be implemented in a transparent and non-
4
discriminatory way.2 Member countries are also encouraged to base their measures on
international standards, but they are not obliged to do so.
TBTs are subject to a triennal review by the WTO. The last occurrence of such review led
to a report published in December 2015 and pointing to the rapid increase in the notification
of such measures (Table 1. Since the first review in 1997 the number of notifications of TBTs
has posted a 254% increase. This figure corresponds to the optimistic assumption that all
member countries actually notify.
Since countries hardly follow WTO recommendation of sticking to international standard,
there is a variety of standards of different restrictiveness for a given category of products
across destinations. No need to stress the cost of duplicating the effort of compliance for
businesses. But progress in alleviating the impact of TBTs is generally made at the regional
level or more generally within the framework of deep integration agreements. The impact
of such attemps on exports of member and third parties has been repeatidly examined in
the literature. Baller (2007) use bilateral trade data at the 3-digit level of the SITC and
examines the trade impact of harmonization (i.e. adoption of a common standard, possibly
more stringent than most of the initial ones) and mutual recognition (granting exporters the
possibility of sticking to the least standard in the region) agreements for testing procedures.
A two-stage gravity estimation provides nuanced results. Mutual recognition is shown to
have a strong and positive effect on parties, in contrast to harmonization. Third countries
do not benefit from harmonization? Chen & Mattoo (2008) adopt a different perspective,
as they analyze the implications for third countries of regional initiatives that deal with
technical barriers. The two options of harmonization and mutual recognition should have
different impacts on trade, in particular for third countries exporters. And the more so
that regional standards are associated with strict rules of origin. Harmonization is shown to
impact positively both the likelihood and the volume of trade within the region. Although
the impact of harmonization on a third country’s exports is positively correlated with the
country’s ability to meet the standards, third countries exports are suffering overall. Mu-
tual recognition has similar positive impacts on intra-regional trade but the effect on third
2In WTO parlance: “ Members shall ensure that technical regulations are not prepared, adopted orapplied with a view to or with the effect of creating unnecessary obstacles to international trade ”
5
countries exports to the region is ambiguous, as tight rules of origin annihilate the benefits
of mutual recognition for third countries exporters. This series of conclusion is however very
much centered on harmonization / mutual recognition among developed countries. Disdier,
Fontagne & Cadot (2015) analyze how harmonization of technical barriers between devel-
oped and developing countries affects international trade. Using a gravity equation, they
show that, as a result of the deep integration associated with standards provisions included
in such economic integration agreements, developing countries partners’ trade expands with
the North, but at the expense of their trade with non-bloc developing partners. Moreover,
harmonization on the basis of regional standards impacts negatively the exports of develop-
ing countries to developed markets. Indeed developing countries often lack infrastructures
and technical capabilities making it possible to easily comply with TBTs. Essaji (2008) uses
US import data at the HS8 level for agricultural, mining and manufacturing imports and
examines the impact of technical regulations. The potential endogeneity of technical regu-
lations is addressed using instrumental variables. Results confirm that technical regulations
substantially hinder developing countries’ exports.
While there is ample evidence that TBTs have negative or positive effects on trade
depending on the measure, its enforcement and the market failure potentially corrected
by the measure,3 little is known about how he different margins of trade are impacted by
TBTs, meaning how technical measures shape the distribution of exporters to the imposing
markets. An exception is Bao & Qiu (2012), who study the TBT effects on trade based
on TBT notifications in the period of 1995 - 2008. Using a two-stage gravity model, they
find that TBTs reduce the extensive margin of export but increase the export intensive
margin. However, this result is based on aggregate data and we know little on how individual
exporters of different productivity actually adjust to stringent TBTs. Data limitation is the
main obstacle to uncovering these complex mechanisms.
3See Disdier, Fontagn & Mimouni (2008) for an illustration on SPS and TBTs. Li & Beghin (2012)perform a meta-analysis of the trade impact of SPS and TBT measures for 27 published articles.
6
3 Data and stylized facts
Our analysis overcomes the data limitations experienced in the previous literature by using
two major databases: a recently-constructed database on specific trade concerns (STCs)
and a database of French firm exports.
3.1 The STCs database and examples of TBT measures
TBT measures cover technical regulations, standards and procedures that are not included
in the SPS scope (which relates to human/animal and plant protection). TBTs apply
to technical requirements introduced for: (i) health or safety purposes, (ii) standardize
products, (iii) ensure quality standard or (iv) avoid consumer deception. See figure 1 for an
illustration on whether a measure belongs to SPS or TBT.
The recently-released WTO database on specific trade concerns records the concerns
that have been raised at the dedicated committees of the WTO is available in a quantitative
format.4 In this paper we focus on the concerns raised in the TBT Committee, who provides
WTO members with a forum to discuss issues related to TBT measures taken by other
members (”specific trade concerns” - STCs). When a country complains about a TBT
measure against another WTO member country (the imposing or maintaining country),
it specifies the product of concern, the type of concern regarding the measure and the
objective of the measure concerned. The advantage of this information is that it provides a
systematic set of all TBT measures that are perceived as sizeable trade barriers by exporters
(i.e. measures important enough that countries whose exports are affected raise a ”concern”
to the TBT committee of the WTO). We thus rely on TBT barriers to trade. This represents
the main advantage of using the STC dataset with respect to other sources. Indeed, other
datasets listing the full TBT measures imposed by countries (TRAINS or Perinorm) mix up
together TBT measures that restrict trade with others that may even increase trade.5
Overall, we have 318 Specific Trade Concerns raised at the TBT committee of the WTO
over the 1995-2011 period. Each STC corresponds to a concern raised by one or more
4The dataset is available at http://www.wto.org/english/res_e/publications_e/wtr12_dataset_
e.htm5Some TBT measures indeed may boost trade by addressing problems of asymmetric information or
network externalities (Moenius 2004, Fontagne, von Kirchbach & Mimouni 2005).
