International Trade, Firms, and the Labor Market (Summary)

Elhanan HELPMAN (Galen L. Stone Professor of International Trade, Harvard University)

Today I will discuss trade and labor market outcomes, a topic on which I have worked for the last few years in various combinations with Oleg Itskhoki and Stephen Redding.

Traditional explanations of international trade have emphasized comparative advantage based on differences in technology and factor endowments. In the 1980s, factor proportions were merged with economies of scale and monopolistic competition to address a greater scope of issues. However, these models adopted the assumption that all firms in an industry are similar to each other, and that all firms engage in export.

More recently, firm heterogeneity has been introduced to international trade models to provide a natural explanation for why only a fraction of industrial firms export, and why exporting firms are bigger and more productive than non-exporting firms. One important example is Melitz, who explained that when trade costs fall, the less productive firms exit the industry, and market share is reallocated to the larger, more productive firms that expand to serve foreign markets. Others showed that firm heterogeneity and selection also influence cross-section patterns of trade and foreign direct investment, so that the ratio of exports to foreign subsidiary sales depends not only on the trade-off between proximity and concentration, but also on the dispersion of firm productivity. Contractual frictions were given as another important determinant of foreign direct investment.

Most of this literature assumes frictionless labor markets, but there are three prominent features of labor markets that are important to my discussion. The first is that there are substantial differences in workforce composition across firms in the same industry. The second is that there is considerable variation in the wages of workers with the same observed characteristics, and the third is that unemployment rate varies across industries.

Moreover, macro studies have found labor market institutions to be influential in shaping a country's responses to shocks, and that changes in labor market institutions need to be examined in order to understand the evolution of unemployment rates in each country. The magnitude of cross-country differences in labor market institutions can be seen in the rigidity of labor market institutions across countries—that is, the ease of hiring and firing workers and the rigidity of hours worked. In the European Union, member states have focused on labor market policies for many years and developed a strategy of conjoining their countries to make their labor markets more flexible and beneficial. To address this sort of issue, we need theoretical models that go beyond the conventional concepts of international trade and labor market features.

The main issues I present today build on a long line of research on trade and labor market frictions and focus on the question, "How do labor market frictions impact interdependence across countries?" In particular, I will discuss the impact of trade on inequality and unemployment and the impacts of one country's labor market frictions on its trade partners. I will also discuss how the removal of trade impediments impacts countries with different labor market frictions, and lastly present a new discussion on labor market policies.

Let me begin with inequality.

Recent research has identified a need to rethink the links between trade and wage inequality, as the traditional framework fails to provide adequate explanation of inequality trends observed around the globe today. In contrast to the traditional framework, our framework is based on the observation that substantial wage inequalities exist across firms and workers within the same sector. To examine the link between trade and inequality, let us consider a differentiated-product sector, where brands are produced by firms that are heterogeneous in productivity. These firms face fixed entry costs and production costs, as well as fixed and variable trade costs. They engage in monopolistic competition in the product market. At the same time, they are subject to search and matching frictions in the labor market, and engage in wage bargaining. In this context, it is important to assume that firms employ many workers and engage in multilateral bargaining.

Workers in this model are homogeneous ex ante . They appear the same, but draw a match-specific ability when matched with a firm in the differentiated sector. This ability, which is observed by neither the worker nor the firm, is drawn from an independent Pareto distribution. Firms too are homogeneous ex ante , but when they enter an industry they become heterogeneous, because they draw a productivity upon entry in the differentiated sector. The production function, or the output of each firm variety, depends on the productivity of the firm, the measure of workers hired, and the average ability of these workers. Another parameter introduces complementarities in worker ability, because a worker's ability affects the productivity of other workers in the firm. Firms also employ a screening process to find out whether a worker's ability is above or below a certain threshold. The cost of hiring is derived from a Cobb-Douglas matching function, where the cost is a function of a parameter, which is increasing in the cost of posting vacancies and decreasing in the efficiency of the matching process, and the tightness of the labor market. This equation is essentially used as a measure of labor market frictions.

