102
 American Economic Journal: Macroeconomics 2012, 4(2): 102?132 
http://dx.doi.org/10.1257/mac.4.2.102
 Capital Market Integration and Wages?
 By A C, P B H,  D S*
 For three years after the typical emerging economy opens its stock 
market to inows of foreign capital, the average annual growth rate 
of the real wage in the manufacturing sector increases by a factor of 
three. No such increase occurs in a control group of countries that do 
not liberalize. The temporary increase in wage growth drives up the 
level of the average worker?s annual compensation by US $487?an 
increase equal to nearly one-fth of their annual pre-liberalization 
salary. Overall, the results suggest that trade in capital may have a 
larger impact on wages than trade in goods. (JEL E25, E44, F16, 
F43, G18, O16)
 The impact of trade on wages occupies a salient space in the collective imagina-tion of the economics profession. When a country opens up to trade with the 
rest of the world, income shifts away from that country?s scarce factor of production 
and toward the one that is abundant (Stolper and Samuelson 1941). Inspired by the 
celebrated Stolper-Samuelson Theorem, economics journals abound with articles 
examining the extent to which trade induces factor price equalization.
 The evidence so far is mixed. The consensus view suggests that trade with develop-
 ing countries is, at best, a modest force behind the large decline in the relative wages 
of low-skilled workers in rich countries (Krugman 1995; Lawrence and Slaughter 
1993; Cline 1997; Lawrence 2008).1 In the case of workers in developing countries, 
the evidence actually runs contrary to the theory. Whereas Stolper-Samuelson predicts 
that trade with rich countries will increase the relative wages of low-skilled workers 
in poor countries, trade liberalization during the 1980s and 1990s actually increased 
wage inequality in the developing world (Goldberg and Pavcnik 2007).
 1 Feenstra and Hanson (2003) provide a dissenting view.
 * Chari: Department of Economics, University of North Carolina at Chapel Hill, CB#3305 Chapel Hill, NC 
27599 and National Bureau of Economic Research (e-mail: achari@unc.edu); Henry: Department of Economics, 
Stern School of Business, New York University, New York, NY 10012 and Brookings Institution (e-mail: pbhenry@
 nyu.edu); and Sasson: Goldman Sachs Asset Management, 200 West Street, New York, NY 10282 (e-mail: diego.
 sasson@gmail.com). Henry gratefully acknowledges nancial support from the W.R. Berkley and Richard R. West 
Professorships, the John A. and Cynthia Fry Gunn Faculty Fellowship, the Stanford Institute for Economic Policy 
Research, and the Stanford Center for International Development. We thank Sandile Hlatshwayo for excellent 
research assistance and Allison Parker for editorial help. We also thank Olivier Blanchard, Steve Buser, Brahima 
Coulibaly, Jonah Gelbach, Pierre-Olivier Gourinchas, Avner Greif, Nir Jaimovich, Pete Klenow, Anjini Kochar, 
John Pencavel, Paul Romer, Robert Solow, Ewart Thomas, and seminar participants at the University of California-
 Berkeley, Brookings Institution, the Chicago Fed, Claremont McKenna, the International Monetary Fund, 
Massachusetts Institute of Technology, NIPFP-DEA, New York University, the Peterson Institute for International 
Economics, the Reserve Bank of India, and Stanford University for helpful comments. All views and remaining 
errors in the paper are our own.
 ? To comment on this article in the online discussion forum, or to view additional materials, visit the article page 
at http://dx.doi.org/10.1257/mac.4.2.102.
VOL. 4 NO. 2 103CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 Moving from trade in goods to trade in factors, an extensive literature also exists 
on the impact of labor ows on wage inequality. Again, the results are mixed. Some 
studies nd that immigration from developing countries exacerbates wage inequal-
 ity in the United States (Borjas, Freeman, and Katz 1997). Others nd little to no 
effect (Card 2009; Ottaviano and Peri 2008).
 While many studies examine the impact of cross-border ows of goods and work-
 ers on relative wages, the literature pays far less attention to the impact of cross-
 border nancial ows on the absolute level of wages. This is surprising for at least 
three reasons.
 First, trade in capital between nations has implications for real wages that are 
every bit as important as cross-border movements of goods and people. In emerging 
economies, where capital is scarce and labor abundant, opening up to free trade in 
capital should reduce the rental rate and increase the real wage.
 Second, examining the absolute level of wages provides information about the 
impact of opening up on the distribution of income between capital and labor that is 
just as important as the information that studies of wage inequality provide about the 
distribution of labor income between high- and low-skilled workers. For instance, 
many emerging economies experienced unprecedented increases in national income 
as a result of globalization in the 1980s and 1990s. If all of the income gains from 
globalization accrued to capital, then the rise in wage inequality documented by 
Goldberg and Pavcnik (2007) necessarily implies that low-skilled workers experi-
 enced income losses. On the other hand, if total labor income grew in line with (or 
faster than) the economy as a whole, then high-skilled workers may have experi-
 enced income gains that did not result in losses for low-skilled workers.
 Third, in the late 1980s emerging economies all over the world began easing 
restrictions on capital inows of all kinds, giving economists a series of before-and-
 after scenarios with which to study the impact of factor ows on factor rewards. A 
large body of research examines the impact of capital market liberalization on asset 
prices, investment, and the growth rate of GDP per capita.2 But to the best of our 
knowledge, this literature is silent about the impact of capital account opening on 
the labor market. Consequently, two decades after the onset of capital market liber-
 alization, we still have no systematic evidence about the impact of this sea change in 
policy on the average level of wages in the developing world.3 This paper provides 
the rst systematic attempt to ll that gap.
 Figure 1 demonstrates that the level of the average annual manufacturing real 
wage in a sample of 25 emerging economies increased signicantly after they lib-
 eralized restrictions on inows of foreign capital between 1986 and 1996.4 Formal 
estimates show that the growth rate of the real wage in local currency terms jumped 
from 1.8 percent per year in the pre-liberalization period to an average of 5.7 per-
 cent in the year liberalization occurred and each of the subsequent three years. The 
2 See Henry (2007) and Obstfeld (2009) for comprehensive surveys of this literature.
 3 Feenstra and Hanson (1997) explore the cost of capital but focus on its impact on relative wages. Aitken, 
Harrison, and Lipsey (1996); Almeida (2007); and Hale and Long (2008) examine FDI and wages, but not the 
general connection between nancial ows and wages vis-?-vis the cost of capital.
 4 In order to have comparable measures of levels of wages across countries, we plot the natural log of the real 
wage in PPP adjusted $US terms.
104 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 3.9 percentage point increase in the growth rate of the real wage during this window 
drives up the level of average annual compensation for each worker in the sample 
of liberalizing countries by the local currency equivalent of US $487?an increase 
equal to 18 percent of their annual pre-liberalization salary.
 One concern about Figure 1 is that an exogenous worldwide shock unrelated to 
capital market opening drove up real wages in liberalizing and nonliberalizing coun-
 tries alike. To distinguish the country-specic impact of liberalization policy from 
that of a common shock, our estimation procedure compares the difference in wage 
growth before and after liberalization for a group of countries that open up to the same 
difference for a group of control countries that do not. Our regressions also include 
year-xed effects to account for the possibility of common shocks that affect only 
the liberalizers and country-xed effects to allow for differences in underlying unob-
 servable factors that may drive variation in wage growth across countries. We also 
control for the impact of contemporaneous macroeconomic reforms, such as ination 
stabilization, trade liberalization, privatization, and Brady Plan debt relief programs. 
In every specication, we nd an economically and statistically signicant increase in 
real wage growth for countries in the liberalization group relative to the control group.
 An open economy interpretation of the neoclassical growth model provides the 
cleanest qualitative explanation of the new facts we uncover. Opening up to capital 
inows reduces the cost of capital in developing countries, and rms respond by 
increasing their rate of investment. For a given growth rate of the labor force and 
total factor productivity, a higher rate of investment increases the ratio of capital per 
effective worker, driving up the marginal product of labor and, in turn, the market-
 clearing wage. Consistent with this chain of logic, Figure 2 demonstrates that the 
growth rate of labor productivity also rises sharply following liberalizations. After 
controlling for other factors, the average growth rate of labor productivity is 9.72 
percentage points higher during the four-year liberalization window than it is in 
nonliberalization years.
  