7
countries in relation to a TBT measure imposed by one or more of their trading partners.
This figure is low, compared to the cumulated number of notifications over the period 1995-
2009 reported in Table 1 – more than 13,000. We are really observing in this committee
the most stringent TBTs, which is exactly our research strategy. For each concern, we have
information on: (i) the country or countries raising the concern and the country imposing the
measure, (ii) the product codes (HS 4-digit) involved in the concern, (iii) the year in which
the concern was raised to the WTO, and (iv) whether it has been resolved and how. Since
we have detailed data on French firm exports, we focus here on a sub-sample of concerns
raised by the EU over the period 1995-2008 (the period for which we have information on
French firm-level exports). Moreover, given the lag structure of our econometric exercise
(see Section 4), in order to be consistent over the entire paper, we rely on the 1996-2008
period.
In Table 2 we show descriptive evidence on the sample if STCs on TBT we use in our
empirical exercise. In column 2 of Table 2 we show that the number of countries targeted
by (at least one) STC is increasing over time. Similarly in column 3, the number of HS-4
chapters under TBT concern (in at least one STC) are exponentially increasing over the
period we analyze. It clearly emerges the increasing use of STCs on TBT, and consequently
the increasing relevance of TBT measures in hampering trade.
In our dataset we also have information on the primary object of the STC on TBTs, this
allows us to characterize the nature of the underlying TBT measure. In Table 3 we show
the top-five most frequent objectives reported in the list of STCs on TBT. Finally, in Table
4 we show the number of HS-4 chapters under STCs by imposing country. It emerges that
European Union and China have been frequently targeted by STCs on TBTs. This does
not mean that EU and China are particularly active in imposing TBTs, it simply indicates
that EU and China, being attractive markets for exporting countries, are often targeted by
STC when they impose TBT measures.
The main difference between TBT and SPS measures relies on the welfare effect. While
SPS measures, by preventing the imports of unhealthy products, have an impact on the
consumers’ welfare in the imposing countries, TBTs have a limited welfare effect since
they simply regulate technical standards. For example, in 1995 Canada, Peru and Chile
8
complained at the WTO TBT committee against a labeling measure imposed by the French
government, who laid down the official name and trade description of scallops. Complainants
claimed that this measure reduced the competition on the French market as their product will
no longer be able to be sold as ”Coquille Saint-Jacques” although there were no difference
(according to complaining countries) between their scallops and French ones in terms of
color, size, texture, appearance and use (it is claimed they are ”like products”). This
concern was solved after one year with mutual recognition and represents an interesting
example of how TBT measure might hamper trade and have no effect on welfare.6 Another
interesting example of labeling measure preventing trade is the concern raised by Peru in
2001 against EU. According to Peruvian government, the EEC regulation 2136/89 prevented
Peruvian exporters to continue to use the trade description ”sardines” for their products.
Peru submitted that, according to the relevant Codex Alimentarius standards (STAN 94-181
rev. 1995), the species ”sardinops sagax sagax” are listed among those species which can
be traded as ”sardines”. Peru, therefore, considered that the above Regulation constituted
an unjustifiable barrier to trade (Articles 2 and 12 of the TBT Agreement and Article XI:1
of GATT 1994). Also this measure was solved with a mutually recognition.7
Similarly, TBTs on packaging might hamper trade with no effect on welfare for the
imposing countries. On 4 April 2012, Honduras requested consultations with Australia
concerning certain Australian laws and regulations that impose trademark restrictions and
other plain packaging requirements on tobacco products and packaging. Honduras chal-
lenged the following measures: (i) an Act to discourage the use of tobacco products, and for
related purposes, Australia’s Tobacco Plain Packaging Act 2011, and its implementing reg-
ulations; (ii) the Trade Marks Amendment (Tobacco Plain Packaging) Act 2011. According
to the Honduras government such measures imposed by Australia were inconsistent with
Australia’s obligations under TRIPS and GATT agreement.8 Many other countries joined
Honduras on this trade concerns soon after in 2012 (Brazil, Guatemala, Nicaragua, New
Zealand, Ukraine, European Union, Canada, Indonesia, Norway, Philippines) and a panel
6See https://www.wto.org/english/tratop_e/dispu_e/cases_e/ds12_e.htm7See https://www.wto.org/english/tratop_e/dispu_e/cases_e/ds231_e.htm8Namely Articles 2.1, 3.1, 15.4, 16.1, 20, 22.2(b) and 24.3 of the TRIPS Agreement; Article 2.1 and 2.2
of the TBT Agreement; and Article III:4 of the GATT 1994.
9
was established to discuss this concern.9
Similar disputes at the TBT committee of the WTO are rather common, suggesting that
indeed TBTs represent impediments to trade with a marginal welfare effect for consumers
in the imposing countries. For this reason, in the present paper we are not interested in the
welfare effect of TBT in neither imposing nor complaining country. We simply use STCs on
TBT measures as examples of increases in (fixed) trade cost.
3.2 French firm-export data
Individual export data on French firms are provided by French Customs for the 1995-2008
period.10 Our estimations will therefore focus on the 1996-2008 period as we lose the starting
year after calculating the firm’s exit probability and the inclusion of lagged variables in the
estimations (for details see the next section).