In this model, workers and firms enter the differentiated sector, and firms discover their productivity. The firms choose to exit or produce, and if they choose to produce, they post vacancies. The vacancies are matched with workers, and firms screen their matched workers by setting a certain screening threshold. Only workers who pass the threshold are hired, and those with abilities below the threshold become unemployed. The firms and their hired workers engage in multilateral wage bargaining. Finally, output is produced and markets clear.

Now, the equilibrium structure of the industry looks similar to what Melitz described when he said that the least productive firms choose to exit and the most productive firms choose to export. In this structure, the more productive firms have larger revenues, post more vacancies, match with more workers, and screen to higher ability thresholds. As a result, they acquire a better workforce composition with higher average ability and pay higher wages. There is thus a positive correlation between wages and the size of a firm.

Wage is also a function of firm productivity, and wages are higher in larger and more productive firms. They are particularly high among exporting firms, because the characteristics of the more productive firms are systematically related to export participation. This means that inequality in wages within an industry stems from the fraction of firms that export. When comparing wage distribution in the trade equilibrium and in autarky, inequality is greater in the trade equilibrium than in autarky when only a fraction of firms export, but the inequality is the same when all firms export. Thus, this highlights a new mechanism for international trade to affect wage inequality: when only some firms export, the increase in wages that occurs at the productivity threshold for exporting raises wage inequality across firms.

The opening of trade also has implications for unemployment. Since the opening of trade reallocates employment within industries towards more productive exporting firms which screen more intensively and hire a smaller fraction of the workers with whom they are matched, the hiring rate becomes lower in the trade equilibrium than in autarky. This reduction in the hiring rate raises unemployment, because there would be more workers who are not matched with a firm, or are not retained by the firm with which they are matched because their match-specific ability draw is below the screening threshold of the firm. It should be noted, however, that while the opening of trade increases both wage inequality and unemployment, such an environment necessarily increases welfare, which is measured as the expected welfare of workers who seek job vacancy.

The second topic I wish to discuss is interdependence between trading countries. How do labor market frictions impact interdependence across countries? And what are the impacts of a country's labor market frictions on its trade partners?

To address the above questions, let us consider a two-country world, say with Countries A and B. Each country has two sectors, one that is differentiated and the other that is homogeneous, but Country A has higher labor market frictions in the differentiated sector. What are the implications?

The implication is that if these countries are allowed to trade with each other, a larger fraction of firms in Country B will export, and Country B will export differentiated products on net and import homogeneous goods. Since the only difference between the two countries is in their labor market frictions, it follows that this pattern of trade is determined by differences in labor market frictions across countries. Moreover, the share of intra-industry trade is smaller, the larger the gap in relative hiring costs.

Another implication is that both countries gain from trade. This is an interesting point that relates to interdependence in labor market frictions. A reduction in labor market frictions in the differentiated sector of a country raises the country's welfare by making the sector more competitive relative to its trade partner, and reduces the trade partner's welfare. However, a simultaneous proportional reduction in the labor market frictions of both countries raises welfare in both countries, because it expands the size of the differentiated sector in each. Unlike these welfare consequences of lower trade frictions, however, the effects on unemployment can differ across countries, because there is no obvious relationship between labor market frictions and unemployment rates or trade frictions and unemployment rates. It is possible for Country A with high frictions to have either a higher or lower unemployment rate compared to Country B, depending on whether the unemployment rate is originally higher or lower in the differentiated or homogeneous sector. In any case, knowledge of relative rates of unemployment across countries is not sufficient to draw inferences about their relative levels of labor market frictions.

Now, let us turn to policy implications and examine the impact of unemployment benefits.

Unemployment benefits impact the cost of hiring by increasing the outside option of workers at the wage bargaining stage. They also affect workers' decisions on whether to search for a job in the differentiated or homogeneous sector. When labor market frictions are higher in the homogeneous sector, unemployment benefits raise the cost of hiring on net. When labor market frictions are higher in the differentiated sector, the differentiated sector has a higher sectoral unemployment rate, so unemployment benefits reduce hiring costs in the differentiated sector and lead to its expansion, as more workers choose to search for jobs in this sector. In other words, unemployment benefits have an uneven effect on sectoral employment, favoring the sector with higher unemployment.

