export market dynamics and plant-level productivity: impact of tariff reductions and exchange-rate...

Scand. J. of Economics, 2012DOI: 10.1111/j.1467-9442.2012.01716.x

Export Market Dynamics and Plant-LevelProductivity: Impact of Tariff Reductionsand Exchange-Rate Cycles∗

John BaldwinStatistics Canada, Ottawa, ON, K1A 0T6, [email protected]

Beiling YanStatistics Canada, Ottawa, ON, K1A 0T6, [email protected]

Abstract

In this paper, we examine how changes in tariff rates and industry-specific real exchangerates affect the entry/exit process to export markets and productivity growth. Using theexperience of the Canadian manufacturing sector over three decades, we find that firms inexport markets enjoy faster productivity growth than non-participants. The size of the growthadvantage depends on whether real exchange rates are increasing or decreasing. The increasein the value of the Canadian dollar during the post-2000 period almost completely offset theproductivity growth advantages enjoyed by new exporters during this period.

Keywords: Entry; exit; productivity growth

JEL classification: F1; F3; L1; O4

I. Introduction

The process of birth and death plays an important role not only in resourcereallocation, but also in the productivity growth that comes from industrialrenewal. While plant closure results in job losses, creative destruction hasbeen shown to improve productivity by replacing the least productive plantswith more productive plants (Foster et al., 2001).

In this paper, we examine the turnover process in one particular market –the export market. We consider how the turnover process is affected bytrade costs, and we study the relationship between productivity growth andturnover in export markets. The entry/exit process in export markets is partof a larger turnover process that occurs as firms renew themselves. Thefocus of most studies in the trade literature has been on only part of this

* Reproduced with the permission of the Minister of Industry. We thank Shenjie Chen, BenTomlin, Dan Ciuriak, the seminar participants at Statistics Canada and at the Department ofForeign Affairs and International Trade, and two anonymous referees for helpful comments.

C© Her Majesty the Queen in Right of Canada 2012. Reproduced with the permission of the Minister of Industry. Publishedby Blackwell Publishing, 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

2 J. Baldwin and B. Yan

process − entry to export markets.1 In this study, we recognize that notall forays into new markets will be successful. We also examine firms thatexit the export market, and we assess how well they perform compared tofirms that survive in export markets.

In this paper, we further the understanding of firm dynamics in an openeconomy by asking how firms’ entry and exit respond to changing tradecosts. In previous studies on exporter survival (Kimura and Fujii 2003;Esteve Perez et al., 2004; Baggs, 2005), the authors have not examinedthe relationship of survival to changes in trade costs, especially to thefluctuations in real exchange rates. Here, we remedy this. We use a uniqueCanadian panel dataset of manufacturing plants over the last three decades,during which the ease of access to foreign markets has changed becauseof shifts in Canada/US bilateral tariffs and real exchange rates. We useindividual firm data to focus on the heterogeneous responses of firms tochanges in different trade costs (i.e., tariffs as opposed to exchange rates).Trade models with heterogeneous firms (Bernard et al., 2003; Melitz, 2003;Baldwin and Gu, 2009) suggest that the impact can vary across plants thathave different levels of efficiency.

Finally, we examine whether export participation improves productivitygrowth. The entry/exit process is intrinsically interesting not only becauseof what it reveals about experimentation with new markets, but also becauseof the way movements into and out of exporting lead to improvements inproductivity. To evaluate whether exporting improves productivity growth,two econometric techniques are used to address non-random sample se-lection problems: a standard ordinary least-squares (OLS) regression thattakes into account plant characteristics, and a propensity-score matchingtechnique together with the differences-in-differences method.

In Section II, we discuss the analytical frameworks that inform thesubsequent empirical analysis. In Section III, we introduce the data. InSection IV, we provide a preliminary comparison of productivity perfor-mance across three decades of adaptation. We find that the productivitygrowth advantage enjoyed by export participants in the earlier periods dis-appeared in the post-2000 period. In Section V, we present multivariateresults. An important finding is that exchange-rate fluctuations explainalmost all of the shifts in the productivity growth of exporters relativeto non-exporters. After allowance is made for changes in tariffs and realexchange rates, export participants nevertheless enjoy faster productivitygrowth than non-participants. Using matching techniques, in Section VI,

1 There are several studies that have also looked at the productivity performances of exiters,for example, Baldwin and Gu (2003) for Canada, Clerides et al. (1998) for Colombia,Bernard and Wagner (1997) for Germany, and Girma et al. (2003) for the UK. For acomplete list, see table A1 in Wagner (2007).

C© Her Majesty the Queen in Right of Canada 2012. Reproduced with the permission of the Minister of Industry.

Export market dynamics and plant-level productivity 3

we further examine two sources of the superior productivity performanceof exporters: the self-selection and the learning-by-exporting effect. Weconclude in Section VII.

II. Analytical Framework

The empirical investigation is guided by both the industrial organizationand international trade literature in order to predict how changes in tariffsand exchange rates are expected to affect both entry to/exit from exportmarkets and productivity growth.

Entry and exit decisions to export markets are explicitly considered inthe heterogeneous-firm-based models of international trade (Bernard et al.,2003; Melitz, 2003; Melitz and Ottaviano, 2008), where the existence ofsunk cost associated with exporting (such as marketing) means that profit-maximizing firms will enter export markets only if the present value of theirprofits exceeds the fixed costs of entry. Likewise, an incumbent exportercontinues only if the present value of expected operating profits exceeds thesunk cost and salvage value. Because firms with higher efficiency levelscan obtain high operating profit, they achieve non-negative net profit evenif they incur sunk costs for exports. The most efficient firms self-select intoexport markets. These are also the firms that survive in export markets.

These models suggest that the factors shaping a firm’s profitability areimportant determinants of export decisions. A reduction in tariffs − ordepreciation in a home currency − makes exporting more profitable. Itdecreases the threshold of efficiency level for exporting and it enhancesthe probability of survival for exporters. Likewise, currency appreciationdecreases operating profits. As a result, the threshold of efficiency levelrequired for survival rises, and plants with lower productivity will be forcedto exit export markets or to shut down.

By itself, this framework leads to the prediction that a reduction in tariffsor a depreciation of a home currency increases the probability that moreefficient plants will enter export markets and it decreases the probabilitythat more efficient plants will exit from export markets.

It has been widely argued that exporting leads to productivity gains. Thishypothesis has received mixed empirical support. It has been found thatexporting has a significant positive effect on a firm’s productivity growthin, for example, Slovenia (De Loecker, 2007), Canada (Baldwin and Gu,2003), Italy (Serti and Tomasi, 2008), and sub-Saharan African countries(Van Biesebroeck, 2005). However, the opposite has been reported forChile, China, Germany, and Mexico, for example.2 These cross-country

2 See Wagner (2007) for a survey of the literature. For other recent surveys, see Lopez (2005)and Greenaway and Kneller (2007).



differences might be the result of variations in the trading environmentsfacing different countries. In particular, new opportunities offered by tradeliberalization, as well as currency depreciation, vary across countries andtime periods. These differences might account for variations in the reportedimpact of exporting on productivity. However, these variations might alsoarise because the differences in the potential gains depend on how firmsadapt to changing trade costs.

Tariff reductions can affect productivity growth among exporters relativeto non-exporters in several ways. First, there is a link between productivitygrowth and the growth in output that exporters obtain from improved ac-cess to foreign markets. Market expansion can be associated with improve-ments in productivity growth, because of traditional scale/scope economies(Kaldor, 1966; Scott, 1989), because of the learning-by-doing effect (Arrow,1962), or because of the effect of changes in capacity utilization (Berndtand Fuss, 1986).