The French firm dataset includes export records at the firm, product and market level
for all French exporters.11 Although the dataset classifies product categories using the
Combined Nomenclature at 8 digits (i.e. CN8 is an 8-digit European extension of the HS6
comprising some 10,000 product categories) we aggregate export data at HS-4 level to be
coherent with STC TBT dataset.
Then, as the EU acts as a single country in WTO committees, we restricted our firm-level
sample to only extra-EU export flows. Given the huge amount of observation, in order to
work with a manageable dataset we selected a subsample of relevant destination countries for
French firms: we calculated total export flows by destination market and retained markets
with above-median exports. Next, we dropped those firms that exported only one of two
years within our time period (churning). This strategy reduces any bias from only occasional
exporters.
As acknowledged by Konings & Vandenbussche (2013), one advantage of individual ex-
porter data is their good quality and the possibility to see clearly whether NTMs affect the
intensive/extensive margins of trade, the exit dynamics from foreign markets, and firms’
export price setting. We can also control for firm characteristics in determining the effect of
9https://www.wto.org/english/tratop_e/dispu_e/cases_e/ds435_e.htm10These data are subject to statistical secrecy and have been acceded at CEPII.11We consider legal units, as defined by their administrative identifier.
10
NTMs. However, since we do not have information on turnover, employment or capital for
the universe of French exporters, we rely on export-based measures of firm characteristics.12
4 Empirical Strategy
The main objective of the paper is to study the asymmetric effect of trade stringent TBTs
(as reflected by STCs on TBT measures) on the export margins of heterogeneous firms with
a focus on the multi-destination status of firms. Recall that we aim at putting to the data
series of clear-cut theoretical predictions suggested by the trade theory with heterogeneous
firms and destinations.
The first prediction is that small and less productive firms will be unable to comply with
stringent standard. Their productivity does not allow them to cope with the additional
fixed/variable cost of the restrictive TBT. Larger and multi-destination firms13 will reorient
their exports towards TBT free markets for which they have already paid the fixed cost of
entry.
Our second prediction is that the most productive and multi-destination exporters will
also look for new markets as a result of stringent TBTs. They will further expand their
geographical scope towards TBT free markets and bear the fixed cost of entry.
In the following we will confront sequentially these two predictions to the data for the
French universe of exporters and the whole set of destinations.
4.1 Trade diversion within the initial geographical scope of in-
cumbent exporters
To test our first prediction, we estimate the following equation:
12Data on French firm characteristics are available only for firms with more than 25 employees. Over 50per cent of exporting firms have fewer than 20 employees. To correctly account for the extensive margin ofexports, we do not use data on French firm characteristics.
13Firms with high productivity can reach more destinations because they can enter into markets withhigher fixed or variable cost.
11
yi,s,j,t = α + β1TBTs,j,t + β2 (TBTs,j,t ∗MultiDesti,t−1) + β3MultiDesti,t−1
+β4 ln(size)i,t−1 + β5 (TBTs,j,t ∗ ln(size)i,t−1)
+β6Ln(tariff + 1)s,j,t + φHS2,t,j + µi + εi,s,j,t, (1)
where the subscripts i, s, j, and t stand respectively for firm, HS 4-digit product category
(or 2-digit sector if HS2), destination country and year. Our dependent variables are: (i) a
dummy variable for the firm exiting a certain product-market (a dummy for the firm not
exporting in the current year but having exported the year before); (ii) a dummy variable
for positive trade flows into a certain product-destination market combination to capture
the (firm-product) extensive margin of trade, or participation; (iii) the firm’s export values
(in logs) to capture the intensive margin of trade;14 and (iv) the price of exported goods
(in logs), proxied by unit export values. Despite the dichotomous nature of some of our
dependent variables, we estimate equation 1 via OLS. We rely on simple linear probability
models (LPM) rather than on non-linear probit (or logit) to avoid the incidental parameter
problem due to the sizeable set of fixed effects we include in all regressions. In addition,
LPM provides simple direct estimates of the sample average marginal effect.
TBTs,j,t, reflects the existence, at time t, of an ongoing (unresolved) TBT concern in
product category s between the EU and an importer country j. Sizei,t−1 is a firm-specific
characteristic and captures heterogeneous export performance across firms related to firm
productivity (which is here unobserved). Since we do not have exhaustive information on
French exporters’ balance sheets, we calculate the size variables in terms of exports and not
total sales.15 We define this variable as: ln(size)i,t−1 =∑s⊂S
∑j⊂J
exportsi,s,j,t−1. We use a
one-year lag in our regressions to reflect that firms’ past performance affects future export
decisions.