7.6
 7.8
 8
 8.2
 8.4
 8.6
 8.8
 ?5          ?4         ?3        ?2          ?1          0            1           2            3            4            5
 Natural log of the real wag
 e
 Year relative to liberalization
   
 Average for 25 liberalizers
 F 1. R W G R   W 
 C A L
	
VOL. 4 NO. 2 105CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 From a quantitative perspective, however, it is less clear whether the neoclassi-
 cal model captures all relevant features of the data. In the standard growth model, 
capital account liberalization works strictly through its impact on capital accumu-
 lation and has no effect on the growth rate of aggregate total factor productivity 
(Gourinchas and Jeanne 2006). The increase in real wage growth present in the data 
is too large to be explained exclusively by capital deepening under conventional 
assumptions about capital shares and the elasticity of substitution between capital 
and labor. One possible explanation stems from the observation that liberalizations 
increase the quantity of capital goods that emerging economies import from indus-
 trial nations (Alfaro and Hammel 2007). If technology diffuses from developed to 
emerging economies through the technology embodied in capital goods imports ? 
la Eaton and Kortum (1999, 2001a,b), then liberalizations may indeed drive up the 
growth rate of total factor productivity.
 While our approach enables us to test previously unexamined real wage impli-
 cations of capital market opening, difference-in-differences estimation requires 
caution because the standard errors are susceptible to serial correlation (Bertrand, 
Duo, and Mullainathan 2004). Opening up to foreign capital increases investment, 
which in turn drives up productivity and wages. Because wages take time to adjust, 
wage growth for a given country may remain elevated above its steady-state rate for 
a number of years after opening, thereby inducing serial correlation in the coun-
 try?s wage-growth residuals. Similarly, many countries open up at approximately the 
same time, possibly inducing cross-country correlation in the residuals. Our empiri-
 cal analysis uses the Petersen (2009) technique to simultaneously adjust the stan-
 dard errors for the potential presence of both types of correlation in the residuals. No 
matter how we compute the standard errors, the impact of capital market opening on 
wages and productivity remains economically and statistically signicant.
 The potential endogeneity of the liberalization decision also raises some con-
 cerns. If prot-maximizing rms in a nancially closed economy face the prospect of 
 
?5          ?4         ?3        ?2          ?1          0            1           2            3            4            5
 Natural log of real value added per worke
 r
 Year relative to liberalization
 Average for 25 liberalizers
 8.5
 8.6
 8.7
 8.8
 8.9
 9
 9.1
   
  
F 2. P R   W 
 C A L
	
106 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 rapidly rising labor costs, they will want to substitute capital for labor. To the extent 
that opening up the capital account would reduce the cost of capital, these rms have 
an incentive to lobby the government to do so. If rising wages cause governments to 
open up, then our estimates will spuriously indicate a strong impact of liberalization 
on wages, when causation in fact runs the other way. While theoretically plausible, 
the endogeneity argument has no empirical support. Figure 1 is not consistent with 
the view that capital market opening occurs in response to rising labor costs. If any-
 thing, wage growth actually falls slightly prior to the opening. (Section IIIC shows 
that mean reversion ? la Ashenfelter 1978 does not drive our results.) The data are 
also not consistent with the explanation that governments liberalize in anticipation of 
higher future labor costs. Although wages rise sharply following liberalization, labor 
productivity rises even faster, so unit labor costs do not increase.
 Finally, with only 25 countries in the sample, one may worry that a few large out-
 liers drive the central nding. This is not the case. Sign tests show that the median 
growth rate of real wages in the post-liberalization period exceeds the pre-liberaliza-
 tion median too often to be explained by chance.
 The rest of the paper proceeds as follows. Section I uses theory to generate test-
 able predictions and explains how we identify real life liberalization episodes. 
Section II describes the data and construction of the control group, and presents 
preliminary ndings. Section III discusses the empirical methodology, main results, 
and alternative interpretations. Section IV examines the consistency of the results 
with the theory. Section V concludes.
 I. Capital Market Integration in Emerging Economies
 This section uses an open economy version of the Solow (1956) model to gener-
 ate previously untested predictions about the impact of capital ows on the time-path 
of the real wage (w). The central theoretical point about capital market integration is 
that it moves emerging economies from a steady state in which their ratios of capital 
to effective labor are lower (and rates of return to capital higher) than in the indus-
 trialized world, toward a steady state in which ratios of capital to effective labor and 
rates of return are equal in both regions.
 Because capital and labor are complements in production, the marginal prod-
 uct of labor (and hence the real wage) rises as countries open up and the process 
of capital deepening sets in. This fundamental insight about capital ows and the 
dynamic path of wages would also hold in an open economy Ramsey model. Since 
the focus of the paper is on wages, not other variables (e.g., the current account) 
that depend on endogenous savings decisions, the Solow model provides the most 
concise exposition.
 A. Theory
 Assume that a country produces output using capital, labor, and a constant-
 returns-to-scale production function with labor-augmenting technological progress:
 (1)  Y = F(K, AL).
VOL. 4 NO. 2 107CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 Let k = K/AL be the amount of capital per unit of effective labor, and y = Y/AL 
the amount of output per unit of effective labor. Using this notation and the homo-
 geneity of the production function, we have
 (2) y = f  (k).
 Also assume that the country saves a constant fraction of national income each 
period and adds it to the capital stock; capital depreciates at the rate ?, the labor 
force grows at the rate n, and total factor productivity grows at the rate g.
 When the economy is in steady state, k is constant at the level  ks.state , and the mar-
 ginal product of capital equals the interest rate (r) plus the depreciation rate:
 (3)  f ?( ks.state ) = r + ?.
 Since the impact of liberalization works through the cost of capital, equation (3) 
has important implications for the dynamics of k and w in the wake of opening up.
 Let r * denote the exogenously given world interest rate. The standard assumption 
in the literature is that r * is less than r, because the rest of the world has more capital 
per unit of effective labor than the developing country. It is also standard to assume 
that the developing country is small, so that nothing it does affects r *. Under these 
assumptions, capital surges in to exploit the difference between r * and r when the 
developing country liberalizes.
 The absence of any frictions in the model means that the country?s ratio of capital 
to effective labor jumps immediately from  ks.state to its post-liberalization, steady-
 state level ( k s.state *  ). In the post-liberalization steady state, the marginal product of 
capital equals the world interest rate plus the rate of depreciation:
 (4)  f ?( k s.state *  ) =  r * + ?.
 Instantaneous convergence implies that interest rates equalize immediately and 
that the country installs capital at the speed of light. Two remarks about this unat-
 tractive feature of the model are in order.
 First, instantaneous capital market convergence is not an artifact of the Solow 
model, but of the small open economy assumption under which liberalization gives 
the country access to an innitely elastic supply of capital at the world interest 
rate. The same counterfactual phenomenon would also occur in an open economy 
Ramsey model. Second, although we do not see equalization of interest rates and 
capital-labor ratios across countries in the real world, a large literature documents 
that the cost of capital drops, and investment booms when developing countries 
remove barriers to capital inows.5
 5 See Henry (2007), Stulz (2005), and the references therein.
108 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 There are a variety of formal methods for slowing down the speed of transi-
 tion (e.g., adjustment costs of capital installation), but all of these methods would 
belabor the exposition without altering the model?s fundamental prediction.6 The 
vital point is that  ? ?k is greater than zero during the country?s transition to its post-
 liberalization steady state. The temporary growth in k has important implications for 
the time path of real wage growth, which we now derive.
 The growth rate of the real wage is the derivative of the natural log of w with 
respect to time, that is,   ? w  _ w =  d _ dt (ln w(t)). Since workers are paid their marginal prod-
 uct of labor, w = A[  f (k) ? k f ?(k)]. This means that the growth rate of the real wage 
is given by   ? w  _ w =  d _ dt (ln (w (t)) =  d _ dt {A[  f (k) ? k f ? (k)]} =   ? A  _ A ?  k f ?(k) ? k _ [  f (k) ? k f ?(k)] . We may 
write this expression as
 (5)    ? w  _ w =   
?
  A 
 
_
  A +  
1
  
_
  ?  ?   
f ?(k)k
  
_
 f (k)  ?   
 