Second, there is a link between imported intermediate inputs and pro-ductivity growth. Bernard et al. (2003) have argued that tariff cuts can leadto productivity gains within surviving plants; while the prices of importedintermediates decline, surviving plants can replace domestically producedinputs with cheaper imported inputs, which brings about within-plant pro-ductivity gains. Whether exporters benefit more than non-exporters dependson their differential ability to substitute cheaper intermediates for workers.The Canadian experience in the period after the Free Trade Agreement(FTA) suggests that two changes occurred in this area: the import con-tent of exports increased over this period (Cross, 2002) and Canadianmanufacturers increased their degree of disintermediation (Baldwin et al.,2004).

Fluctuations in exchange rates affect the productivity growth of exportersrelative to non-exporters in several ways. A rising value of the Canadiandollar relative to the US dollar (equivalent to lowering home tariffs andraising foreign tariffs) reduces export sales by making Canadian exportsmore expensive in the US. This also makes imported intermediate inputscheaper. Using the same logic as in the above discussion, currency ap-preciation can therefore have a negative demand effect and an ambiguousimport effect, on the relative productivity growth of exporters.

In addition, the appreciation of the Canadian dollar has, in the past, beenassociated with resource booms that have led to the expansion of the mar-ket for non-exporters relative to exporters (Macdonald, 2008; Chen et al.,2010). If market growth is associated with productivity growth because oftraditional scale/scope economies, because of the learning-by-doing effect,or because of the effect of capacity utilization, then exporting plants shouldbe expected to perform relatively worse than domestic-oriented plants dur-ing periods when the Canadian dollar has appreciated.



Thus, these various considerations indicate that the impacts of tariffcuts and exchange-rate movements on the productivity growth of exportparticipants relative to non-participants depend on the relative stimulus thatcomes from output markets as opposed to input markets. If plants substitutecheaper imported imports for labour (as in the model of Bernard et al.,2003), tariff cuts and currency appreciation are expected to generate within-plant productivity gains. Whether exporters benefit more than non-exportersdepends on their differential ability in substituting cheaper intermediateinputs for workers. However, if market growth is positively associated withproductivity growth, because of traditional scale/scope economies, becauseof the learning-by-doing effect, or because of the effect of changes incapacity utilization, tariff cuts and currency depreciation are expected togenerate more productivity growth gains for export participants than fornon-participants.

III. Data

Data Sources

The plant-level data used in this study come from the Annual Survey ofManufacturers (ASM) by Statistics Canada; this is a longitudinal databasethat tracks Canadian manufacturing plants over time. In this paper, we usethe entire sample from the ASM. The ASM database contains informationon shipments, value-added, employment, age of plants, export status, andindustry affiliation.3 Labour productivity is defined as real value-addedoutput per employee, where real value-added is calculated using plant-levelnominal value-added output deflated by corresponding industry deflators.4

Information on export status is available in 1979, 1984, 1990, 1993, 1996,1997, 1998, and 1999 for plants that filled out the long form, and it isavailable annually from 2000 onwards for all plants. Small plants for the1984–1990 and 1990–1996 periods (those who filled in the short-formquestionnaires) are assumed to be non-exporters.5

The plant-level micro data are linked to industry-level data on tariffsand real exchange rates that cover 236 four-digit Canadian manufacturing

3 Industry affiliation is classified using the 1980 four-digit Canadian Standard IndustrialClassification (SIC) from 1979 to 1997, and the six-digit North American Industry Clas-sification System (NAICS) from 1997 onwards. In this paper, we use the SIC version ofthe ASM for the 1984–1990 and 1990–1996 periods, and the NAICS version of the ASMfor the post-2000 period (2000–2006). For the post-2000 micro data, some records are im-puted. These imputed micro records have problematic measures of relative value-added andemployment. Therefore, they are excluded from parts of this analysis.4 The nominal value-added in the ASM includes payments for services.5 According to a 1974 survey that collected export data for all plants, only 0.4 percent ofplants that filled in the short-form questionnaires reported exports (Baldwin and Gu, 2003).



industries. Bilateral tariffs between Canada and the US are those createdand used by Trefler (2004).6 The industry-specific real exchange rate eit

is constructed as the normal exchange rate Nt (expressed in terms of USdollars per Canadian dollar), deflated by relative US and Canadian industryprices, pu

it and pcit, respectively (i.e., eit = Nt (pc

it/puit), where i and t

denote industry and year, respectively). The nominal exchange rate is fromthe CANSIM database of Statistics Canada. Canadian industry prices pc

it

are gross output prices from the KLEMS productivity database maintainedby the Economic Analysis Division at Statistics Canada, while the USgross output prices pu

it are derived from the productivity databases of theUS National Bureau of Economic Research (NBER) and the US CensusBureau’s Center for Economic Studies (CES).7

Three Decades of Adaptation

Three panels of continuing plants, which differ in terms of the trading envi-ronment that each faced, are used to examine the linkages between export-ing and productivity growth. The first panel covers the period 1984–1990,the second panel covers the period 1990–1996, and the third panel coversthe period 2000–2006.

The three panels cover the periods prior to, during, and after the imple-mentation of the FTA between Canada and the US. Tariff rates fell in thefirst two periods, but reductions became larger in the second period, follow-ing the introduction of the FTA in 1989. In the 2000−2006 period, the FTAtariff reductions with the US had been implemented. More importantly, thissecond period was marked by a resource boom that was associated withan appreciation of the Canadian dollar against the US dollar, which madeexporting to the US market less advantageous.

6 We are grateful to Alla Lileeva for providing us with the tariff data. For details on thesources and construction of the tariff data, see the appendix in Trefler (2004).7 Other studies have sometimes used an alternative industry-specific real exchange rate, bycalculating the exchange rate as a weighted average of exchange rates between Canada andits trading partners, with weights being countries’ trade shares for each industry (Baggset al., 2009). There are two problems with the use of this rate for studies of the reaction offirms to movements in exchange rates. First, the Canadian trade-weighted real exchange ratessuffer from the lack of variability across industries because the US is Canada’s predominanttrading partner across manufacturing industries. Second, it is assumed that the price adjust-ments to the nominal exchange-rate movements are the same across industries. Baldwin andYan (2007) have found large heterogeneity in the responses of industries. The price-adjustedreal exchange rate is a better indicator of an industry’s international competitiveness. It mea-sures the price spread between an industry’s product price and the landed price charged byindustries in other countries. When the real exchange rate increases, Canadian manufacturers− because they lose competitiveness − are placed under more pressure by US exporters.When the real exchange rate has a downward trend, then Canadian industries become morecompetitive, relative to their foreign counterparts.



Table 1. Average annual changes in tariff rates and real exchange rates (inpercentage points)

1984–1990 1990–1996 2000–2006

Real US/Canada exchange rate 1.6 −1.9 5.5(1.5) (1.5) (3.1)

Canadian tariff against US −0.4 −0.8(0.3) (0.4)

US tariff against Canada −0.2 −0.4(0.5) (0.7)

Average tariff between Canada and US −0.3 −0.6(0.3) (0.5)

Source: Compilation by the authors from various data sources: Statistics Canada’s CANSIM Table 176–0064;NBER productivity database; Statistics Canada’s gross output deflator from KLEMS productivity database; tariffrates from Trefler (2004).Notes: Average annual changes are calculated as differences between the first and last years of the variable,divided by the number of years. Standard deviations are in parentheses.