To investigate how TBT concerns shape the adjustment of multi-destination exporters
14The dependent variable in this regression includes only positive trade values.15The empirical literature has extensively shown that export values are a good proxy for the overall size of
the firm: big exporters are usually larger and more-efficient than are non-exporters (see Mayer & Ottaviano(2008))
12
we interact the TBT dummy with the number of destination markets free of TBT served by
firm i at time t− 1 - MultiDesti,t−1. According to our hypothesis they might prefer leaving
the market imposing the measure. We test this channel of adjustment by introducing in
Equation 1 the interaction between TBT and MultiDesti,t−1. Multi-destinations exporters
are also more productive and larger because of their higher productivity. Thus we need to
control for size since the effect of MultiDesti,t−1 might well absorb the effect of TBT for big
firms. So in Equation (1) we also include the interaction between the TBT dummy and the
size of the firm at t− 1 (in log). Heterogeneous-firm trade models16 suggest that the effect
of an TBT measure on export performance may depend on the size of a firm, provided that
size is associated with productivity, and hence with the ability to overcome the additional
costs of exporting. In other words, not all firms will be able to cope with the higher costs of
new TBT regulations. In order to isolate the effect of TBT concerns from traditional tariff
protection we also control for applied tariffs at the product level (Ln(tariff + 1)s,j,t).17
Finally we add two sets of fixed effects. We include a set of firm fixed effects (µi)
to control for firm-specific and time-invariant unobserved characteristics that might affect
the trade performance of exporters. We also include a set of three-way fixed effects (HS2-
Destination-Year) - φHS2,t,j, to control for country-time-HS2-level varying factors that may
affect trade, such as business cycles, import-demand shocks and multilateral trade resistance
(as highlighted by Head & Mayer (2014)). These three-way fixed effects also control for the
geographic orientation of French exports that may affect the probability of raising a concern.
Moreover, HS2-Destination-Year fixed effects also control for measures imposed by a country
in response to a negative domestic shock in a given sector.
The inclusion of these two wide sets of fixed effects drastically reduces any endogeneity
concern due to the omitted variables bias. Moreover, the inclusion in Equation 1 of Sizei,t−1
captures the impact of firm specific time-varying characteristics on export performance. The
reverse causality arises if the government of a certain destination market introduces a TBT
measure in response to high levels of imports from a specific French firm (and then a concern
16See e.g. Chaney (2008), Arkolakis (2010) and Spearot (2013)17The tariff data come from the TRAINS database. Since TRAINS provides tariff in percentage points
(i.e. 10% ad-valorem tariff listed as 10), we divide tariff by 100 and then compute the price equivalenttransformation log(tariff+1)
13
is raised by the EU against this measure). This is a remote possibility. Even more if we
consider that we use TBT concerns raised by the EU as a whole (and not STCs raised
specifically by France). However, as a robustness check, we estimate our regressions lagging
the TBT variable by one year: given that a concern raised in t− 1 is related to a measure
introduced in t − 2 or earlier, there is low chance that such concern would be driven by
exports at time t.
4.2 Trade diversion beyond the initial geographical scope of in-
cumbent exporters: new markets
In presence of TBT concern in a given destination country, the firm may well exit the
market but add a new TBT-free destination, as suggested by our reasoning above. This will
happen as soon as the fixed cost of exporting into a new destination is below the fixed and
variable cost of staying in the market imposing the TBT and complying with such stringent
regulation. In this case we should see a positive effect on the number of the new destinations
countries at (t+1).
To investigate this second channel of adjustment, we collapse the dataset at firm-product-
year, so that for each firm-product combination we keep track of the total number of new
TBT-free destinations. For each firm-product we then compare the set of destinations at
time t with those at time t− 1 and count the number of new TBT-free destinations.
An alternative strategy of identification of this channel is to compute the probability
for an exporter to serve a new TBT-free destination when confronted to a concern on an
existing destination (for the same HS4 indeed). We just code a dummy if the number of
new TBT-free destinations is ≥ 1.
We thus use the above mentioned variables as dependent variables in the following esti-
mating equation:
yi,s,t = α + β1TBTi,s,t−1 + β2MultiDesti,s,t−1 +
β3 (TBTi,s,t ∗MultiDesti,s,t−1) + φHS4,t + µi + εi,s,j,t, (2)
14
The main explanatory variable here is a dummy being equal to one if the firm-product had
at (t-1) at least one destination with active TBT concern. We also include as a control the
number of destinations served by firm i on product s at time t− 1. This approximates the
multi-destination status of the firm at time t−1. We also interact the two above mentioned
variables to test the peculiar behavior of multi-destination firms facing TBT concerns in
terms of destination portfolio. Finally, we include firm and product-by-year fixed effects to
control for firms specific characteristics and any product specific unobserved shock.
5 Results
5.1 The Effects of TBT on firms’ margins of trade
Results for the estimated Equation 1 are reported in Tables 5, 6, 7 and 8 for respectively
the exit, the extensive margin, the intensive margin and the price charged (here proxied by
the unit value).
In Table 5 we show the results for the exit probability of firms. The mean effect of
the presence of an active TBT concern within an HS4 is to push firms out of the market
(column 1). Interestingly, we observe in column 2 that multi-destination exporters (i.e.
firms exporting in alternative TBT-free destinations) drive this result: the coefficient on
the interaction term between TBT and MultiDesti,t−1 is positive and highly significant in
column 2, while the coefficient on TBT is no longer significant. Finally, we show in column
3 that large firms are less prone to exit in response to a TBT concern, as the interaction
between TBT and firm size is negative and highly significant. While big firms are more able
to cope with TBT (reduced probability of exit the market, as expected), multi-destination
firms are still more likely to exit the market when they face a TBT concern, which is our
focus. Finally, in column 4, as robustness, we use only manufacturing sectors are results
remain unchanged. The rationale for using only manufacturing as robustness check comes
from results in Table A1 showing that the effect of TBT measures concerns mainly the
manufacturing sector.
In Table 6 we show the effect of TBT concerns on the extensive margin of French ex-
porters (participation). Column 1 shows that the presence of TBT concerns reduces the
15
probability of exporting into the destination imposing the measure on the considered HS4
product category. This results is robust across all the specifications reported in Table 6.