?
  k
  
_
  k ,
 where ? = ? f ?(k)[  f (k) ? k f ?(k)]  _f (k) f ? (k)k   is the elasticity of substitution.
 The right-hand side of equation (5) demonstrates that the growth rate of the real 
wage depends on the sum of two terms. The rst term, the growth rate of total fac-
 tor productivity ( ? A /A), is not affected by capital account policy in the canonical 
version of the neoclassical growth model. In Section IV, we discuss the implica-
 tions of recent work that adopts a more catholic view of the relationship between 
capital account liberalization and total factor productivity. For now, we proceed as 
though the impact of liberalization works strictly through the second term, which 
is the product of the inverse of the elasticity of substitution (1/?), capital?s share in 
national income ( f ?(k)k/f (k)), and the growth rate of the ratio of capital per unit of 
effective labor ( ? k/k).
 Prior to liberalization, the ratio of capital to effective labor is constant at the level 
k
  s.state
  , so that  ? k/k equals 0, and w simply grows at the same rate as total factor produc-
 tivity. Since  ? k/k is greater than 0 during the transition to  k s.state * , the growth rate of the 
real wage also increases temporarily. Figure 3 illustrates the hypothetical time paths of 
r and the natural logs of k and w under the assumption that the interest rate converges 
immediately upon liberalization, but the ratio of capital to effective labor does not.
 Again, previous work documents that the actual responses of the cost of capital 
and the quantity of capital to liberalization resemble their hypothetical time paths. 
Figure 1 demonstrates that the growth rate of the real wage also behaves in accor-
 dance with the theory. In Section IV, we examine whether the size of the real wage 
increase is consistent with the magnitude of the previously documented increases in 
the growth rate of capital. The next subsection explains how we identify the real life 
liberalization episodes used to construct Figure 1.
 6 See Barro and Sala-i-Martin (1995, chapter 2) and Henry (2007, section 4.1).
VOL. 4 NO. 2 109CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 B. Reality
 An ideal test of the prediction that real wage growth will rise following the removal 
of restrictions on capital inows requires information on capital account liberaliza-
 tion dates that is more precise than one can generally obtain. In theory, opening the 
capital account is as simple as pulling a single lever. In reality, the capital account 
has many components, so trying to determine exactly when a country liberalizes 
(as in Subsection IIA) is not a trivial task. In fact, the difculty of determining 
 
r
 0
  
 
0
  
 
r*
 t
 Panel A. The cost of capital
 Panel B. Ratio of capital to effective labor
 In(k)
 In(k*
 s.state
 )
 In(k
 s.state
 )
 t
 Panel C. Real wage
 In(w)
 0
 t
 F 3. H I 
 L
	   C 
 C, I   R W
110 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 precise liberalization dates causes most papers in the literature to ignore the prob-
 lem (Eichengreen 2001). Instead of asking whether opening the capital account has 
an impact on a country?s growth rate (as theory clearly dictates), most published 
studies examine whether openness and long-run growth are positively correlated 
across countries. The distinction matters greatly. Testing for the effect of openness 
on growth can produce spurious results that tell us nothing about the true impact of 
liberalization (Henry 2007).
 In contrast to the previous literature, which makes no attempt to nd periods of 
opening, this paper identies liberalization dates using the point in time when coun-
 tries rst permitted foreigners to purchase shares of companies listed on the domes-
 tic stock market. Relative to the most general conception of the capital account, at 
rst blush the lifting of restrictions on foreign investment in the stock market may 
seem like a narrow way to dene capital account liberalization. But deeper reection 
reveals that stock market liberalizations serve as observable de facto indicators of 
harder-to-pinpoint de jure policy changes.7
 For instance, the establishment by a foreign nancial institution of an equity 
mutual fund is the modal means through which countries rst liberalize their stock 
markets (see Table 1). If country-fund opening dates are valid proxies for the occur-
 rence of broader, undocumented liberalizations that took place in the late 1980s and 
early 1990s, then signicant quantities of capital that are not associated with any 
particular fund may ow in to the country as a consequence of the opening. Three 
facts suggest that our stock market liberalization dates provide nonspecious indica-
 tors of a larger move toward open capital markets.
 First, a steady stream of country funds, issuances of American Depository 
Receipts, and other vehicles for foreign savers to buy domestic stocks typically 
follow the initial stock market liberalizations (Karolyi 2004; Gozzi, Levine, and 
Schmukler 2010). As a case in point, Chile liberalized its stock market in May 1987 
through the Toronto Trust Mutual Fund, a Canadian closed-end fund with a net asset 
value of US $37.7 million.8 In the following ve years, six additional country funds 
with a cumulative net asset value of $991.8 million were established in Chile.9
 Moving beyond Chile, Figure 4 demonstrates the extraordinary change wrought 
by stock market liberalizations in emerging economies. Net inows of equity capi-
 tal to the developing world, practically nonexistent in the 1970s and early 1980s, 
accelerated sharply after the median stock market liberalization date in our sample 
(1989). Policymakers often raise concerns about the potential volatility of portfolio 
capital, worrying that net inows of equity can easily turn to net outows at the 
rst sign of trouble. The facts do not substantiate this concern. While Figure 4 indi-
 cates that the pace of equity inows to developing countries slowed following the 
Mexican Crisis in 1994?1995 and the Asian Crisis of 1997?1998, there is nothing 
that remotely resembles a reversal of equity inows.10
 7 See Kose et al. (2006) for a detailed discussion of de facto versus de jure indicators.
 8 See Park and Van Agtmael (1993), Price (1994), and Wilson (1992).
 9 See Park and Van Agtmael (1993), Price (1994), and Wilson (1992).
 10 Concerns about the volatility of debt ows, on the other hand, are quite well founded. See Henry (2007) and 
Rogoff (1999).
VOL. 4 NO. 2 111CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 It is also important to note that stock market liberalizations account for a substan-
 tial fraction of foreign direct investment (FDI), the most stable form of foreign capi-
 tal. A common misconception views FDI solely as green eld investment, where 
a foreign company builds from scratch a new manufacturing plant in an emerging 
economy. As a matter of ofcial statistics, FDI includes any stock transaction (i.e., a 
cross-border merger or acquisition) that results in the purchaser owning 10 percent 
or more of the voting shares.11 In fact, from 1991?2000, cross-border mergers and 
acquisitions accounted for 48 percent of FDI in Latin America and 61 percent in 
East Asia (Chari, Ouimet, and Tesar 2004).
 11 The Organisation of Economic Co-operation and Development (OECD) denes foreign direct investment 
(FDI) as ?A category of cross-border investment made by a resident in one economy (the direct investor) with the 
objective of establishing a lasting interest in an enterprise (the direct investment enterprise) that is resident in an 
economy other than that of the direct investor. The ?lasting interest? is evidenced when the direct investor owns at 
least 10 percent of the voting power of the direct investment enterprise? (OECD 2008).
 T
 1? C A L
	 O A  S T   
O M E R