Average tariff reductions between Canada and the US in the first twoperiods were large, with annual average rates of 0.3 and 0.6 percent-age points during 1984–1990 and 1990–1996, respectively. While tariffdeclines reduced trade costs between Canada and the US, movements inthe exchange rate sometimes complemented these reductions, and at othertimes countered them. In the first period, the real Canadian dollar appre-ciated at an annual average rate of 1.6 percentage points; in the secondperiod, it depreciated at an average annual rate of 1.9 percentage points.Then, it increased at a rate of 5.5 percentage points from 2000 to 2006.The standard deviations for the real exchange rates are large, indicatingsubstantial variations across industries (Table 1). Therefore, the decade ofthe 1990s differs from both the 1980s and the post-2000 period in terms ofthe incentive to export. In the 1990s, both tariff reductions and depreciationin the real exchange rate were experienced. In the 1980s and the post-2000period, there was an exchange-rate appreciation.

Plant Groupings by Export Transition

In this paper, we focus on export decisions and performances among con-tinuing plants, and we make two comparisons: (1) continuing non-exportersare compared to entrants into export markets; (2) continuing exporters arecompared to exiters from export markets. At the beginning of a period,producers consist of either exporters or non-exporters. At the end of theperiod, they might have maintained the status quo (i.e., continuing exportersor continuing non-exporters) or they might have changed their export sta-tus (i.e., exiters from export markets or entrants to export markets). Theproductivity growth of those that retained their status quo is compared



Table 2. Differentials in productivity performance (in percentage points)

Labour productivity level Labour productivity growth(log of real value-added per worker) (annual log changes in real value-added

at the beginning of a period output per worker) during a period(1) (2)

Mean differences: entrants less non-exporters1984–1990 0.18∗ 2.00∗1990–1996 0.22∗ 5.29∗2000–2006 0.06∗ 0.13

Mean differences: exiters less continuingexporters1984–1990 −0.05∗ −3.06∗1990–1996 −0.20∗ −6.87∗2000–2006 −0.17∗ −0.16

Source: Compilation by the authors from the Canadian ASM data.Notes: Mean differences are computed from regressions in the form of Y = αi+β1 Dpt , where Y is the level orgrowth of labour productivity, and Dpt is a dummy variable that takes the value of one if a plant p enters exportmarkets during the period and zero if it remains a non-exporter. Similarly, Dpt equals one if a plant p exitsexport markets during the period and zero if it remains an exporter. The regression is run with industry-specificfixed effects (αi). ∗denotes significance at 5 percent level or better.

to those that changed their export status. If exporting leads to better pro-ductivity performance, entrants should have higher productivity growth thancontinuing non-exporters, and continuing exporters should have higher pro-ductivity growth than those exiting the export markets.

IV. Preliminary Comparison of Productivity Performance

Table 2 summarizes the average productivity performance of plants withdifferent transitions to export markets. The results are consistent with aself-selection process − only the more productive plants enter and survivein export markets (see Column 1 of Table 2). Over all three periods, theproductivity level of entrants to export markets is significantly higher at thebeginning of a period than it is for non-exporters. This agrees with previousstudies such as Das et al. (2007), Eaton et al. (2008), and Ruhl and Willis(2008). The results also show that exporters who survive in export marketsenjoy significantly higher productivity levels at the beginning of a periodthan those who subsequently stop exporting.

However, exporting is not always associated with higher productivitygrowth (see Column 2 of Table 2). Productivity growth is higher for con-tinuing exporters than for exiters for the first two periods (1984–1990and 1990–1996), but the difference becomes statistically insignificantfor the 2000–2006 period. Moreover, the magnitude of the differencevaries across the first two periods. On average, annual labour produc-tivity growth was around 6.9 percentage points higher for continuing ex-porters in the early 1990s (1990–1996), when tariffs were falling and the



exchange rate was depreciating. In comparison, the annual labour produc-tivity growth was only 3.1 percentage points higher in the late 1980s(1984–1990), when tariffs were falling but the exchange rate was ap-preciating. Similar patterns emerge when non-exporters and entrants arecompared.

Thus, the size of the gap in productivity growth between export par-ticipants and non-participants varies depending on the period examined.The gap was largest in the early 1990s, when new opportunities in ex-port markets were the greatest because of the size of tariff cuts and thecoincident depreciation in the exchange rate. The superior performancewas diminished in the late 1980s when the appreciation of the Canadiandollar partially offset the effect of a decline in tariffs. However, signifi-cantly, the difference in performance became statistically insignificant inthe post-2000 period (2000–2006), when the primary external influence oncompetitiveness was an appreciating dollar.

V. Results on the Impact of Changing Tariffs and ExchangeRates

In order to understand the forces behind these cross-period differences,we use a multivariate analysis to examine the impact of tariff barriers andexchange rates on plant dynamics. We pool two panels of continuing plants,one over the 1984–1990 period and the other over the 1990–1996 period.8

Impact on Entry/Exit Dynamics in the Export Market

The probability of entering and exiting export markets is estimatedas a function of industry-level shocks and plant-specific characteris-tics. Industry-level shocks include four-digit industry-wide average annualchanges in tariffs �τ it, real exchange rates �eit, and real gross output�lnQit.9 Plant-level characteristics Zpt0 include relative productivity levelLPpt0 (relative to the mean productivity level of plants in the same SICfour-digit industry), relative employment Lpt0 (relative to mean industryemployment), age, and nationality of ownership (domestic versus foreign-controlled) at the start of a period. Interaction terms between changes intariffs or real exchange rates and initial labour productivity level LPpt0 and

8 The 2000–2006 panel data are excluded because we do not have tariff data for this period.Tariff changes between Canada and the US during this period were close to zero during thispost-FTA and post-North American FTA (NAFTA) period.9 To prevent possible endogeneity, we measure industry-specific real gross output as the sumof real shipments at the four-digit SIC level minus the real shipment of the plant itself.



plant size Lpt0 are also included.

Prob(Dpt = 1) = �(ai + at + ß1�τi t + ß2�eit + ß3�τi t∗LPpt0 + ß4�τi t

∗L pt0

+ ß5�e∗i t L Ppt0 + ß6�e∗

i t L pt0 + ß7�lnQit + γ Z pt0) (1)

Here, � is the standard normal cumulative distribution, and Dpt is adummy variable that takes the values of one, if a plant p enters exportmarkets during the period, and zero, if it remains a non-exporter. Simi-larly, Dpt equals one if a plant p exits export markets during the period,and it equals zero if it remains an exporter. The probit model controlsfor three-digit industry-specific fixed effects αi and period-specific fixedeffects αt.

Because coefficients in probit models with interaction terms are hard tointerpret (Ai and Norton, 2003; Norton et al., 2004), the results presentedhere focus on the marginal effects. Marginal effects for interaction termsare calculated according to

∂2�

∂(x1)∂(x2)= �′(xβ)

∂2(xβ)

∂x1∂x2+ �′′(xβ)

∂(xβ)

∂x1

∂(xβ)

∂x2. (2)

The overall marginal effect of a variable is

∂�

∂(x)= �′(xβ)

∂(xβ)

∂(x). (3)

All marginal effects are evaluated at mean values of covariates.Table 3 reports the marginal probability of a non-exporter entering export

markets and the probability of an exporter exiting export markets (the probitcoefficients are given in Table A1). There are four significant findings.First, plants that are more productive, larger, and older are more likely toenter export markets, and they are less likely to exit export markets. Thisconfirms the preliminary findings reported in Section IV.