In column 2 we show that the mean negative effect of TBT concern is exacerbated for
multi-destination firms: these exporters are even less incline than single-destination firms to
export in destination countries with active TBT concerns. Comparing the stringency and
related cost of the TBT, they are incline to divert their exports towards other destinations
of their geographic portfolio whereby (stringent – as revealed by concerns) TBTs are absent
for this HS4 category of product. Again, this interaction might capture the size effect. So in
column 3 we include also the interaction between size and TBT. The sign of this interaction
is positive and significant, suggesting that big firms are better equipped to comply with
stringent TBTs, but the sign of the interaction between TBT and the number of destination
served by the firm remains negative and significant. Control variables have the expected
sign, firm size is positively related with the export probability, while tariff in destination
country (when significant) negatively affects the participation probability.
In Table 7 we show results for the intensive margin of firms. In this case the presence of
a TBT concern has a positive and significant effect on the intensive margin (of incumbent
firms), and this across all specifications in Table 7.18 This suggests that (stringent) TBTs
act as a barrier to entry reducing the toughness of competition in the imposing country
for the surviving exporters. The positive effect of TBT on the intensive margin is however
reduced for multi-destination exporters (see columns 3 and 4) and magnified for big firms
(see interaction with firm size in columns 3). This is coherent with the evidence showed
above: multi-destination firms exit from the market imposing the measure and so reduce
their presence there (because the have wider portfolio of destinations to exploit), although
big firms are more able to comply with the measure, remain in the market and enjoy the
reduced competition. We accordingly expect a positive average effect of TBT on the export
price of firms, but magnified for big firms. This is what we find in Table 8, where the TBT
dummy has always a positive and significant impact on prices charged by exporters – this
effect being magnified for big firms (column 3). Finally, in Table 9 we show results by using
the same specification as in Equation 1 but using a one year lagged TBT dummy. This
18This result is is coherent with findings of Bao & Qiu (2012)
16
robustness check aims to reduce any endogeneity concern of the TBT concern. Results are
qualitatively the same as those described above.
5.2 The dynamics of multi-destination firms
The evidence reported in Subsection 5.1 suggests a peculiar behavior of multi-destination
firms facing TBT concerns: the presence of a stringent TBT increases the exit probability of
firms, in particular for multi-destination firms (who have wider portfolio of destinations to
exploit). In this sub-section we now want to go further on this line of reasoning: we study
the adjustment of firms at the extensive margin channeling through the inclusion of new
destinations in their geographic portfolio. Namely, does the firm enter new markets after
exiting from the TBT imposing market?
Results for the estimation of Equation 2 are reported in Table 10. As expected, when a
firm faces the presence of a TBT concern on product s in (at least) one of its destinations at
(t-1), she starts shipping the considered HS4 product category at least in one new – TBT-
free – destination in t (see columns 4 to 6 where the probability of serving a new destination
is used as dependent variable). This result is in line with our hypothesis: in presence TBTs,
exporters will balance the cost of complying with this regulation against the fixed cost of
entering in a new market and make the decision of serving a new market when the regulation
is really stringent (as revealed by the concern). More productive exporters, already serving
several destinations, are shown to be more prone to make this decision of entering a new
market. This result is suggesting economies of scope in terms of prospection, networks of
distribution, or even branding of the product.
In order to confirm this unexpected effect, based on the simple theoretical framework
used above, we examine how the number of new destinations served, and thus the extent of
the redeployment of the exporter on a wider geographic scale. Here again, we show that the
number of new TBT-free destinations is positively impacted when the exporter is hit by a
stringent TBT, for the same HS4 product category, on an existing destination (column 1).
But here, this effect seems to be specific to those firms being multi-destination at (t-1) - see
column 3.
17
6 Conclusion
Aiming to uncover the adjustment channels of heterogenous exporters confronted to stringent
technical barriers on certain markets, this paper combined information on TBT concerns in
the dedicated WTO committee with with custom data at firm level for the universe of French
exporters. Theory suggests that small and less productive exporters will be unable to cope
with the additional fixed/variable cost of the restrictive TBT. They will exit. In contrast,
multi-destination firms will reorient their exports towards markets free of TBT concerns (for
the same HS4 category of products). Theory also suggest that a sub-sample of firms will
be productive enough to envisage new destinations, expand their geographical scope, and
bear the associated fixed cost, provided that these markets have less stringent TBTs. These
predictions are confirmed by our data. Stringent TBTs actually drive small players out of
the market; Competition is reduced to the benefit of survivors. Besides multi-destination
players are firstly encouraged reorienting their shipments towards other destinations, clear
of TBT concerns, for which they already paid the fixed cost of entry. On these markets they
adjust at the intensive margin. At the same time, they go for new markets, where pay the
induced fixed cost of entry. The latter channel of adjustment, at the extensive margin, is
magnified for exporters initially diversified geographically. This is suggesting the presence
of economies of scope in foreign market access.
18
Bibliography
Arkolakis, C. (2010), ‘Market penetration costs and the new consumers margin in interna-
tional trade’, Journal of Political Economy 118(6), 1151 – 1199.
Baller, S. (2007), ‘Trade effects of regional standards liberalization: A heterogeneous firms
approach’, World Bank policy research working paper (4124).
Bao, X. & Qiu, L. D. (2012), ‘How Do Technical Barriers to Trade Influence Trade?’, Review
of International Economics 20(4), 691–706.
Chaney, T. (2008), ‘Distorted gravity: the intensive and extensive margins of international
trade’, American Economic Review 98(4), 1707–1721.
Chen, M. X. & Mattoo, A. (2008), ‘Regionalism in standards: good or bad for trade?’,
Canadian Journal of Economics 41(3), 838–863.