Under these circumstances, one country's unemployment benefits also affect its trade partners. That is, by raising unemployment benefits a country makes its differentiated sector more competitive in world markets if this sector has the higher sectoral rate of unemployment, in which case the unemployment benefits hurt the country's trade partners. Alternatively, by raising unemployment benefits a country benefits its trade partners when the country's labor market frictions are higher in the homogeneous sector.

Whether a country gains from raising unemployment benefits depends on structural features of the labor market. When alpha (α) is the relative weight of workers in the matching technology, lambda (λ) is the relative weight of firms in the bargaining game, and α x λ = 1, a condition known as the Hosios, unemployment benefits reduce welfare. When the product of α and λ is high, unemployment benefits initially raise and then reduce welfare. When it is low, unemployment benefits reduce welfare. It follows that when αλ is large, welfare is maximized at a positive level of unemployment benefits, and when αλ is small, the optimal level of unemployment benefits equals zero.

Let me move to my last point on optimal policies.

In the design of optimal policies, we need to consider achieving a constrained Pareto optimum, where workers are allocated to firms through the matching response. Only if the Hosios condition is satisfied, which requires the relative bargaining weight of employers to equal their relative weight in the matching technology, tightness in the labor market is optimal, and there is no need to intervene with labor market policies. If the Hosios condition is not satisfied, subsidies to posting vacancies or to the cost of hiring should be used to achieve the efficient level of labor market tightness. Unemployment benefits could also be used to achieve optimal labor market tightness under some circumstances, but more direct interventions are preferable on informational grounds. In fact, a combination of interventions is required to correct the numerous distortions in the labor and product markets and achieve the constrained efficient allocation.

Importantly, the same optimal policies in the differentiated sector apply to all firms. In other words, they equally apply to low- and high-productivity firms, and to exporters and nonexporters alike. This means that optimal policies do not discriminate between firms based on productivity, size, or export status.

In conclusion, I wish to summarize a few points.

Labor market frictions can be a source of comparative advantage, and while trade with labor market frictions can raise both unemployment and inequality, it is beneficial. The point here is that unemployment and inequality are not necessarily good indications of welfare.

Increased wage inequality due to unobserved worker heterogeneity may result from technological change that increases the dispersion of firm productivity or from declining costs of international trade.

Differences in unemployment rates across countries do not necessarily reflect differences in welfare and labor market frictions.

Unemployment benefits as an isolated policy instrument can be beneficial or detrimental to one's own country, as well as to its trade partners. In which way it plays out depends on various parameters.

There exists a set of policies from which optimal policies should be designed to achieve a constrained Pareto optimum.

Lastly, conventional macro models are not satisfactory for dealing with unemployment and interdependence across countries. They need to be generalized to form a multisectoral perspective of interdependence.

Richard E. BALDWIN (Professor of International Economics, The Graduate Institute, Geneva)

This paper I am giving today is a theory paper I prepared jointly with Anthony Venebles from Oxford. It examines the unbundling or relocating of the value chain, and what it means for policy in general, and for labor markets in particular.

Let me start by putting this value chain unbundling into the context of globalization. Actually, I invite you to think about globalization as two unbundlings affecting two types of cost. The first unbundling is associated primarily with transportation cost. The development of steam power shrunk the world and allowed factories to be a long way from consumers. In other words, the first unbundling was factories from consumers. Globalization took a different turn during the second unbundling. Until then, a factory's various functions were clustered for reasons that had more to do with the cost of coordinating complex activities than the cost of shipping goods, but the ICT revolution lowered the cost of coordinating complex activities at distance and unbundled the factories themselves.

It is important to think about globalization as being driven first by the falling cost of transportation, and second by the falling cost of coordination, although it should be noted that coordination costs are still very much present. Even with trade costs coming down and coordination costs coming down, we still see local clustering, which shows that in no way is distance dead. We therefore need to think about a progressive movement of intermediate trade and coordination costs.