 Country
 Capital account 
liberalization Stabilization
 Trade 
liberalization Privatization
 Brady plan  
debt relief
 Argentina Nov-89 Nov-89 Apr-91 Feb-88 Apr-92
 Brazil Mar-88 Jan-89 Apr-90 Jul-90 Aug-92
 Chile May-87 Aug-85 1976 1988 NA
 Colombia Dec-91 NA 1986 1991 NA
 Egypt Feb-91 Apr-91 Apr-91 Apr-91 NA
 Greece Jul-94 Jul-89 Apr-53 Nov-90 NA
 India Jun-86 Nov-81 1994 1991 NA
 Indonesia Sep-89 May-73 1970 1991 NA
 Israel Oct-89 Jul-85 Feb-52 Jan-86 NA
 Jordan Dec-95 May-94 1965 Jan-95 Jun-93
 Malaysia May-87 NA 1963 1988 NA
 Mexico May-89 May-89 Jul-86 Nov-88 Sep-89
 Morocco Dec-92 Jan-84 Sep-83 1993 NA
 Nigeria Aug-95 Jan-91 NA Jul-88 Mar-91
 Pakistan Feb-91 Sep-93 2001 1990 NA
 Philippines May-86 Oct-86 Nov-88 Jun-88 Aug-89
 Portugal Jan-93 Oct-90 Jan-60 Apr-89 NA
 South Africa Mar-95 Mar-86 Apr-94 Apr-94 NA
 South Korea Jun-87 Jul-85 1968 NA NA
 Spain Jan-93 Jan-78 Jul-59 1985 NA
 Taiwan May-86 NA 1963 NA NA
 Thailand Sep-87 Jun-85 Always open 1988 NA
 Turkey Aug-89 Jul-94 1989 1988 NA
 Venezuela Jan-90 Jun-89 May 1989** Apr-91 Jun-90
 Zimbabwe Jun-93 Sep-92 NA 1994 NA
 Notes: The capital account liberalization dates identied in this table are the dates on which the 25 countries in col-
 umn 1 eased restrictions prohibiting foreign ownership of domestic stocks. Technically speaking, Greece, Israel, 
Portugal, and Spain are not emerging economies but to maintain consistency with other studies we include them in 
our sample. The liberalization dates in column 2 are an amalgamation of those in Henry (2000), Levine and Zervos 
(1998b), and Bekaert and Harvey (2000). Columns 2?6 list the dates of major economic reforms that occurred 
around the same time as the capital account liberalizations. The stabilization program dates in column 3 come from 
Henry (2002) and various issues of the IMF Annual Reports. Column 4 lists trade liberalization dates from Sachs 
and Warner (1995). The privatization dates in column 5 come from the Privatization database maintained by the 
World Bank. Finally, column 6 lists the month and year that each country received debt relief under the Brady Plan. 
The debt relief dates come from Cline (1995), Lexis Nexis, and various issues of the Economist Intelligence Unit. 
 **Venezuela reversed its trade liberalization reforms in 1993. 
112 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 Second, the facilitation of cross-border nancial ows and ownership through 
stock market liberalization also induces large inows of physical capital. Stock mar-
 ket liberalizations in emerging economies coincide with a signicant increase in 
their imports of capital goods. In a sample of 25 countries that liberalized their stock 
markets between 1980 and 1997, liberalization led to a 9 percent increase in capital 
goods as a fraction of total imports, and the share of total machine imports to GDP 
rose by 13 percent (Alfaro and Hammel 2007). Because developing countries do not 
produce a signicant portion of the capital goods that they use, the observation that 
imports of capital goods rise in concert with the advent of portfolio equity inows 
increases condence in earlier work on liberalization and aggregate investment.
 Third, with the sole exception of Malaysia during the Asian Crisis, none of the 
stock market liberalizations from Table 1 were followed by a reversal of freedom of 
foreign access.
 Taken together, these three facts conrm that stock market liberalizations signify 
the beginning of a steady march toward greater freedom of capital inows and provide 
the closest empirical analogue to the textbook example in Section IA. Accordingly, 
we use the stock market liberalization dates in Table 1 as the empirical counterpart to 
year [0] in the model of Section IA (Figure 3). Standard and Poor?s Emerging Markets 
Database covers 53 emerging economies with stock markets. Of these 53 countries, 
25 have stock market liberalization dates that are consistently used elsewhere in the 
literature and veriable from primary sources. Column 1 of Table 1 lists these 25 
countries and the year in which they liberalized.12 Table 2 presents summary statistics 
on the behavior of real wages in each of the 25 liberalizing countries. The next section 
explains the source and construction of the wage data.
 12 For further details about the complexities of determining liberalization dates see Henry (2007, section 5).
 Billions of US dollar
 s
 Net equity inflows
  