Second, tariff reductions, on average, do not have much of an impact onthe export dynamics among continuing plants. The overall effects of tariffcuts on entry to and exit from export markets are statistically insignificant.Baldwin and Yan (2011) have separately examined the impact of tariffreductions on plant closure for exporters, and they have found that tariffreductions do have an effect here. Taken together, these results suggest thattariff cuts affect the survival of exporters, by causing the closure of plantsrather than by simply switching production from the export market to thedomestic market.

Third, a real appreciation of the Canadian dollar, on average, makesexport markets less attractive (i.e., non-exporters become less likely tostart exporting, and exporters become more likely to stop exporting andto serve only the domestic market). This complements the findings ofBaldwin and Yan (2011), who have reported that an appreciation of the



Table 3. Probability of entering and exiting export markets (marginal impact)

Entry Exit

Coefficient Std Err. Coefficient Std Err.

Overall marginal impactAverage tariff changes −0.007 (0.007) −0.004 (0.023)US/Canada real-exchange-rate changes −0.010∗∗ (0.002) 0.012∗∗ (0.004)Relative labour productivity 0.031∗∗ (0.003) −0.041∗∗ (0.008)Relative employment 0.043∗∗ (0.002) −0.106∗∗ (0.012)Age 0.002∗∗ (0.000) −0.004∗∗ (0.001)Foreign control 0.004 (0.007) −0.071∗∗ (0.012)Industry real-gross-shipment growth 0.000 (0.000) −0.001 (0.001)

Marginal effect of interaction termsAverage tariff changes∗ relative labour productivity −0.024∗∗ (0.009) 0.034∗∗ (0.017)Average tariff changes∗ relative employment −0.022∗∗ (0.005) 0.068∗∗ (0.033)Real-exchange-rate changes∗ relative labour productivity 0.003∗ (0.002) 0.004 (0.004)Real-exchange-rate changes∗ relative employment −0.007∗∗ (0.001) 0.010∗∗ (0.005)

Notes: Marginal impacts are calculated according to equations (2) and (3), evaluated at mean values of covariates.Standard errors are reported in parentheses. ** and * denote significance at the 5 and 10 percent levels,respectively.

Canadian dollar leads to plant shut-down among exporters.10 Therefore,currency appreciation reduces the survival rate of exporters, by increasingthe probability that they will switch from serving export markets to serv-ing only domestic markets and by increasing the plant death rate amongexporters.

Fourth, the impact of tariff reductions and the changes in real exchangerates vary across plants. Whether a particular plant changes its export status,following changes in tariff cuts and currency depreciation, depends on theefficiency level of the plant. The more efficient non-exporters, measuredeither by size or by labour productivity, are more likely to start exportingthan the less efficient non-exporters (significant negative interaction terms),while the more efficient exporters are less likely to stop exporting thanthe less efficient exporters (significant positive interaction terms). Baldwinand Yan (2011) have found similar patterns for plant death (i.e., the moreefficient plants have a better chance of survival in the face of tariff cuts orcurrency appreciation).

10 Baggs et al. (2009) have also found that survival and sales are negatively associated withappreciations in the Canadian dollar. Baldwin and Yan (2011) have further looked at howproductivity growth is affected by changes in tariffs and exchange rates, and at how survivalreactions and performances differ across groups of plants (e.g., exporters as opposed tonon-exporters, and domestic-controlled as opposed to foreign-controlled plants).



Table 4. Impact of tariffs and real exchange rates on labour productivitygrowth

Entrants vs. continuingnon-exporters

Exiters vs. continuingexporters


Average tariff changes 0.26 (0.269) 0.09 (0.448)US/Canada real-exchange-rate changes −0.69∗∗ (0.068) −0.77∗∗ (0.122)Dummy 4.14∗∗ (0.356) −5.70∗∗ (0.370)Dummy∗ average tariff changes 0.38 (0.398) −0.29 (0.554)Dummy∗ real-exchange-rate changes −0.69∗∗ (0.074) 0.54∗∗ (0.122)Relative labour productivity −4.36∗∗ (1.408) −5.37∗∗ (0.821)Relative employment 0.38∗∗ (0.094) 0.45∗∗ (0.081)Age −0.06∗∗ (0.014) 0.00 (0.025)Foreign-control 2.81∗∗ (0.969) 1.20∗∗ (0.284)Industry real-gross-shipment growth 0.05∗∗ (0.009) 0.02 (0.014)Number of observations 34,243 10,030R2 0.20 0.24

Notes: “Dummy” is a dummy variable that takes the value of one if a plant enters export markets during theperiod and zero if it remains a non-exporter. Similarly, Dummy equals one if a plant exits export markets duringthe period and zero if it remains an exporter. Standard errors are reported in parentheses. ∗∗ and ∗ denotesignificance at the 5 and 10 percent levels, respectively.

Impact on Within-Plant Productivity Growth

In order to examine the impact of changes in the trading environmenton productivity growth across plants that differ in terms of their exporttransition status, we model the plant-level productivity growth as a functionof tariff changes �τ it, changes in the real exchange rate �eit, a dummyvariable indicating export transition status Dpt, and their interactions. Themodel also controls for industry-level real gross output growth �lnQit, andplant-specific characteristics Zpt0.

�ln(LPpt ) = αi +αt + β1�τi t +β2�eit +β3 Dpt + β4 D∗pt�eit

+β5 D∗pt�τi t +β6�lnQit + γ Z pt0 (4)

Here, �ln(LPpt) is the average annual log growth of labour productiv-ity for plant p during period t. All other variables are defined as inequation (1).

Four conclusions are noteworthy (see Table 4). First, plants that arelarger and foreign-controlled have faster productivity growth. Also, thoseplants that started with a higher initial level of productivity have slowerproductivity growth, indicating reversion to the mean. These results arestatistically significant at the 5 percent level.

Second, if there had been no changes in tariffs and real exchange rates,plants that entered export markets would have had 4.1 percentage points



faster productivity growth, on average, than non-exporters (significant pos-itive coefficient on the dummy variable for entrants) and plants that exitedexport markets would have had 5.7 percentage points slower productivitygrowth, on average, than continuing exporters (significant negative coeffi-cient on the dummy variable for exiters).

Third, tariff reductions have little impact on the average productivity per-formance of plants, whether they are export participants or non-participants.This is in contrast to Trefler (2004) and Lileeva (2008). They use the sameplant-level dataset and industry tariff rates, but they use a longer periodbetween 1980 and 1996, and they report that US tariff cuts are associatedwith plant-level productivity gains. The difference in the two results stemsfrom the use of longer period, and it is caused by two factors. First, thelonger period has a larger cumulative decline in tariff rates, and therefore itallows for measurement of the tariff effect across a wider range of valuesof the independent variable. Secondly, the use of the longer period reducesthe observation set to a sample of larger and more successful plants thatsurvive for more than 15 years. The sample set over the 15-year period isonly 10,000 plants, compared to an average of 20,000 per five-year periodconsidered here. When the sample over the longer period is adopted here, adecline in tariffs significantly increases productivity growth. Also, changesin the real exchange rate continue to be significant. More importantly,an interaction term between plant size and tariff cuts indicates that largerplants benefit to a greater extent than small plants. As Lileeva and Trefler(2007) have shown, Canadian plants that gain from tariff cuts are those whoengage in innovation. It is the large plants that tend to be more innova-tive. Large plants are associated with greater financial, informational, andtechnology-absorptive capabilities. Exporting leads to changes in the typeof innovation activity in which large firms have a comparative advantage.Also, exporting is associated with a greater use of advanced manufacturingtechnologies, with increases in investments in R&D and in staff trainingand innovation activities. All of these features are more intensive in largerfirms than in smaller firms (Schumpeter, 1942; Baldwin and Hanel, 2003;Baldwin and Gu, 2004).