Based on the above, let us look into the second unbundling, and think about which parts get unbundled first.

We shall work within a conceptual framework, which tries to think about what happens when production processes unbundle and get offshored or outsourced. However, to connect this to the labor market later, I shall talk a little bit about a paper I wrote with Frederic Robert-Nicoud, which integrates the Grossman and Rossi-Hansberg trade-in-tasks into the Heckscher-Ohlin framework so we can see how offshoring of various things leads to labor market effects, production effects, and wage effects, at the end.

This paper attempts to study the process of the second unbundling by taking seriously the fact that engineering details of the supply chain dictate the way in which different stages of production fit together.

In a stylized version of a supply chain, various parts are used to create components, which come together for final assembly to be sold domestically or abroad. There may be movements within a country or unbundled movements between plants in different countries. The engineering details involved here can be broken down into two configurations: spiders and snakes. The spider represents a situation where multiple limbs (parts) come together to form a body (assembly), which may be a component or the final product itself. The snake, on the other hand, is linear, and is more similar to a classic assembly line where a part moves down the assembly line with value (labor or capital) added at each stage. Most production processes are complex mixtures of the two.

To characterize how the supply chain would unbundle, let us assume an extremely simple framework of perfect competition and constant returns, and to give some sort of agglomeration force to the model, let us consider a two-country world, North and South, and place all consumption in the North. Here, there are two types of costs. The first is the shipping costs of the final good to the North, or traditional trade costs. The other is the offshoring cost for each part, made up of costs of coordination and management.

These off-shoring costs explain why factories are bundled spatially even within nations. When the ICT revolution came along, the cost of offshoring dropped, and some parts were able to be offshored. However, offshoring costs created two forces. One is an agglomeration force, which binds related stages together from the perspective of coordination cost. The other is a dispersion force, which encourages dispersed production from the perspective of production cost. The interaction of these forces determines the location of different parts of a value chain. Each part therefore has two aspects: How much cheaper would it be to produce it in the South? And how expensive would that be in terms of coordination cost?

There are two thresholds depending on where assembly is, and this gives us three sets of parts, say N, NS, and S. N is always cheapest to produce in the North regardless of the location of assembly. A comparative advantage dominates even if assembly is in the South. S is always cheapest to produce in the South regardless of the location of assembly. NS is cheapest to produce in the same place as assembly. So, when will an agent (assembler) choose to locate assembly in the North or in the South? This also depends on engineering detail—snakes or spiders.

In the spider, when an assembler determines the location of assembly and all parts in order to minimize total costs, in a situation referred to as "single-agent cost minimization," the answer is given by the comparison of total costs. For example, assembly would be located in the South if total costs of assembly are greater in the North even taking into account the shipment of parts. One reason for this is because the natural engineering detail of the spider allows parts to be ordered according to comparative advantage.

The relationship between comparative advantage and offshoring costs can create what is referred to as "offshore overshooting." For example, let us assume that all offshoring costs are equal for all parts. When offshoring costs are high, all parts are made in the North and the assembly is in the North, because that is where the consumers are located. As trade costs start coming down, more parts start moving to the South. At some point it becomes cost minimizing to relocate assembly to the South, so assembly also moves to the South, taking with it many more parts than comparative advantage would suggest natural. As the second unbundling proceeds, the comparative advantage of the North returns, and a re-onshoring of parts occurs; that is, some parts come back to the North. This is what offshore overshooting is about.

In contrast to the above, in a situation where there is Nash equilibrium and parts producers independently decide where to locate given the location of other firms and not the assembler, multiple equilibria may arise. When offshoring is expensive, it is an equilibrium that all the parts and assembly are done in the North, but it is also equilibrium to have all parts in the South, in terms of saving on trading costs. Interestingly, this is basically the goal of governments who are trying to create clusters. They are trying to create a fait accompli which will make their location attractive to more people.