0
 50
 100
 150
 200
 250
 300
 350
 400
 450
 500
 Year
  
197
 0
 197
 2
 19
 74
 19
 76
 197
 8
 198
 0
 198
 2
 198
 4
 198
 6
 198
 8
 199
 0
 199
 2
 199
 4
 199
 6
 199
 8
 200
 0
 200
 2
 200
 4
 200
 6
 F 4. E I
 S W E E L
	  S M
VOL. 4 NO. 2 113CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 II. Data
 The wage data come from the Industrial Statistics Database of the United Nations 
Industrial Development Organization (UNIDO). UNIDO provides data on total 
wages and salaries, total employment, and output for the manufacturing sector. For a 
given year, wages and salaries include all payments to employees in cash or in kind. 
Payments include: direct wages and salaries; remuneration for time not worked, 
bonuses and gratuities, housing allowances and family allowances paid directly by 
the employer, and payments in kind. Excluded from wages and salaries are employ-
 ers? contributions on behalf of their employees to social security, pension, and insur-
 ance schemes, as well as the benets received by employees under these schemes 
and severance and termination pay.
 Conceptually, total wages and salaries equal W ? L ? H, where W is the hourly 
wage rate, L is the stock of labor, and H is total hours worked for the year. Since 
UNIDO provides no data on the number of hours worked or the hourly wage, we 
divide total wages and salaries by total employment (L) to compute the average 
annual wage (W ? H) of each worker in the manufacturing sector of each country 
(more on this point in Section IIA). UNIDO reports the value of wages and salaries 
in local currency terms. We deate each country?s nominal annual wage in local 
currency by the local consumer price index (CPI) to create a local currency-denom-
 inated real wage.
 In addition to information on wages, employment, and output, we would like 
to have data on the manufacturing capital stock. Unfortunately, UNIDO only pro-
 vides data on investment. The standard approach to an absence of capital stock data 
converts investment ows to capital stocks with the perpetual inventory method by 
making assumptions about the initial level of capital in a given year and using the 
investment ows to interpolate the subsequent time path of the capital stock.
 Interpolation is methodologically sound when the focus is on long-run relation-
 ships where assumptions about the initial stock of capital make little difference. 
In contrast, this paper focuses on short-run dynamics and therefore requires a 
clear picture of the trajectory of the capital stock during the liberalization win-
 dow. Simply put, it would be inappropriate to interpolate the growth rate of the 
capital stock during liberalization episodes when trying to measure the impact of 
liberalization on capital stock growth. Moreover, the UNIDO dataset is missing 
more than 50 percent of the country-year observations for investment in the aggre-
 gate manufacturing sector, and many of these missing observations fall within the 
liberalization window. In the absence of reliable capital stock data, we will use 
estimates of capital stock growth from previously published work (in Section IV) 
to check the consistency of our results with the theoretical channel from capital 
growth to wages.
 For each country in our sample, the annual wage data generally run from 1960 
to 2003, with the exact dates differing by country. After taking the difference of the 
natural log to compute growth rates, we have a total of 758 country-year observa-
 tions with which to identify the impact of liberalization on real wage growth. Table 
1 shows that the timing of liberalizations is correlated across countries, so these 758 
observations are not independent. For instance, liberalizations may coincide with an 
114 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 exogenous global productivity shock that drives up wages in all countries, irrespec-
 tive of whether or not they liberalize.
 To address whether it is the case that an exogenous shock unrelated to open-
 ing up drives the temporary increase in real wage growth, we select a group of 
T
 2?S S 
 R W G D  
C A L
	 E
 Panel A
 Country
 Liberalization real 
wage change
 Liberalization wage 
change relative to  
preliberalization 
wages
 Liberalization wage 
change relative to 
control group
 Difference-in-
 differences wage 
change
  (1) (2) (3) (4)
 Argentina ?2.88% 3.10% 18.70% 16.26%
 Brazil 25.15% 23.78% 38.17% 19.12%
 Chile 20.18% 6.36% 30.98% 1.50%
 Colombia 9.75% 9.25% 22.76% 4.37%
 Greece 9.33% 13.11% 30.19% 33.61%
 India 13.18% 9.79% 23.97% 2.32%
 Indonesia ?4.54% ?14.53% 12.25% ?2.16%
 Israel 16.38% 14.64% 19.93% 5.51%
 Jordan 14.12% 23.63% 13.39% 13.93%
 Malaysia 2.57% ?3.71% 27.02% 22.98%
 Mexico 23.00% 39.07% 39.80% 44.93%
 Morocco 23.68% 19.77% 32.72% 22.49%
 Nigeria ?51.99% 25.87% ?51.68% 0.13%
 Pakistan 3.01% ?10.01% 5.54% ?22.24%
 Philippines 14.15% 6.25% 27.17% 3.22%
 Portugal 45.40% 59.56% 60.26% 66.57%
 South Africa 6.77% 3.06% 17.56% ?4.33%
 South Korea 49.05% 27.45% 69.91% 54.30%
 Spain 6.35% 8.30% 11.01% 3.88%
 Taiwan 42.79% 17.19% 57.64% 34.13%
 Thailand 17.16% -8.00% 35.18% 12.61%
 Turkey 59.46% 65.14% 76.26% 73.59%
 Venezuela ?13.13% 35.15% ?2.35% 17.75%
 Zimbabwe 51.08% 99.55% 60.29% 87.78%
 Mean 15.83% 19.74% 28.19% 21.35%
 Panel B
 Right-hand  
side variable
 Liberalization real 
wage change
 Liberalization wage 
change relative to  
preliberalization 
wages
 Liberalization wage 
change relative to 
control group
 Difference-in-  
differences wage 
change
  (1) (2) (3) (4)
 Constant  0.1583***  0.1915***  0.2687***  0.2015***
  (0.048)  (0.051)  (0.054)  (0.054) 
Observations 25 25 25 25
 Notes: Panel A shows the following summary statistics for each country. Column 1 shows the log real wage change 
from event year [0] to [3] for each country?s liberalization episode. Column 2 presents the liberalization real wage 
change expressed relative to the country?s mean log real wage change from [?3] to [?1]. Column 3 is the liberal-
 ization log real wage change expressed relative to the contemporaneous mean log real wage change for the control 
group countries between years [0] and [3]. Column 4 presents the difference-in-differences log real wage change 
that results from subtracting the difference in the pre-liberalization wage change between the treatment and control 
groups from the quantity in column 3. In panel B, for columns 1?4, we test whether the mean differs signicantly 
from zero and report a heteroscedasticity-consistent estimate of the standard error in parentheses.
  *** Signicant at the 1 percent level.
  ** Signicant at the 5 percent level.
   * Signicant at the 10 percent level.
VOL. 4 NO. 2 115CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 control countries. The control group consists of developing countries that are 
similar to the liberalizing countries except that the control countries did not open 
their stock markets to foreign investment. Appendix A provides a list of countries 
that have stock markets but never liberalized. These nations comprise the control 
group against which we compare the real wage growth of the countries in our 
treatment group.
 A. Descriptive Findings and Data Concerns
 Figure 1 exhibits a steep positive inection after year [0], indicating a sharp 
increase in the growth rate of the real wage. However, with only 25 countries in the 
sample, an important question is whether a few outliers drive the increase.
 The descriptive statistics in Table 2 suggest that this is not the case. Only four 
countries?Indonesia, Malaysia, Pakistan, and Thailand?have mean growth rates in 
the aftermath of liberalization that do not exceed their full sample mean (column 2). 
Turning from means to medians, we also performed simple Wilcoxon signed-rank 
tests on the data for each country in the sample (Wilcoxon 1945). The procedure 
tests the equality of matched pairs of observations by using the Wilcoxon matched-
 pairs, signed-ranks test. Applied to the current context, the null hypothesis is that the 
distributions of pre- and post-liberalization wage growth are the same. In the three-
 year post-liberalization period, ve countries experience median real wage growth 
that falls below the median growth rate of their real wage in the pre-liberalization 
period. Given the null hypothesis, the probability (p-value) of nding no more than 
5 countries with wage growth rates below their pre-liberalization median is 0.0015.
 Table 2 (panel A) also shows the following quantities: (a) for each country, col-
 umn 1 gives the change in the natural log of the wage over the liberalization window 
(years [0] to [3]); and denotes this variable ?liberalization log wage change? as a 
convenient shorthand expression; (b) Column 2 gives the liberalization log wage 
change for each country expressed relative to the country?s mean log wage change 
over the pre-liberalization window (years [?3] to [?1]); (c) Column 3 gives the 
liberalization log wage change for each country expressed relative to the contempo-
 raneous mean log wage change for the control group over the liberalization window 
(years [0] to [3]); (d) Column 4 gives the difference-in-differences log real wage 
change that results from subtracting the difference in the pre-liberalization wage 
change between the treatment and control groups from the quantity in column 3.
 The average cumulative log wage change for the treatment group during the lib-
 eralization window is 15.8 percent (column 1). Relative to the pre-liberalization 
window, the cumulative log wage change is 19.1 percent (column 2). Relative to the 
control group during the liberalization window, the cumulative log wage change is 
26 percent (column 3). The average difference-in-differences estimate of the cumu-
 lative log wage change is 20.2 percent (column 4).
 For each of the quantities in columns 1?4, we test whether the mean differs signif-
 icantly from zero and report a heteroscedasticity-consistent estimate of the standard 
error in Table 2, panel B. These simple tests support the observation that liberaliza-
 tion leads to large effects on the level of real wages, lending credence to the more 
sophisticated regression results that follow.
116 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 To provide additional descriptive evidence, Table 3 presents estimates of the 
impact of liberalization on the growth rate of real wages over time. Panel A of Table 
3 presents estimates of 1 coefcient only, but allowing progressively more periods 
after the reform (from year [2] up to [5]). For example, the coefcient on the vari-
 able DUMMY03 estimates the average effect of liberalization on real wage growth 
in years [0], [1], [2], and [3]. The coefcient estimate of 0.0369 suggests that the 
average real wage growth in that 4 year span is 3.69 percent per year, or an over-
 all increase of 14.76 percentage points (3.69 ? 4). The average estimated effect of 
liberalization on wage growth differs between specications. Consistent with the 
theoretical prediction that liberalization will produce a temporary increase in wage 
growth, the average effect becomes progressively smaller as we add more years sub-
 sequent to the liberalization window. We also estimated the coefcient estimates on 
the liberalization dummy over the [0, 2], [2, 4], and [4, 6] event windows. However, 
the interpretation of the statistical difference of the coefcients across the different 
windows is difcult given their overlapping nature.
 Panel A also presents an estimate of the ?Ashenfelter? dip. The variable 
?Dummy ([?3], [?1]),? which takes on a value of 1 for the 3 years before liber-
 alization, shows that the pre-reform dip in wage growth, which is ?2.37 percent, 
compared to the post-reform average real wage growth of 3.69 percent in the four 
years following liberalization.
 T
 3?T I 
 L
	  R W G  T  D  T
 Panel A
 Time  
window ([?3], [?1]) ([0], [2])  ([0], [3]) ([0], [4]) ([0], [5]) ([0], [6])
 
  (1)  (2)  (3)  (4)  (5)  (6) 
Liberalize
  ?0.0237** 0.0341*** 0.0369*** 0.0314***  0.0190* 0.012
  (0.011)  (0.013)  (0.012)  (0.011)  (0.011) (0.011)
 Constant 0.0154*** 0.006 0.004 0.004 0.006 0.008
 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)
 N 437 437 437 437 437 437
 R2 0.139 0.145 0.15 0.147 0.138 0.135
 Panel B
  Year [0] Year [1] Year [2] Year [3] Year [4] Year [5]
 