Fourth, fluctuations in real exchange rates have a significant impact onthe relative productivity growth of export participants and non-participants.A real appreciation of the Canadian dollar decreases productivity growthfor both non-exporters and entrants, but significantly more so for the latter.On average, a 1 percentage point appreciation in the real exchange ratedecreases productivity growth of non-exporters by 0.7 percentage points,compared to 1.4 percentage points for plants that entered export markets.These estimates from the 1984–1996 period serve to explain the narrowingof the productivity growth gap in Canada between non-exporters and en-trants post-2000. An appreciation of 5.9 percentage points annually would



reduce the superior performance of entrants over non-exporters to zero.The dramatic increase in the real value of the Canadian dollar duringthe 2000–2006 period (an average annual rate of 5.5 percentage points)meant that the difference in productivity growth between entrants andnon-exporters became smaller and statistically insignificant during this pe-riod (Table 2).

Similarly, a real appreciation of the Canadian dollar decreases produc-tivity growth for both exiters and continuing exporters, but significantlymore for the latter. On average, a 1 percentage point appreciation in thereal exchange rate decreases the productivity growth of continuing ex-porters by 0.8 percentage points, compared to only 0.2 percentage pointsfor plants that exited export markets. When the Canadian dollar appreci-ates against the US dollar, the productivity growth gap between continu-ing exporters and exiters is diminished. In the post-2000 period, the realexchange rate appreciated by 5.5 percentage points; this was enough toclose the gap between continuing exporters and exiters by 3.3 percentagepoints.

The negative correlation between currency appreciation and productivitygrowth, particularly among exporters, could be the result of several factors,such as the scale effect, the effect of a decline in capacity utilization, ora measurement issue that arises from the use of an average industry pricerather than a plant-specific price for deflating output to derive estimates ofproductivity growth.11

To evaluate whether these impacts have economic significance, a coun-terfactual experiment is conducted (see Table 5). First, changes in tariffsand real exchange rates are set to zero. Under this scenario, the resultsfrom Table 4 indicate that entrants would have enjoyed an advantage of4.1 percentage points over non-exporters in terms of average annual labourproductivity growth, while exiters would have lagged behind continuing

11 As one of the referees has pointed out, the measured productivity growth (real value-added per worker) will consequently consist of the actual productivity growth as well aschanges in the mark-ups (Bernard et al., 2003). If, in the face of an appreciation of thehome currency, exporters cut their prices in terms of the domestic currency to keep theirexports competitively priced in foreign markets, then their profit margin − and consequentlytheir measured productivity − might decrease relative to the industry’s average price. It isquite likely that exporters might react to currency appreciations by cutting domestic prices.Schembri (1989) and Baldwin and Yan (2007) have found that Canadian exporters adjusttheir output prices in Canadian dollars to fluctuations in the exchange rate, so that the exportselling prices in the foreign currency (US dollars) remain relatively stable. In pricing theirproducts to the foreign market, Canadian exporters absorb at least some of the fluctuationsin the exchange rate through their profits. However, as Baldwin and Yan (2007) have shown,this is also the time that domestic margins fall as a result of the foreign competition thatarises from exchange-rate appreciation. Therefore, the industry’s average price might divergevery little from the export price in periods of appreciation.



Table 5. Contribution to productivity growth gaps: a counter-factual exercise(in percentage points)

1984–1990 1990–1996 2000–2006

Entrants less continuing non-exportersActual average gaps 2.0 5.3 0.1Predicted gaps 3.0 5.4 0.3Gaps if there had been no changes in

tariffs and real exchange rates4.1 4.1 4.1

Gaps resulting from tariff changes 0.0 0.0 0.0Gaps resulting from real-exchange-rate

changes−1.1 1.3 −3.8

Exiters less continuing exportersActual average gaps −3.1 −6.9 −0.2Predicted gaps −4.8 −6.7 −2.7Gaps if there had been no changes in

tariffs and real exchange rates−5.7 −5.7 −5.7

Gaps resulting from tariff changes 0.0 0.0 0.0Gaps resulting from real-exchange-rate

changes0.9 −1.0 3.0

Source: Calculation by the authors.Notes: Actual average gaps are from Table 2. Gaps when there had been no changes in tariffs and real exchangerates are from Table 4. Gaps resulting from tariff changes and real-exchange-rate changes are calculated usingmarginal impacts from Table 4 and actual changes of the variables from Table 1. Gaps resulting from tariffchanges are set to zero, because the marginal impacts of tariffs are not statistically different from zero.

exporters by 5.7 percentage points. The productivity growth gap under thisscenario is caused by factors other than changes in tariffs and real exchangerates, resulting from either inherent differences between export participantsand non-participants or a possible learning-by-exporting effect. Second, thepredicted gaps induced by changes in tariffs and real exchange rates arecalculated. The predicted gaps are estimated using the marginal impactsreported in Table 4 and the actual changes in tariffs and real exchangerates from Table 1. Because the marginal impacts of tariffs are not sta-tistically different from zero, the predicted gaps because of tariff cuts areassumed to be zero. Third, the predicted gaps are compared to the actualgrowth gaps, which include both the in-sample comparison (1984–1990and 1990–1996 periods) and out-of-sample comparison (2000–2006period).

This counterfactual experiment shows that the fluctuations in real ex-change rates explain almost all the shifts in productivity growth gaps be-tween export participants and non-participants over the three decades. In thecase of export entrants and non-exporters, the depreciation in the real ex-change rate increases the relative advantage of entrants by 1.3 percentagepoints during the 1990–1996 period. However, the superior productivityperformance of entrants is offset partially during the 1984–1990 periodand almost entirely during the post-2000 period, when the Canadian dollar



appreciated. In the case of exiters and continuing exporters, a depreciationof the real exchange rate increases the growth gap by 1 percentage pointduring the 1990–1996 period, but the appreciation during the 1984–1990and 2000–2006 periods closed the gap by 0.9 and 3.0 percentage points,respectively.

VI. Self-Selection or Learning-by-Exporting Effects?

While the difference in the productivity growth of export participants andnon-participants varies systematically across periods, it is nevertheless pos-itive after allowance is made for changes in tariffs and real exchangerates (i.e., plants that enter export markets do better than those that donot, and plants that survive in export markets do better than those thatfail).

There are at least two explanations about why exporting has a positivecorrelation with productivity growth. The first is the self-selection hypoth-esis: larger, more productive, and more innovative plants self-select intoexport markets. These plants are more likely to be successful and to havehigher productivity growth in general, both before and after entry. Thesecond is the learning-by-exporting hypothesis. Exporting might improveproductivity, because expansion to foreign markets offers opportunities toexpand plant size and to learn how to exploit scale economies, and oppor-tunities to learn about new technologies and products and to become moreinnovative. Intense international competitive pressure also forces plants toimprove efficiency. In this case, productivity performance increases becauseof various learning effects.

In Section V, we have shown that exporters enjoy higher productivitygrowth, even after accounting for plant characteristics such as size andproductivity. However, the regression analysis used for this purpose mightsuffer from a selection-bias problem. The binary variable that accounts forthe differences between the two samples is essentially calculated as theeffect at the mean of the population − both exporters and non-exporters.Calculating the average of the treated group (exporters) to the average ofthe entire population might yield biased estimates of the treatment effectif the treated group is selected in a non-random way. This can introduce aselection bias into the estimates of the effect of exporting.