Compared to the spider, the snake is slightly more difficult to deal with, because products cannot be freely ordered by comparative advantage. The relative cost of parts in the South and its offshoring costs within the linear chain of the snake can "wander," because parts in the upstream may be labor-intensive, or parts in the downstream may be labor-intensive, with parts that are skill-intensive between them. Moreover, offshoring costs are not monotonic. It may be cheap to offshore some parts and more expensive to offshore others. This makes the snake fundamentally harder to deal with but perhaps the more interesting.

The decision to offshore is, again, cost minimization. If a certain segment is offshored, all parts of the stage are moved to the South. This saves on factor costs corresponding to that segment, but incurs offshore costs. When the cost savings are weighed against offshore costs, the assembler might consider offshoring another part upstream or downstream to offset the costs of offshoring, and thus create agglomeration.

A generic property of this model is that it does not produce small segments of supply chain, but rather it creates cluster tendencies. This also means offshore overshooting could arise. If one stage is already offshored, trade costs favor production of the immediate upstream and downstream stages, and once a threshold is reached, the stages will move to the North as offshoring costs fall.

Now, let us think about labor market implications.

Assume that the North has Hicks neutral technological advantage. In an equilibrium where wages are lower in the South and technology is superior in the North, offshoring allows the North to take its good technology to the South, employ southern labor, do some processing, and re-import it into the North. In other words, the North can produce the same outputs with fewer resources. This type of essential change in the production structure of the North is called "shadow migration."

In particular, if the offshoring is in labor-intensive products, it will keep or bias the production structure in the North towards labor-intensive sectors. For example, a country can continue exporting a product that is very labor intensive by offshoring some of the most labor-intensive parts. In relation to our discussion on what bits go first, there is a tendency to move out the labor-intensive bits, which feeds into the general equilibrium production structure. The same applies to wages. The amount of labor used in the North changes with offshoring, because by doing some of the processing abroad, the original technology matrix changes. By multiplying the amount of labor with wages in the North and adding what is paid to laborers in the South for the offshore bits, you get the pricing.

Now, in terms of cost savings by sector due to the offshoring, we see a Stolper-Samuelson-like effect. What matters is not whether the cost savings is in the labor-intensive industry or the capital-intensive industry, or the nature of the offshoring itself, but which sector it is in. This means that general equilibrium effects on wages depend upon where the cost savings is greatest, not the nature of cost savings per se. In fact, an example by Grossman and Rossi-Hansberg has shown that offshoring labor-intensive steps or stages raises the wage of labor in the offshoring country.

This theory is still in its early stage of development, but we think it provides a framework for thinking about what will be offshored next and what the determinants are. It is not as easy as one might think, however, because it is not just a matter of labor cost, but also clustering, agglomeration, etc. Moreover, the theory needs guidance from facts on unbundling in specific industries, because there will surely be bifurcation according to industries. As a result, we may end up with models that look a lot like snakes or those which look a lot like spiders, and in between may be a model that puts them together.

YANG Tzu-Han (Professor, National Taipei University)

Large MNCs have grown rapidly in size of assets and sales and have facilitated global and regional economic integration through FDI and intra- and inter-firm trade. There is a high correlation between MNCs and investment, such that countries with larger MNCs invest more abroad and receive more investment from abroad.

There is also a regional gravity effect that bilateral FDI intensity is high among Japan, NIEs and ASEANs, which may induce higher trade intensity in the East Asia region. However, the structure of this East Asian cluster has changed so that the inner core is now China and Japan. This change implies that the sequence of flying geese pattern has changed, and that China may have caught up and passed over the ASEAN4. RCA index analysis results also show that China has inherited more industries directly from NIEs and Japan-US than from the ASEAN4.

ITO Banri (Fellow, RIETI / Lecturer, Department of International Economics, School of Economics, Senshu University)

I have three comments in regard to the discussion by Professor Yang.

First, FDI intensity was defined as the bilateral flow between countries to the total volume of world trade, but is it necessary to standardize FDI intensity to take into account the change in composition of sample countries? And is FDI intensity also computed by "flow"? Perhaps "stock" is better to construct FDI intensity.