Disdier, A.-C., Fontagn, L. & Mimouni, M. (2008), ‘The impact of regulations on agricultural
trade: Evidence from the sps and tbt agreements’, American Journal of Agricultural
Economics 90(2), 336–350.
Disdier, A.-C., Fontagne, L. & Cadot, O. (2015), ‘North-South Standards Harmonization
and International Trade’, World Bank Economic Review 29(2), 327–352.
Essaji, A. (2008), ‘Technical regulations and specialization in international trade’, Journal
of International Economics 76(2), 166 – 176.
Fontagne, L., von Kirchbach, F. & Mimouni, M. (2005), ‘An assessment of environmentally-
related non-tariff measures’, The World Economy 28(10), 1417–1439.
Head, K. & Mayer, T. (2014), Gravity equations: Workhorse, toolkit, and cookbook, Vol.
4, Handbook of International Economics, Gita Gopinath, Elhanan Helpman and Kenneth
Rogoff editors, chapter 4.
Konings, J. & Vandenbussche, H. (2013), ‘Anti-dumping protection hurts domestic ex-
porters’, Review of World Economics 149, 295–320.
Li, Y. & Beghin, J. C. (2012), ‘A meta-analysis of estimates of the impact of technical
barriers to trade’, Journal of Policy Modeling 34(3), 497 – 511.
Mayer, T. & Ottaviano, G. (2008), ‘The happy few: The internationalisation of european
firms’, Intereconomics: Review of European Economic Policy 43(3), 135–148.
Moenius, J. (2004), Information versus product adaptation: The role of standards in trade,
Technical report, International Business and Markets Research Center Working Paper.
Spearot, A. C. (2013), ‘Variable demand elasticities and tariff liberalization’, Journal of
International Economics 89(1), 26 – 41.
19
7 Tables and Figures
Figure 1: TBT vs SPS measures.
Source: WTO website.
Table 1: Total notifications of TBTs per triennal review period
1995-97 1998-00 2001-03 2004-06 2007-09 2010-12 2013-15
1737 2012 2096 2658 4673 5845 6150Source: WTO,Seventh triennal review of the operation and implementation
of the agreement on technical barriers to trade under article 15.4. 3 December 2015, p.12.
20
Table 2: TBT concerns by year
year Number of Countries Number of HS4 items under Average number of TBT
with at least one STC on TBT TBT STC in at least one country concerns by country
1996 3 33 1,06
1997 7 342 16,19
1998 5 355 21,84
1999 3 402 13,35
2000 6 247 10,03
2001 11 280 25,58
2002 11 335 23,61
2003 12 556 38,1
2004 12 454 21,06
2005 13 467 17,61
2006 11 454 18
2007 20 408 16,16
2008 17 495 25,19
Table 3: The top-5 motives for TBT concerns
Objective nb occurrences across STCs Frequency over the total specified objectives*
Protection of Human Health or safety 135 22,60%
Protection of the Environment 67 11,20%
Consumer Information or Protection 48 8,00%
Labelling 28 4,70%
Quality 26 4,30%
Each STC has one or more objectives. One occurrence is thus the combination STC-objective. The total number of occurrencesin the dataset is 598. The frequency reported in the column is thus the ratio between the number of occurrences of aobjective over the total number of occurrences in the dataset.
21
Table 4: Number of HS4 items object of a STC in TBT by imposing country and year
Country 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Argentina 0 0 0 0 0 0 0 161 161 0 0 7 7
Brazil 0 0 158 0 0 167 3 5 2 2 0 10 15
Canada 1 1 0 0 0 0 0 0 0 0 0 1 2
Chile 0 0 0 0 0 167 0 0 0 0 0 0 0
China 0 0 0 0 0 1 210 214 3 31 225 30 197
Colombia 0 0 0 0 0 0 0 0 0 5 0 0 1
Ecuador 0 0 0 0 0 0 0 4 4 0 0 0 0
Egypt 0 2 166 166 3 0 0 0 0 0 0 0 0
European Union 4 179 16 81 82 277 251 426 429 258 261 273 274
Hong Kong 0 0 0 0 0 1 0 0 0 0 0 4 0
India 0 0 0 0 0 2 171 10 14 0 7 18 14
Indonesia 0 0 0 0 0 4 0 0 0 2 0 0 6
Israel 0 3 0 0 0 0 0 0 0 0 0 1 1
Japan 0 0 0 0 167 167 5 2 0 0 14 0 2
Korea 0 0 9 0 49 0 14 10 3 16 5 5 8
Kuwait 0 0 0 0 0 0 0 131 0 0 0 0 0
Malaysia 0 0 0 0 0 0 0 0 0 18 0 0 0
Mexico 0 315 328 0 0 0 0 0 2 2 0 0 0
Moldova 0 0 0 0 0 0 0 0 0 0 0 2 8
New Zealand 0 0 0 167 4 4 4 4 4 4 4 4 0
Peru 0 1 0 0 0 0 0 0 5 11 5 5 5
Philippines 0 0 0 0 0 0 0 0 0 0 2 2 0
Qatar 0 0 0 0 0 0 0 0 0 0 0 8 0
South Africa 0 0 0 0 0 0 3 0 0 170 0 0 161
Switzerland 0 0 0 0 0 0 0 0 5 5 5 5 0
Taipei 0 0 0 0 0 0 0 0 0 0 0 52 1
Thailand 0 1 0 0 0 2 2 0 0 0 0 4 4
Turkey 0 0 0 0 0 0 0 0 0 0 0 2 0
United States 28 0 0 0 6 1 25 170 21 22 29 67 75
Uruguay 0 0 0 0 0 0 0 0 0 0 1 1 0
Venezuela 0 0 0 0 0 0 44 44 0 0 0 0 0
22
Table 5: Exit probability estimation
Dep. Var. Exit Dummy
(1) (2) (3) (4)
TBT 0.033** 0.021 0.028* 0.016
(0.016) (0.016) (0.016) (0.016)
TBT*Log(TBT free destinations)t−1 0.016*** 0.020*** 0.026***
(0.005) (0.005) (0.007)
Log(TBT free destinations)t−1 0.010*** 0.010*** 0.010***
(0.000) (0.000) (0.000)
Log(Firm Size)t−1*TBT -0.007***
(0.002)
Log(Firm Size)t−1 0.025*** 0.022*** 0.022*** 0.023***
(0.000) (0.000) (0.000) (0.000)
Ln(tariff+1) 0.008 0.007 0.007 -0.009
(0.009) (0.009) (0.009) (0.012)
Sample Full Full Full Manuf.