In this section, we make use of the propensity-score matching approachto choose a sample for the control group in order to reduce the potentialeffects of selection bias. This approach has recently been applied to theanalysis of exporting and firm performance (Wagner, 2002; Girma et al.,2004; De Loecker, 2007) to test for a causal relationship between exportparticipation and productivity.



Methodology

In order to evaluate the impact of becoming an exporter, an estimate isrequired of the difference between the productivity growth of plants thatchanged their export status (entered or exited export markets) and theoutcome for the same plants had they not changed their status. However,the latter outcome is an unobserved counterfactual.

Propensity-score matching provides a way of constructing the counterfac-tual, by estimating the difference between plants that undergo a treatment(entry or exit) and those that do not experience treatment. From a poolof continuing non-exporters or continuing exporters, the technique selectsplants that share similar characteristics as plants that changed export status,and then it calculates the productivity growth difference between plants thatchanged status (treated plants) and the selected group of control observa-tions (control or untreated plants). If the matching process is successful, acausal interpretation can be given to the average difference in productivitygrowth between treated and control groups.

The control group is created on the basis of observable plant charac-teristics, such as size, labour productivity, age, and ownership status, aswell as characteristics that potentially influence the outcome of interest inthe treated group, such as industry-wide changes in tariffs, real exchangerates, and output and industry-specific effects. Technically, this is done bymatching treated plants to control plants with the same or a very similarpropensity score in order to identify a set of similar plants in the controlgroup to those who received the treatment, defined here as entry to or exitfrom export markets. The propensity score is the predicted probability ofentering or exiting export markets. It collapses the set of characteristicsthat determine whether a plant entered or exited export markets to a singlecomposite number that is used to identify plants in the control group thatare similar in all respects to those treated, except that they did not receivethe treatment (i.e., in the case of this investigation, entry to or exit fromexport markets).

Propensity-score matching controls for selection bias by restricting thecomparison to differences between treated and control plants with similarobservable characteristics. However, this method is still vulnerable toproblems of non-random selection bias because of potential unobservablecharacteristics in the treated group. To address this, a differences-in-differences method is used, which controls for time-invariant unobservablecharacteristics.

The combination of the matching and differences-in-differences ap-proaches allows us to ask whether there is a divergence in the paths ofproductivity growth between plants that changed export status and the



matched control plants that have similar observable and unobservable (butconstant) attributes.

Results

In order to avoid conflating the effects of export market entry and exit,plants that have changed export status in some earlier periods are ex-cluded from the analysis. Exporter starters and non-exporters are defined asfollows: plants that did not export during the 1984–1990 period, but didstart exporting during the 1990–1996 period are classified as export starters;non-exporters are those that did not export during either the 1984–1990 or1990–1996 periods. Similarly, exporter stoppers and continuing exportersare defined as follows: plants that were exporters during the 1984–1990period, but stopped exporting during the 1990–1996 period are classified asexport stoppers; continuing exporters exported during both the 1984–1990and 1990–1996 periods.

Table A2 presents the probit results that are used in the propensity-scoreapproach to determine the probability of entry and exit from export marketsduring the 1990–1996 period.12 The probability of entering and exiting ex-port markets during the 1990–1996 period depends on plant characteristicsat the beginning of the period and changes in tariffs, exchange rates, andindustry real-shipment growth during the period. These equations are thenused to determine a score to be used to choose a set of matching plants inthe control group.

The difference in each predictor used in the probit model between thetreated and the control groups was examined in order to assess how wellthe propensity-score matching performed. Differences are expected beforematching but not after matching if the covariates are balanced. If the testsfor any predictor showed that there were significant differences betweentreated and control units, the probit model was modified by adding higher-order terms of the covariate. Table 6 shows that all the differences aftermatching are small and are statistically insignificant. Of the population of7,539 non-exporters, 1,410 are found to be good matches for the 1,410export starters. Similarly, of the population of 1,853 continuing exporters,402 are found to be good matches for the 403 export stoppers. Thus,about one in five non-exporters (or continuing exporters) is deemed todisplay observable characteristics similar to the characteristics of thosewho subsequently entered (or exited) the export market.

12 One-to-one nearest-neighbour matching techniques without replacement and with commonsupport are used. That is, there are both treated and non-treated plants for each characteristicthat is compared. If the common support is not satisfied in the treatment group, then theseplants are dropped.



Table 6. Balancing test: comparisons of means for unmatched and matchedsamples

Export starters vs.non-exporters

Export stoppers vs. continuingexporters

P-value on P-value onExport Non- t-test of Export Continuing t-test of

Means starters exporters differences stoppers exporters differences

Unmatched samplesSample size 1,410 7,539 403 1,853Relative labour

productivity1.33 1.03 0.00 0.99 1.03 0.33

Relativeemployment

2.57 0.93 0.00 0.76 1.53 0.00

(Relativeemployment)square

15.68 5.56 0.15 1.61 4.77 0.00

Age 14.05 12.90 0.00 15.43 15.39 0.81Foreign-controlled 0.13 0.05 0.00 0.26 0.37 0.00Canadian tariff

changes−0.81 −0.74 0.00 −0.57 −0.53 0.24

US tariff changes −0.34 −0.29 0.00 −0.29 −0.26 0.04Real-exchange-rate

changes−1.97 −1.92 0.16 −1.43 −1.45 0.79

Real industryshipmentchanges

0.88 0.63 0.19 3.02 2.79 0.58

Matched samplesSample size 1,410 1,410 402 402Relative labour

productivity1.33 1.29 0.29 0.99 0.99 0.92

Relativeemployment

2.57 2.33 0.10 0.76 0.88 0.15

(Relativeemployment)square

15.68 26.90 0.49 1.62 2.42 0.43

Age 14.05 14.22 0.23 15.42 15.38 0.84Foreign-controlled 0.13 0.12 0.43 0.26 0.26 0.81Canadian tariff

changes−0.81 −0.81 0.93 −0.56 −0.58 0.66

US tariff changes −0.34 −0.35 0.51 −0.29 −0.30 0.51Real-exchange-rate

changes−1.97 −2.00 0.41 −1.43 −1.40 0.84

Real industryshipmentchanges

0.88 0.84 0.90 3.02 2.82 0.74

Source: Calculation by the authors.

The primary results of interest are the average differences in labourproductivity growth in the matched samples, net of the average initialdifferences before changes in export status (Column 3 of Tables 7 and8. The results reveal a causal effect of export participation on produc-tivity growth. Productivity in plants that entered export markets grew by



Table 7. Comparison of export starters and non-exporters: unmatched andmatched samples (in percentage points)

Labour productivity growth

Pre-entry period Post-entry period Difference: post-entry(1984−1990) (1990−1996) less pre-entry periods

Means (1) (2) (3)

Unmatched samplePlants that changed export

status (export starters,n = 1,410)

−0.9 2.3 3.2

Plants that did not changeexport status(non-exporters, n = 7,539)

−1.4 −2.4 −1.0

Difference: exporter startersless non-exporters

0.5 4.7∗ 4.2∗

Matched samplePlants that changed export

status (export starters,n = 1,410)

−0.9 2.3 3.2

Plants that did not changeexport status(non-exporters, n = 1,410)

−0.8 −1.6 −0.8

Difference: exporter startersless non-exporters

−0.1 3.9∗ 4.0∗

Source: Calculation by the authors.Notes: ∗ denotes significance at the 5 percent level or better.