Second, to compare how the growth of the East Asian bloc differs from other blocs, I suggest that a simple descriptive examination may be feasible through calculations using relative intensity and homogeneity in a cluster.

The third is my personal thought. I think the robustness check for the relationship between trade and economic integration can be achieved by estimating the gravity equation with the "cluster category dummy." I also think that an analysis on the difference in magnitude between the institutional regional dummy and cluster category dummy may be an interesting issue.

AHN Joyup (Senior Research Fellow, Korea Labor Institute)

In trying to understand why nonstandard workers are so prevalent in the Korean economy, some say it is induced by globalization or openness, which creates competition so that firms must maximize their profits by utilizing more nonstandard workers. However, since there is little empirical evidence of this, I am testing the null hypothesis that globalization measures have no significant effect on the share of nonstandard workers.

So far, data analysis indicates that there is no significant close relationship between openness and the share of nonstandard workers even when nonstandard workers are classified into three categories: contingent work, part-time work, and an alternative employment arrangement. Therefore, I have also examined trends in nonstandard employment by type, industry size, and firm size in relation to various other variables. My study is ongoing, but my best guess is still that openness is not so directly related with the share of nonstandard employment.

TSURU Kotaro (Senior Fellow, RIETI)

I have four comments concerning Professor Ahn's paper, in comparison with the situation regarding nonstandard workers in Japan.

First, why does the share of contingent work have a downward trend while part-time and alternative employment show an upward trend in Korea? I wonder if the regulatory change concerning fixed-term contract workers made in 2007 has had any effect.

Second, what is the impact of the recent economic crisis on nonstandard workers in Korea? In Korea, the number and share of nonstandard workers did not decrease in 2009 after the Lehman shock like it did in other countries.

Third, I found some differences with the case of nonstandard work in Japan, especially by firm size and by industry.

SATO Hitoshi (Fellow, RIETI)

We attempted to understand how economic globalization might shift the labor demand from permanent to temporary workers and derived the following empirical findings.

First, volatility raises the temp ratio. Firms prefer less volatile revenue fluctuations and try to accommodate them by using temporary workers who have low dismissal costs.

Second, larger firms have more products than smaller firms, and larger firms tend to have lower revenue volatility than smaller firms. In addition, larger firms have a lower temp ratio than smaller firms. This suggests that larger firms can diversify the risk of revenue fluctuations more than smaller firms by pooling the risk of revenue fluctuations in their larger number of products.

Finally, high-export firms tend to decrease their number of products and raise their temp ratio. This can be interpreted such that trade openness creates tough competition so that firms tend to streamline their products. This may raise the risk of revenue fluctuations and increase demand for temporary workers.

ONO Yukako (Keio Economic Observatory, Keio University)

• The volatility measure already incorporates the effect of international trade. So what other effect is being addressed by including the interaction term between volatility and export intensity?
• If multi-product firms are less volatile because they are more diversified, why do we include a separate scale measure?
• Why are data at the plant level used? It might make more sense to use firm-level data.
• When looking at the use of temporary workers in manufacturing, it may be also interesting to consider the role of inventory.
• Sato-san talked about the causality coming from the global market conditions to the labor market outcomes, but we hear a lot about labor market conditions determining global market outcomes. This should perhaps be discussed in the empirical section because we may be capturing the reverse causalities.

ICHIDA Toshihiro (Associate Professor of International Trade, School of Commerce, Waseda University)

This paper incorporates a job assignment model of individual comparative advantage with human capital investments. When workers have multi-dimensionally heterogeneous skills, each worker has a choice in making either a general or specific skill investment. This paper looks at questions such as how a worker makes his or her investment decision and what the directional incentives for their decisions become given the different degrees of risk aversion. If agents are risk-neutral, a specific investment is made only in the agents' innately strong skills, because risk-neutral agents simply try to maximize expected income rather than expected utility. This is particularly true of workers with strong comparative advantage. For agents with a smaller comparative advantage, general investment may occur when the general investment parameter exceeds a certain threshold value.