  (1)  (2)  (3)  (4)  (5)  (6) 
Liberalize 0.0002 0.0476*** 0.0718*** 0.0316 ?0.0047 ?0.0456
 (0.02) (0.015) (0.021) (0.02) (0.023) (0.032)
 Constant 0.0114** 0.0064 0.0078 0.0099* 0.0116**  0.0134*** 
 (0.01)  (0.01)  (0.01)  (0.01)  (0.01)  (0.01) 
N 437 437 437 437 437 437
 R2 0.132 0.149 0.152 0.136 0.132 0.139
 Notes: Panel A presents coefcient estimates on the liberalization dummy over time or the time prole of the impact 
of liberalization on the growth rate of real wages. Panel A presents estimates of the coefcient on the liberalization 
dummy allowing progressively more periods after the reform (from 2 up to 6) and for the pre-liberalization period 
([?3], [?1]). The left-hand-side variable is the natural log of the real wage change. Panel B presents the coefcient 
estimates on the liberalization dummy by individual year ranging from the liberalization year [0] up to ve years 
following the liberalization [5]. N is the number of observations. Robust standard errors appear in parentheses.
 *** Signicant at the 1 percent level.
  ** Signicant at the 5 percent level.
   * Signicant at the 10 percent level.
VOL. 4 NO. 2 117CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 Panel B presents the coefcient estimates on the liberalization dummy by indi-
 vidual year, ranging from the liberalization year [0] up to ve years following the 
liberalization [5]. Columns 2 and 3 indicate that the impact of liberalization on real 
wages is positive and signicant in both the rst [1] and second [2] years follow-
 ing the liberalization. The coefcient estimates on the liberalization dummy are not 
signicant in years [3], [4], [5]. The estimates in years [1] and [2] are signicantly 
different from the coefcients in the other years.
 Although the numbers in Tables 2 and 3 suggest a reasonably consistent increase 
in real wage growth across countries, two other questions about the data remain.
 Hours Worked.?First, the necessity of using annual data instead of hourly wages 
raises a potential measurement concern. If the average number of annual hours 
worked per employee increases following liberalizations, then total annual compen-
 sation may rise without any change in the implied hourly wage. In other words, the 
rise in average annual labor income (W ? H) documented in Figure 1 could be the 
result of an increase in hours worked rather than an increase in the hourly wage rate. 
To interpret the impact of liberalization on total annual compensation as an increase 
in labor?s compensation per unit of time, we need to know that the average number 
of hours worked does not rise signicantly following liberalizations.
 In an attempt to address this concern we ran into nontrivial constraints that forced 
us to rely on data for a subset of the countries in our sample. UNIDO does not 
provide information on hours worked. Data available from the International Labor 
Organization (ILO) is also not helpful because the ILO?s denition of hours worked 
is inconsistent across countries. Within a given country, the ILO?s denition of hours 
worked sometimes varies across sectors and over time as well. In the end, we used 
data provided by the Groningen Growth and Development Center (GGDC) because 
GGDC seemed to take most seriously the problems associated with trying to con-
 struct a consistent cross-country measure of hours worked. In their own words, the 
GGDC?s estimates of hours worked are based on ? ? a country-by-country ? judg-
 ment of which sources made the most appropriate adjustments to achieve the pre-
 ferred concept of actual hours worked per person employed? (GGDC 2011). The 
GGDC data include paid overtime and exclude paid hours that are not worked due 
to sickness, vacation, and holidays.
 Specically, the data on hours worked come from GGDC?s Total Economy 
Database, which extends the work of Maddison (1980). Although the Total Economy 
Database contains annual numbers on GDP, population, employment, hours, and 
productivity for about 125 countries from 1950 to 2008, the series on hours worked 
per person are available for only 43 countries. Twelve of those 43 countries are also 
in our dataset: Argentina, Brazil, Chile, Colombia, Greece, Korea, Mexico, Portugal, 
Spain, Turkey, Taiwan, and Venezuela.
 To assess whether the rise in average annual compensation is driven by an 
increase in hours worked, Figure 5, which plots the natural log of hours worked 
in liberalization time, illustrates that hours worked are invariant to liberalization. 
Between years [?5] and [?1] the average natural log of hours worked is 7.64. In 
years [1]?[5], the average is 7.62. In levels, these numbers translate to an average of 
2,094 hours worked per year prior to liberalization and 2,066 hours worked per year 
118 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 after liberalization. Dividing 2,094 and 2,066 by the number of weeks in a year (52, 
not adjusting for vacation time) gives an estimate of roughly 40 hours to the average 
work week in these 12 countries before and after opening up. That number seems 
entirely reasonable and reinforces our condence in the GGDC data. Looking at 
medians instead of means does not alter the story. The median natural log of hours 
worked before liberalization is 7.59. The median after liberalization is also 7.59. In 
short, the number of hours worked per year does not change with liberalization and 
does not drive the increase in real wage growth documented in Figure 1.13
 Concurrent Economic Events.? A second concern is that capital market open-
 ings may coincide with major economic events that could have a signicant impact 
on wages, independent of any effects of liberalization. Union activity provides a 
case in point. If agitation for higher pay by organized labor coincides with liber-
 alization, then the estimates may be overstated. A survey of labor market events 
during capital account liberalization episodes revealed that signicant union activ-
 ity to secure higher wages was present in only 3 of the 25 countries in our sample 
(Brazil, Turkey, and South Korea). When these countries are dropped, the cumu-
 lative raw difference-in-differences estimate is not substantively changed at 17.39 
percent. Also, when we control for unionization in the formal regression analysis, 
the coefcient on the capital account liberalization dummy remains unchanged. In 
13 As a nal check, we also used the GGDC data to construct a measure of hourly wages for the subset of 12 
countries. Regressing the change in the natural log of the hourly wage on the same right-hand-side variables that 
appear in equation (6), we nd results that are qualitatively identical to those reported in panel A of Tables 4?6.
 F 5. T A N
 