3.2 percent, while similar plants that remained in the domestic market grewby only −0.8 percent (Table 7). Thus, plants that started exporting enjoyeda productivity growth advantage over the control group of 4 percentagepoints. Plants that exited export markets experienced slower productivitygrowth than plants) that had an equivalent probability of exiting exportmarkets but did not. Notably, productivity growth was much slower in theperiod when the exit occurred. Plants that exited export markets experienceda loss in productivity of 6.8 percent, while similar plants that remained inexport markets had a 0.3 percent growth. This resulted in a disadvantageof 7.1 percentage points for plants that stopped exporting, compared to thecontrol group (Table 8). The differences are all statistically significant at a5 percent level.

VII. Conclusion

Productivity growth in a globalized economy is affected by the nature ofthe reaction of different producers to events that shape the world tradingsystem, which include changes in tariffs associated with trade liberalization



Table 8. Comparison of export stoppers and continuing exporters: unmatchedand matched samples (in percentage points)

Labour productivity growth

Pre-exit period Post-exit period Difference: post-exit(1984−1990) (1990−1996) less pre-exit periods

Means (1) (2) (3)

Unmatched samplePlants that changed export

status (export stoppers,n = 403)

1.1 −5.7 −6.8

Plants that did not changeexport status(continuing exporters, n= 1,853)

1.2 2.1 0.9

Difference: exportstoppers less continuingexporters

−0.1 −7.8∗ −7.7∗

Matched samplePlants that changed export

status (export stoppers,n = 403)

1.1 −5.7 −6.8

Plants that did not changeexport status(continuing exporters, n= 402)

1.0 1.3 0.3

Difference: exportstoppers less continuingexporters

0.1 −7.0∗ −7.1∗

Source: Calculation by the authors.Note: ∗ denotes significance at the 5 percent level or better.

and movements in exchange rates. In this paper, we have looked at howentry into and exit from export markets affect productivity growth, andhow entry and exit are affected by changes in the trading environment ascharacterized by changing tariff rates and real exchange rates. We haveexamined the experience of Canadian manufacturing firms over three sep-arate periods, which feature different combinations of changing tariff ratesand trends in the real exchange rate.

We have confirmed previous findings. Plants self-select into export mar-kets (i.e., more efficient plants are more likely to enter and to survive inexport markets). However, the trading environment is found to affect thedegree of experimentation. Tariff reductions and currency depreciation in-crease the probability that more efficient non-exporters will start to export,and they decrease the probability that more efficient exporters will stopexporting.



We also find that exporting is associated with improvements in therelative productivity growth of exporters. This finding is robust to theestimation technique used. The first technique is an OLS multivariate anal-ysis, and the second is a combination of a propensity-score matching tech-nique and the difference-in-differences method. Both techniques take intoaccount plant characteristics and industry-specific changes in tariffs andreal exchange rates, and both find that plants that enter export mar-kets have higher productivity growth (by about 4 percentage points inboth cases) than those not doing so. Similarly, plants that exit ex-port markets have slower growth than similar plants that remain in ex-port markets (a difference of 5.7 percentage points in the multivari-ate analysis and 7.1 percentage points in the propensity-score matchinganalysis).

Baldwin and Gu (2004) have examined the mechanisms behind thelink between participation in export markets and increases in productiv-ity growth. Besides the effect of increasing plant specialization and theexposure to more intense international competition, they have also foundevidence in support of learning-by-exporting. First, exporters learn frominternational best practices by an increase in the use of foreign technologyat plants, by an increase in the incidence of R&D collaboration agreementswith foreign buyers, and by improving the flow of information about for-eign technologies to Canadian plants. Second, plants that move into exportmarkets increase investments in R&D and training to develop capacities forabsorbing foreign best practices. Third, exporting leads to changes in thenature of innovation (i.e., a greater use of advanced technologies, whichthereby increases the intensity of process innovation).

While exporting is associated with faster productivity growth, the mag-nitude of the gains is affected by what is happening to exchange rates atthe time; this is reinforced when the home currency depreciates, and itis attenuated when the home currency appreciates. In this paper, we havedemonstrated that the superior performance of Canadian export startersor continuing exporters was reinforced in the 1990–1996 period, whenthe Canadian dollar depreciated. However, the advantage was reduced in1984–1990 and 2000–2006 when the Canadian dollar appreciated. In par-ticular, the dramatic increase in the value of the Canadian dollar duringthe post-2000 period almost completely offset the advantages enjoyed byexport-market participants. This could be a result of the effects of loss ofscale economies or declines in capacity utilization.

These findings can explain why in the literature it has not been consis-tently found that exporting is associated with productivity gains. Differencesin the trading environment − as measured by changing tariffs and exchangerates − reinforce or attenuate the gains that, in normal circumstances, areenjoyed by exporters.



Appendix

Table A1. Probit coefficient estimates of entering and exiting export markets

Entry Exit


Average tariff changes 0.154∗∗ (0.056) −0.317∗∗ (0.113)US/Canada real-exchange-rate changes −0.043∗∗ (0.011) −0.005 (0.020)Average tariff changes × relative

labour productivity−0.095∗∗ (0.040) 0.102∗∗ (0.051)

Average tariff changes × relativeemployment

−0.087∗∗ (0.022) 0.205∗∗ (0.095)

Average real-exchange-rate changes ×relative labour productivity

0.019∗∗ (0.007) 0.012 (0.012)

Average real-exchange-rate changes ×relative employment

−0.019∗∗ (0.005) 0.028∗∗ (0.015)

Relative labour productivity 0.094∗∗ (0.026) −0.077∗∗ (0.029)Relative employment 0.143∗∗ (0.015) −0.223∗∗ (0.061)Age 0.009∗∗ (0.002) −0.013∗∗ (0.003)Foreign-control 0.017 (0.031) −0.219∗∗ (0.038)Industry real-gross-shipment growth 0.000 (0.001) −0.003 (0.002)Number of observations 36,683 10,137Log pseudo-likelihood −14,974 −5,428

Notes: Standard errors are reported in parentheses. ∗∗ and ∗ denote significance at the 5 and 10 percent levels,respectively.

Table A2. Propensity-score matching: probit results

Dependent variable

Entry = 1 Exit = 1

Relative labour productivity 0.15∗∗ (0.019) −0.14∗∗ (0.060)Relative employment 0.39∗∗ (0.015) −0.70∗∗ (0.059)Relative employment2 −0.01∗∗ (0.001) 0.04∗∗ (0.005)Age 0.01∗∗ (0.005) 0.03∗∗ (0.011)Foreign control 0.03 (0.071) −0.18∗∗ (0.083)Canadian tariff changes 0.13 (0.083) −0.34∗ (0.209)US tariff changes −0.18 (0.142) 0.22 (0.321)Real-exchange-rate changes −0.01 (0.033) 0.01 (0.031)Real industry shipment changes 0.00 (0.003) 0.01 (0.005)No. of observations 8,949 2,256Log likelihood −3200.7 −872.5Pseudo-R square 0.18 0.18

Source: Calculation by the authors.Notes: Standard errors are reported in parentheses. ∗∗ and ∗ denote significance at the 5 and 10 percent levels,respectively.