Observations 3,848,704 3,848,704 3,848,704 3,451,593
R-squared 0.070 0.070 0.070 0.057
Firm and destination-chapter-year Fixed Effects always included.
Clustered standard errors by destination-HS4-year in parentheses.
*** p < 0, 01; ∗ ∗ p < 0, 05; ∗p < 0, 1.
23
Table 6: Extensive margin estimation
Dep. Var. Participation Dummy
(1) (2) (3) (4)
TBT -0.070*** -0.035* -0.041** -0.041*
(0.019) (0.019) (0.019) (0.021)
TBT*Log(TBT free destinations)t−1 -0.014* -0.018** -0.010
(0.008) (0.008) (0.011)
Log(TBT free destinations)t−1 0.166*** 0.166*** 0.168***
(0.001) (0.001) (0.001)
Log(Firm Size)t−1*TBT 0.007**
(0.003)
Log(Firm Size)t−1 0.119*** 0.071*** 0.071*** 0.070***
(0.000) (0.000) (0.000) (0.000)
Ln(tariff+1) -0.008 -0.026** -0.026** -0.003
(0.012) (0.012) (0.012) (0.017)
Sample Full Full Full Manuf.
Observations 3,848,704 3,848,704 3,848,704 3,451,593
R-squared 0.117 0.157 0.157 0.146
Firm and destination-chapter-year Fixed Effects always included.
Clustered standard errors by destination-HS4-year in parentheses.
*** p < 0, 01; ∗ ∗ p < 0, 05; ∗p < 0, 1.
24
Table 7: Intensive margin estimation
Dep. Var. Log of export value
(1) (2) (3) (4)
TBT 0.144* 0.242*** 0.170* 0.287***
(0.082) (0.092) (0.098) (0.090)
TBT*Log(TBT free destinations)t−1 -0.055 -0.106*** -0.084**
(0.042) (0.038) (0.036)
Log(TBT free destinations)t−1 0.841*** 0.841*** 0.837***
(0.003) (0.003) (0.003)
Log(Firm Size)t−1*TBT 0.073***
(0.026)
Log(Firm Size)t−1 0.246*** 0.024*** 0.023*** 0.024***
(0.004) (0.004) (0.004) (0.004)
Ln(tariff+1) -0.128** -0.224*** -0.223*** -0.253***
(0.063) (0.060) (0.060) (0.079)
Sample Full Full Full Manuf.
Observations 2,271,564 2,271,564 2,271,564 2,014,243
R-squared 0.314 0.373 0.373 0.360
Firm and destination-chapter-year Fixed Effects always included.
Clustered standard errors by destination-HS4-year in parentheses.
*** p < 0, 01; ∗ ∗ p < 0, 05; ∗p < 0, 1.
25
Table 8: Export price estimation
Dep. Var. Log of Trade Unit Value
(1) (2) (3) (4)
TBT 0.161*** 0.142*** 0.104** 0.145**
(0.055) (0.055) (0.048) (0.059)
TBT*Log(TBT free destinations)t−1 0.015 -0.012 0.014
(0.013) (0.015) (0.021)
Log(TBT free destinations)t−1 -0.033*** -0.032*** -0.040***
(0.002) (0.002) (0.002)
Log(Firm Size)t−1*TBT 0.039***
(0.014)
Log(Firm Size)t−1 0.013*** 0.021*** 0.021*** 0.023***
(0.002) (0.002) (0.002) (0.002)
Ln(tariff+1) -0.348*** -0.345*** -0.344*** -0.362***
(0.036) (0.036) (0.036) (0.057)
Sample Full Full Full Manuf.
Observations 2,271,564 2,271,564 2,271,564 2,014,243
R-squared 0.769 0.770 0.770 0.745
Firm and destination-chapter-year Fixed Effects always included.
Clustered standard errors by destination-HS4-year in parentheses.