In contrast, if agents are risk-averse, agents with smaller comparative advantages (who have similar abilities in two sectors) may make a general or specific investment depending on the degree of risk aversion. When the parameter of risk aversion is larger than one, those who have the smallest comparative advantage will not only make a specific investment, but might also do so in the direction of strengthening their weak skills.

JINJI Naoto (Associate Professor, Faculty of Economics, Kyoto University)

I have one question and three comments.

The question is whether individual choices of specific and general investments are socially efficient. Mr. Ichida posed a question but did not provide an answer.

My first comment is about implications of the results. For example, I wish to know more about whether the results explain any specific economic phenomenon, or whether policy implications can be obtained from them.

My second comment is about the relative value of talent. In addition to relative comparative advantage, I think the cost of human capital investment should also be factored in to provide consistency with real world observations.

My final comment is about extension of the model. The paper relies on a lot of symmetries, but providing discussions about relaxing at least some of the assumptions might make the paper more interesting.

TOMIURA Eiichi (Faculty Fellow, RIETI / Professor, Faculty of Economics, Yokohama National University)

As the economy globalizes, labor demand becomes more elastic, and economic activities that cross the border expand from imported goods to offshoring.

We have empirically evaluated the impacts of offshoring on employment flexibility by the share of regular full-time workers in total employment, and found that offshoring is significantly related to flexible employment even after firm characteristics are controlled for.

We also found that the impact of offshoring, such as declining regular employment, depends not on offshoring in general, but on the type of offshore task. While offshoring of professional services has a particularly strong potential negative impact, production offshoring loses its significance once the effect of labor market deregulation is considered. It is important therefore to separate globalization offshoring effects from domestic institutional effects.

The measure and the magnitude of the impacts remain to be addressed in future research.

ENDOH Masahiro (Professor, Faculty of Business and Commerce, Keio University)

Tomiura sensei's paper differs from other studies on the relationship between trade and labor competition in that it focuses on offshore sourcing. It finds that offshoring firms have decreased their regular worker ratio mainly as a factor of the year dummy "After," and that a sizeable effect of offshoring on the regular worker ratio can also be expected in the near future. This analysis can perhaps be applied to translog functions in reference to wage data. It might be difficult to connect plant-level wage data and the firm-level data of this analysis, but it should be worth doing. When doing so, I think it is important to clarify whether firm size is defined by sales, as a function of labor, capital and intermediate inputs, or by value-added, which is a function of labor and capital.

I also propose that, since this survey is very unique and original, the data should perhaps be shared with other researchers.

LI Hongbin (Professor, School of Economics and Management, Tsinghua University)

One inequality that is garnering a lot of attention in China in recent years is educational inequality.

In order to get into college, almost all students need to take the College Entrance Exams. About 70% are admitted, but the admission rate to top elite universities is extremely low. The fate of CEE takers is determined by various biases. These include gender bias, for one, as most top science students in elite universities are males. There is also the hukou (urban vs. rural) bias, which works in favor of urban students, and an affirmative policy that gives preference to minorities. Training bias shows that CEE takers who take the test twice improve their score, and home bias shows high local admission rates. The high school bias is evident in that only students from the top high schools have a chance to go to an elite college.

TODO Yasuyuki (Faculty Fellow, RIETI / Professor, Department of International Studies, Graduate School of Frontier Sciences, the University of Tokyo)

China has grown rapidly in this globalized world owing to technological progress that is attributable to competent Chinese people cultivated through fair and tough competition.

However, I have a few questions and suggestions for possible future research. What are the sources of possible biases in Chinese exams? What are the incentives of the biases? And does equality based on CEE scores really mean educational equality? Perhaps these issues could be examined more closely.

In Japan, recent changes in college admissions have obscured the criteria for getting into good schools. Students do not know what to do, and become reluctant to make efforts. This could lead to an economic slowdown. As a lesson from Japan, I wish to say that educational equality is important in the sense that rules are made clear to everyone and students believe that their efforts will be rewarded.