ReferencesAi, C. and Norton, E. C. (2003), Interaction Terms in Logit and Probit Models, Economics

Letters 80, 123−129.Arrow, K. (1962), The Economic Implications of Learning-by-Doing, Review of Economics

Studies 29, 155−173.Baggs, J. (2005), Firm Survival and Exit in Response to Trade Liberalization, Canadian

Journal of Economics 38, 1364−1383.Baggs, J., Beaulieu, E., and Fung, L. (2009), Firm Survival, Performance, and the Exchange

Rate, Canadian Journal of Economics 42, 393−421.Baldwin, J. R. and Gu, W. (2003), Export Market Performance and Productivity Performance

in Canadian Manufacturing, Canadian Journal of Economics 36, 634−57.Baldwin, J. R. and Gu, W. (2004), Trade Liberalization: Export-Market Participation, Pro-

ductivity Growth, and Innovation, Oxford Review of Economic Policy 20, 372−392.Baldwin, J. R. and Gu, W. (2009), The Impact of Trade on Plant Scale, Production-Run

Length and Diversification, in T. Dunne, J. B. Jensen, and M. J. Roberts (eds.), ProducerDynamics: New Evidence from Micro Data, University of Chicago Press, Chicago.

Baldwin, J. R. and Hanel, P. (2003), Innovation and Knowledge Creation in an OpenEconomy, Canadian Industry and International Implications, Cambridge University Press,Cambridge.

Baldwin, J. R. and Yan, B. (2007), Exchange Rate Cycles and Canada/US ManufacturingPrices, Review of World Economics (Weltwirtschaftliches Archiv) 143, 508−533.

Baldwin, J. R. and Yan, B. (2011), The Death of Canadian Manufacturing Plants: Hetero-geneous Responses to Changes in Tariffs and Real Exchange Rates, Review of WorldEconomics (Weltwirtschaftliches Archiv) 147, 131−167.

Baldwin, J. R., Jarmin, R. S., and Tang, J. (2004), Small North American Producers GiveGround in the 1990s, Small Business Economics 23, 349−61.

Bernard, A. B. and Wagner, J. (1997), Exports and Success in German Manufacturing,Review of World Economics (Weltwirtschaftliches Archiv) 133, 134−157.

Bernard, A. B., Eaton, J., Jensen, J. B., and Kortum, S. (2003), Plants and Productivity inInternational Trade, American Economic Review 93 (4), 1268−1290.

Berndt, E. R. and Fuss, M. A. (1986), Productivity Measurement Using Capital Asset Valu-ation to Adjust for Variations in Utilization, Journal of Econometrics 33, 7−29.

Chen, Y-C., Rogoff, K., and Rossi, B. (2010), Can Exchange Rate Forecast CommodityPrices?, Quarterly Journal of Economics 125, 1145−1194.

Clerides, S. K., Lach, S., and Tybout, J. (1998), Is Learning by Exporting Important? Micro-Dynamic Evidence from Columbia, Mexico and Morocco, Quarterly Journal of Economics113, 903−947.

Cross, P. (2002), Cyclical Implications of the Rising Import Content of Exports, CanadianEconomic Observer, Catalogue no. 11-010-XPB.

Das, M., Roberts, M., and Tybout, J. (2007), Market Entry Costs, Producer Heterogeneityand Export Dynamics, Econometrica 75, 837−873.

De Loecker, J. (2007), Do Exports Generate Higher Productivity? Evidence from Slovenia,Journal of International Economics 73, 69−98.

Eaton, J., Eslava, M., Krizan, C. J., Kugler, M., and Tybout, J. (2008), A Search and LearningModel of Export Dynamics, http://www.econ.gatech.edu/files/seminars/Eaton_F2009.pdf.

Esteve Perez, S., Sanchis Llopis, A., and Sanchis Llopis, J. A. (2004), The Determi-nants of Survival of Spanish Manufacturing Firms, Review of Industrial Organization 25,251−273.

Foster, L., Haltiwanger, J., and Krizan, C. J. (2001), Aggregate Productivity Growth: Lessonsfrom Microeconomic Evidence, in C. R. Hulten, E. R. Dean, and M. J. Harper (eds.), NewDevelopments in Productivity Analysis, University of Chicago Press, Chicago.



Girma, S., Greenaway, D., and Kneller, R. (2003), Export Market Exit and PerformanceDynamics: A Causality Analysis of Matched Firms, Economics Letters 80, 181−187.

Girma, S., Greenaway, D., and Kneller, R. (2004), Does Exporting Increase Productivity?A Microeconometric Analysis of Matched Firms, Review of International Economics 112,855−866.

Greenaway, D. and Kneller, R. (2007), Firm Heterogeneity, Exporting and Foreign DirectInvestment, Economic Journal 117 (517), F134−F161.

Kaldor, N. (1966), Causes of the Slow Rate of Economic Growth of the United Kingdom,Cambridge University Press, Cambridge. Reprinted in Kaldor, N. (1978), Further Essayson Economic Theory, Duckworth, London.

Kimura, F. and Fujii, T. (2003), Globalizing Activities and the Rate of Survival: Panel DataAnalysis on Japanese Firms, Journal of the Japanese and International Economies 17,538−560.

Lileeva, A. (2008), Trade Liberalization and Productivity Dynamics: Evidence from Canada,Canadian Journal of Economics 41, 360−390.

Lileeva, A. and Trefler, D. (2007), Improved Access to Foreign Markets Raises Plant-LevelProductivity . . . for Some Plants, NBER Working Paper no. 13297.

Lopez, R. A. (2005), Trade and Growth: Reconciling the Macroeconomic and MicroeconomicEvidence, Journal of Economic Surveys 19, 623−648.

Macdonald, R. (2008), The Terms of Trade and Domestic Spending, Canadian EconomicObserver, Catalogue no. 11-010, Statistics Canada, Ottawa.

Melitz, M. (2003), The Impact of Trade on Intra-Industry Reallocations and AggregateIndustry Productivity, Econometrica 71, 1695−1725.

Melitz, M. and Ottaviano, G. (2008), Market Size, Trade and Productivity, Review of Eco-nomic Studies 75, 295−316.

Norton, E. C., Wang, H., and Ai, C. (2004), Computing Interaction Effects and StandardErrors in Logit and Probit Models, Stata Journal 4, 154−167.

Ruhl, K. J. and Willis, J. L. (2008), New Exporter Dynamics, manuscript, NYU Stern Schoolof Business, http://www.econ.yale.edu/seminars/IntTrade/it09/ruhl-090422.pdf

Schembri, L. (1989), Exchange Rates and Export Prices: An Industry Approach, in R.Feenstra (ed.), Trade Policy and International Competitiveness, University of ChicagoPress, Chicago, 403−468.

Schumpeter, J. (1942), Capitalism, Socialism and Democracy, Harper and Row, New York.Scott, M. C. (1989), A New View of Economic Growth, Clarendon Press, Oxford.Serti, F. and Tomasi, C. (2008), Self-Selection and Post-Entry Effects of Exports: Evidence

from Italian Manufacturing Firms, Review of World Economics 144, 660−694.Trefler, D. (2004), The Long and Short of the Canada−US Free Trade Agreement, American

Economic Review 94(4), 870−895.Van Biesebroeck, J. (2005), Exporting Raises Productivity in Sub-Saharan African Manufac-

turing Firms, Journal of International Economics 67, 373−391.Wagner, J. (2002), The Causal Effect of Exports on Firm Size and Labour Productivity: First

Evidence from a Matching Approach, Economics Letters 77, 287−292.Wagner, J. (2007), Exports and Productivity: A Survey of the Evidence from Firm-Level

Data, The World Economy 60, 60−82.

First version submitted October 2009;final version received January 2011.


export market dynamics and plant-level productivity: impact of tariff reductions and exchange-rate...

Documents