*** p < 0, 01; ∗ ∗ p < 0, 05; ∗p < 0, 1.
26
Tab
le9:
Rob
ust
nes
sch
eck
usi
ng
lagg
edT
BT
dum
my
Exte
nsi
veE
xit
Inte
nsi
veT
UV
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
TB
Tt−
1-0
.074***
0.0
08
0.0
21
0.0
12
0.1
22
0.4
03***
0.1
75**
0.1
58**
(0.0
25)
(0.0
22)
(0.0
20)
(0.0
19)
(0.0
96)
(0.0
92)
(0.0
69)
(0.0
68)
TB
Tt−
1*L
og(T
BT
free
des
tin
atio
ns)
t−1
-0.0
15**
0.0
27***
-0.0
62*
0.0
10
(0.0
07)
(0.0
04)
(0.0
33)
(0.0
12)
Log
(TB
Tfr
eed
esti
nat
ion
s)t−
10.1
66***
0.0
10***
0.8
41***
-0.0
33***
(0.0
01)
(0.0
00)
(0.0
03)
(0.0
02)
Log
(Fir
mS
ize)
t−1
0.119***
0.0
71***
0.0
25***
0.0
22***
0.2
46***
0.0
24***
0.0
13***
0.0
21***
(0.0
00)
(0.0
00)
(0.0
00)
(0.0
00)
(0.0
04)
(0.0
04)
(0.0
02)
(0.0
02)
Ln
(tar
iff+
1)-0
.008
-0.0
25**
0.0
08
0.0
07
-0.1
29**
-0.2
23***
-0.3
48***
-0.3
45***
(0.0
12)
(0.0
12)
(0.0
09)
(0.0
09)
(0.0
63)
(0.0
60)
(0.0
36)
(0.0
36)
Ob
serv
atio
ns
3,848,7
04
3,8
48,7
04
3,8
48,7
04
3,8
48,7
04
2,2
71,5
64
2,2
71,5
64
2,2
71,5
64
2,2
71,5
64
R-s
qu
ared
0.1
17
0.1
57
0.0
70
0.0
70
0.3
14
0.3
73
0.7
69
0.7
70
Clu
ster
edst
an
dard
erro
rsby
des
tin
ati
on
-HS
4-y
ear
inp
are
nth
eses
.F
irm
an
dd
esti
nati
on
-ch
ap
ter-
yea
rF
ixed
Eff
ects
alw
ays
incl
ud
ed.
***p<
0,0
1;∗
∗p<
0,0
5;∗
p<
0,1
.
27
Tab
le10
:T
he
Dynam
iceff
ect
ofT
BT
Nu
mb
erof
New
Pro
bab
ilit
yto
serv
ea
new
TB
T-f
ree
des
tin
ati
on
TB
T-f
ree
des
tinati
on
(1)
(2)
(3)
(4)
(5)
(6)
TB
Tt−
10,1
25***
0,0
87***
0,0
07
0,3
49***
0,2
55*
**
0,1
90***
(0,0
04)
(0,0
04)
(0,0
04)
(0,0
05)
(0,0
05)
(0,0
08)
N.
ofd
esti
nat
ions t−1
0,0
52***
0,0
51***
0,1
32*
**
0,1
31***
(0,0
01)
(0,0
01)
(0,0
01)
(0,0
01)
TB
Tt−
1*N
.of
des
tin
atio
ns t−1
0,0
67***
0,0
55***
(0,0
04)
(0,0
06)
Ob
serv
atio
ns
1920748
1920748
1920748
1920816
19208
16
1920816
R-s
qu
ared
0,1
15
0,1
33
0,1
34
0,1
54
0,1
76
0,1
76
Clu
ster
edst
an
dard
erro
rsby
firm
-HS
4in
pare
nth
eses
.F
irm
an
dH
S4-y
ear
Fix
edE
ffec
tsalw
ays
incl
ud
ed.
***p<
0,0
1;∗
∗p<
0,0
5;∗
p<
0,1
.
28
Appendix
29
Tab
leA
1:T
he
effec
tof
TB
Ton
man
ufa
cturi
ng
and
agri
cult
ure
sect
ors.
Exte
nsi
vem
arg
inE
xit
pro
bab
ilit
yIn
ten
sive
marg
inT
rad
eU
nit
Valu
e
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
TB
T-0
.075
***
0.0
01
0.0
36**
0.0
03
0.1
51*
0.1
07
0.1
63***
0.0
65
(0.0
21)
(0.0
28)
(0.0
17)
(0.0
17)
(0.0
89)
(0.1
04)
(0.0
59)
(0.0
83)
Log
(Fir
mS
ize)
t−1
0.11
8***
0.1
39***
0.0
26***
0.0
19***
0.2
40***
0.2
94***
0.0
12***
-0.0
04
(0.0
00)
(0.0
02)
(0.0
00)
(0.0
01)
(0.0
04)
(0.0
15)
(0.0
02)
(0.0
06)
Ln
(tar
iff+
1)0.
029*
-0.0
61***
-0.0
07
0.0
29**
-0.0
75
-0.2
34**
-0.3
71***
-0.3
52***
(0.0
17)
(0.0
18)
(0.0
12)
(0.0
12)
(0.0
84)
(0.0
95)
(0.0
57)
(0.0
41)
Sam
ple
Man
uf.
Agri
.M
anu
f.A
gri
.M
anu
f.A
gri
.M
anu
f.A
gri
.
Ob
serv
atio
ns
3,45
1,59
3372,5
91
3,4
51,5
93
372,5
91
2,0
14,2
43
241,2
02
2,0
14,2
43
241,2
02
R-s
qu
ared
0.10
40.2
41
0.0
57
0.2
07
0.3
00
0.4
85
0.7
45
0.7
78
Fir
man
dd
esti
nati
on
-ch
ap
ter-
yea
rF
ixed
Eff
ects
alw
ays
incl
ud
ed.
Clu
ster
edst
an
dard
erro
rsby
des
tin
ati
on
-HS
4-y
ear
inp
are
nth
eses
.
***p<
0,0
1;∗
∗p<
0,0
5;∗
p<
0,1
.
30