 H W D  R  L
	
 In
  (hours worked per yea
 r) Subset of 12 liberalizers
 7.4
 7.5
 7.6
 7.7
 7.8
 7.9
 ?5          ?4         ?3        ?2          ?1          0            1           2            3            4            5
 Year relative to liberalization
VOL. 4 NO. 2 119CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 two other countries?Chile and South Africa?the government thwarted protestors 
demanding wage hikes. The remaining 20 countries in the sample did not experience 
any signicant union activity during their liberalization episodes.
 While union activity does not drive the increase in wages, a separate concern 
is that liberalizations often coincide with major economic reforms that could have 
a signicant impact on wages apart from any effects of liberalization. Stabilizing 
ination, removing trade restrictions, and privatizing state-owned enterprises are 
all reforms that may affect real wages through their impact on the efciency of 
domestic production. Indeed, Table 1 demonstrates that the timing of these reforms 
makes it plausible that they, not capital account liberalization, are responsible for 
the increase in real wages apparent in Figure 1. The next section, which presents our 
formal empirical methodology and results, uses the information in Table 1 to control 
directly for the impact of other reforms and to address a host of lingering concerns 
and alternative explanations.
 III. Empirical Methodology and Results
 We evaluate the statistical signicance of the temporary increase in wage growth 
by estimating the following difference-in-differences panel regression specication:
 (6) ?ln  w it dif =  a0  + COUNTR Y i +  a 1 ? LIBERALIZ E it +  a 2  ? TRAD E it 
 +  a 3  ? STABILIZ E it +  a 4 ? PRIVATIZ E it 
 +  a 5 ? BRAD Y it +  ? it .
 The left-hand-side variable, ? ln  w it dif , is the change in the natural log of the real 
local currency value of annual compensation for country i in year t minus the aver-
 age change in the natural log of the real wage for the group of control countries in 
year t. Moving to the right-hand side of equation (6), the variable LIBERALIZ E it 
is a dummy variable that takes on a value of 1 in the year that country i liberalizes 
([0]) and each of the subsequent three years ([1], [2], and [3]). This means that the 
coefcient  a 1 measures the average annual deviation of the growth rate of the real 
wage in the treatment group from the growth rate of the real wage in the control 
group during the three-year liberalization episode.
 The right-hand side of equation (6) also contains four additional country-spe-
 cic dummy variables?STABILIZE, TRADE, PRIVATIZE, and BRADY?that are 
designed to prevent country-specic shocks in the shape of economic reforms from 
articially inating the coefcient on LIBERALIZE. We treat reforms and liberal-
 ization symmetrically, constructing dummy variables that take on the value 1 in the 
year a reform program begins and each of the three subsequent years.
 Turning at last to the error term  ? it , it is important to note that when the residu-
 als are correlated across observations, OLS standard errors can be biased and 
may overestimate or underestimate the true variability of the coefcient estimates. 
Specically, the standard distributional assumptions needed for valid statistical 
inference will not hold in the presence of: (a) correlation of the residuals across 
120 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 countries within a given time period (cross-sectional dependence), or (b) correla-
 tion of the residuals within a given country over time (time-series dependence). 
Point (a) matters because liberalizations often occur at the same time for differ-
 ent countries, possibly inducing correlation in the wage-growth residuals across 
countries at a given point in time. Point (b) matters because it takes time for wages 
to adjust to their new trajectory. For a given country, wage growth may remain 
elevated above its steady-state rate for a number of years in the post-liberalization 
period, thereby inducing serial correlation in the country?s wage-growth residuals. 
To compute accurate standard errors, we employ various clustering procedures 
described below.
 A. Benchmark Estimates: A Difference-in-Differences Specication
 Table 4, column 1 presents the results from our estimate of equation (6), restrict-
 ing the sample to data three years before the liberalization and three years after. We 
cluster the standard errors by year to account for potential cross-country correlation 
in the wage growth residuals. Column 1 of Table 4 shows that after accounting for 
the effects of ination stabilization, trade liberalization, the Brady Plan, and priva-
 tization, the coefcient on LIBERALIZE is 0.031. This implies that relative to the 
control group, the average growth rate of the typical country?s real wage exceeds 
its long-run mean by 3.1 percentage points per year during liberalization episodes.
 The impact of other economic reforms on the growth rate of the real wage is not 
as robust as that of liberalization, but we do nd some signicant effects of ina-
 tion stabilization and privatization when the regression specication includes these 
variables individually.14 The coefcient estimates demonstrate that controlling for 
the other economic reforms that tend to accompany liberalization does not reduce 
the impact of capital account opening on the growth rate of the real wage. This rein-
 forces our condence in the accuracy of the reform dates and the relevance of the 
corresponding dummy variables as controls.
 Column 2 of Table 4 estimates equation (6) using data that spans the entire sample 
to show that the impact of liberalization on real wage growth is robust to potential 
concerns about the length of the pre-liberalization window used for estimation. The 
coefcient on LIBERALIZE is 0.036 and is signicant at the 1 percent level.
 Turning from economic reforms to statistical issues, researchers do not always 
know whether the precise form of the dependence in residuals is time-series or 
cross-sectional in nature. As a way of addressing this concern, Petersen (2009) 
suggests a less parametric estimation approach that clusters on two dimensions 
simultaneously (e.g., country and time). Petersen?s approach uses the following 
estimate of the variance-covariance matrix, which combines the standard errors 
clustered by country with the standard errors clustered by time: VCountry&Time 
= VC + VTime - VWhite.15
 14 The estimate of the coefcient on the liberalization dummy ranges 0.031 to 0.043 and is statistically signi-
 cant at the 1 percent level in nearly every specication that includes the economic reform dummies one-by-one. 
These additional results are available in the online Appendix.
 15 Proposed by Cameron, Gelbach, and Miller (2006a), and Thompson (2006).
VOL. 4 NO. 2 121CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 The rst matrix on the right-hand side allows standard errors to be clustered by coun-
 try, capturing the unspecied correlation between observations on the same country in 
different years (e.g., correlations between ?it and ?is). The second matrix on the right-
 hand side, allows standard errors to be clustered by time, capturing the unspecied cor-
 relation between observations on different countries in the same year (e.g., correlations 
between ?it and ?kt). Since both the country- and time-clustered variance-covariance 
matrices include the diagonal of the variance-covariance matrix, the Petersen procedure 
subtracts off the White variance-covariance matrix to avoid double counting. This method 
allows for both a country and a time effect, although observations on different countries 
in different years are assumed to be uncorrelated. Petersen (2009) demonstrates through 
simulation that clustering by two dimensions produces less biased standard errors.
 Table 4, columns 3 and 4, presents estimates of equation (6) that use Petersen?s 
(2009) procedure to simultaneously cluster the standard errors by year (to adjust for 
cross-country correlation) and by country (to adjust for serial correlation).16 Since 
16 For another discussion of multi-way clustering see Cameron, Gelbach, and Miller (2006b).
 T
 4? L
	 T I  G R 
 R W  P
 Dependent 
variable Real wage (log-difference 
relative to control group)
 Real wage (log-difference 
relative to control group)
 Real wage 
(log- 
difference)
 Real value-added per 
worker (log-difference  
relative to control group)
 Time  
window ([?3], [+3]) Full sample ([?3], [+3]) Full sample Full sample ([?3], [+3]) ([?3], [+3])
 
  (1)  (2)  (3)  (4)  (5)  (6)  (7) 
Liberalize 0.0314*** 0.0362** 0.0394** 0.0474*** 0.0388** 0.0972*** 0.1026***
 (0.011) (0.014) (0.015) (0.015) (0.015) (0.034) (0.016)
 Trade 0.0051 0.0172 0.0211 0.0197 0.0170 0.0509 0.0694***
 (0.021) (0.014) (0.021) (0.015) (0.014) (0.045) (0.024)
 Stabilize 0.0228 ?0.0303 ?0.0288 ?0.0289 ?0.0278 ?0.0577 ?0.0622
 (0.023) (0.02) (0.022) (0.023) (0.017) (0.099) (0.067)
 Privatize 0.0285 0.0165 0.0298 0.0100 0.0055 ?0.0603 ?0.0430
 (0.023) (0.016) (0.027) (0.016) (0.019) (0.049) (0.041)
 Brady 0.0405 0.0168 ?0.0265** ?0.0226 0.0155 ?0.0011 ?0.0676
 (0.031) (0.025) (0.012) (0.025) (0.023) (0.116) (0.09)
 Constant 0.0104 0.0211*** 0.0196 0.0215*** 0.0180*** ?0.0332 ?0.0361
 (0.014) (0.006) (0.016) (0.008) (0.006) (0.09) (0.071)
 Standard errors
 Clustered
 (country)
 Clustered
 (country and year) Robust
 Clustered 
(year)
 Clustered 
(country  
and year)
 Country xed effects Yes Yes No No Yes Yes No
 Year xed effects No No No No Yes No No
 Observations 152 722 152 722 758 146 146
 R2 0.38 0.12 0.092 0.03 0.1730 0.19 0.07
 Notes: The estimation procedure is ordinary least squares. LIBERALIZE is a dummy variable that takes on the value 
of 1 in the year that country i liberalizes (year [0]) and each of the subsequent three years ([1], [2], and [3]). TRADE, 
STABILIZE, PRIVATIZE, and BRADY are dummy variables that take on the value 1 whenever a trade liberalization, 
ination stabilization, privatization, or Brady plan program takes place during country i?s capital account liberaliza-
 tion episode. ([?3],[+3]) denotes a time window of 3 years prior to and following the liberalization year. Standard 
errors appear in parentheses.
 *** Signicant at the 1 percent level.
  ** Signicant at the 5 percent level.
   * Signicant at the 10 percent level.
122 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 it is not possible to include country xed effects while simultaneously clustering 
the standard errors by year and country, the magnitudes of the estimates in columns 
3 and 4 are not identical to those in columns 1 and 2, but they are very similar. 
Focusing, then, on the precision of the estimates, we see that the coefcient estimate 
on LIBERALIZE (in columns 3 and 4) is signicant at the 5 percent and 1 percent 
condence levels, respectively. This suggests that our general nding is robust to 
concerns about both serial and cross-country correlation in the error terms.
 An Alternative Specication: Country and Year Fixed Effects.?As an alternative 
to our difference-in-differences estimates, we regress the change in the natural log 
of the real wage or a full set of country xed effects and year xed effects plus the 
reform dummies. Specically, we estimate
 (7) ?ln wit =  a 0 + COUNTRYi + YEA R t +  a 1 ? LIBERALIZ E it 
 +  a 2 ? TRAD E it +  a 3 ? STABILIZ E it +  a 4 ? PRIVATIZ E it 
 +  a 5 ? BRAD Y it +  ? it .
 The left-hand-side variable, ?ln w it  , is the natural log of the real local currency 
value of annual compensation for country i in the treatment group in year t minus the 
same variable in year t?1. The variable YEARt is shorthand for the set of year-xed 
effects. The right-hand side of equation (7) also contains four additional dummy 
variables?STABILIZE, TRADE, PRIVATIZE, and BRADY. The standard errors are 
clustered by year.
 Table 4, column 5 presents the results from our estimate of equation (7). The 
standard errors are clustered by year to account for potential cross-country correla-
 tion in the wage growth residuals. The impact of liberalization on real wage growth 
is economically large. Controlling for the effects of ination stabilization, trade lib-
 eralization, the Brady Plan, and privatization, the coefcient on LIBERALIZE is 
0.0388 and signicant at the 5 percent level.17 This means that during liberalization 
episodes, the average growth rate of the typical country?s real wage exceeds its long-
 run mean by 3.88 percentage points per year. Accounting for the other economic 
reforms that tend to accompany liberalization also does not affect the impact of 
capital account opening on the growth rate of the real wage.
 B. Alternative Explanations
 One interpretation of the evidence says that wages rise following liberalizations 
because of an increase in labor demand stemming from a capital-deepening induced 
rise in productivity. Alternatively, the increase in wages may be due to a reduction 
in labor supply. The argument runs as follows. If workers perceive the impact of 
17 The estimate of the coefcient on LIBERALIZE ranges from 0.038 to 0.043 in additional regressions that 
include the reform dummies individually. The coefcients on the liberalization dummy are statistically signicant 
at the 1 percent level. These results are available in the online Appendix.
VOL. 4 NO. 2 123CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 liberalization on wages to be permanent, then they effectively receive a positive 
shock to their permanent income and may reduce their labor supply accordingly. If 
this is the case then the observed increase in wage growth may stem from a decrease 
in labor supply as well as an increase in labor demand.18
 The employment data are not consistent with a decrease in labor supply. There is 
no discernible change in the growth rate of employment following liberalizations. 
The regression specications in Table 5 examine the change in the natural log of 
employment on the same right-hand-side variables that appear in equation (7). The 
liberalization dummy is never signicant. Also, if labor supply decreases, then we 
would also expect a decline in the number of hours worked. Again, Figure 4 demon-
 strates that this is not the case. Overall, the evidence does not suggest that workers 
reduce their labor supply in response to liberalization. While we do not formally 
estimate the labor supply decision and cannot exclude the possibility that part of the 
wage increase results from a decrease in labor supply, if this alternative explanation 
is at work, its overall impact appears to be second order.
 C. Economic Interpretation
 There are two ways to examine the economic signicance of the results. First, 
consider the magnitude of the growth rate of the real wage during liberalization epi-
 sodes relative to the growth rate of the real wage over the entire sample. To do this, 
use the estimate of the constant and the liberalization dummy from the regression 
that controls for other economic reforms (column 5 in Table 4). Real wages grow 
by an average of 1.8 percent per year over the entire pre-liberalization sample. The 
estimate of the coefcient on the liberalization dummy is 0.039. Adding the average 
pre-liberalization real wage growth to the coefcient on the liberalization dummy 
gives the average growth rate of the real wage during liberalization episodes?5.7 
percent per year. This means that in the year the liberalization occurs and each of 
the subsequent three, the average growth rate of the real wage is almost three times 
as large as in nonliberalization years (5.7 versus 1.8).
 Of course, the increase in the growth rate of the real wage is temporary, so a 
second way of assessing economic signicance is to compute the impact of liberal-
 ization on the permanent level of the real wage. For the countries in the treatment 
group, the average level of annual compensation in the year before liberalization 
(year [?1]) is $2,686 PPP-adjusted dollars. During the three-year liberalization 
window the real wage grows at 5.7 percent per year, so that by the end of year [3] 
the average level of the real wage is 2,686e0.057?4 = $3,374 PPP-adjusted dollars. 
Now, assume that in the absence of liberalization the real wage would have grown 
at the sample mean of 1.8 percent per year. In that case, the level of the real wage at 
the end of year [3] would be 2,686e0.018?4 = $2,887 PPP-adjusted dollars. In other 
words, by the time the impact of liberalization has run its course, the average worker 
in the manufacturing sector has annual take-home pay that is $487 dollars higher 
18 An alternative view is that labor supply is relatively inelastic (see Pencavel 1986 on this point). If this is 
the case, workers may not reduce the number of hours that they want to work in response to the increase in their 
expected future income.
124 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 ($3,374 minus $2,887) than it would have been in the absence of liberalization. This 
change in the level of the wage is equal to nearly one-fth of the average manufac-
 turing worker?s pre-liberalization, PPP-adjusted take-home pay.
 It is also important to note that the results are not an artifact of mean reversion 
following a temporary fall in earnings ? la Ashenfelter (1978). Figure 1 shows a 
decline in the level of the real wage from an average of $2,968 PPP-adjusted dol-
 lars 5 years prior to liberalization to an average of $2,686 PPP-adjusted dollars 1 
year prior to liberalization. A few hypothetical calculations demonstrate that the 
results do not simply reect a bounce-back effect. Suppose that instead of declin-
 ing, on average, for the next four years (as they do in the data), real wages grew 
at their (continuously compounded) long-run rate of 1.8 percent per year over the 
pre-liberalization window. Under that scenario, the real wage in year [3] would have 
been $3,485 PPP-adjusted dollars. The actual level of the average real wage in year 
[3] is $4,050 PPP-adjusted dollars, roughly 20 percent higher than the level that 
would have prevailed had wages simply continued to grow at their long-run rate.
 D. The Impact of Liberalization on Productivity
 The results suggest that the response of wages to capital account liberalization 
is large. To scrutinize the plausibility of our estimates we cross-checked the results 
against data on labor productivity. The model in Section I demonstrates that liber-
 alization induces capital deepening, and through the increase in capital per worker, 
drives up productivity, the demand for labor, and the real wage. If this chain of 
T
 5?L
	 D N S
 I E
  (1)  (2)  (3)  (4)  (5)  (6) 
Liberalize ?0.015 ?0.0141 ?0.0154 ?0.0189 ?0.0061 ?0.0106
 (0.021) (0.022) (0.021) (0.023) (0.02) (0.021)
 Trade ?0.0159 ?0.0101
 (0.083) (0.089)
 Stabilize ?0.0148 0.0106
 (0.019) (0.028)
 Privatize 0.0115 0.0173
 (0.02) (0.027)
 Brady ?0.1201 ?0.1260
 (0.079) (0.087)
 Constant ?0.0443** ?0.0421** ?0.0448** ?0.0456** ?0.0211 ?0.0203
 (0.019) (0.021) (0.019) (0.019) (0.027) (0.03)
 Observations 783 783 783 783 783 783
 R2 0.12 0.12 0.12 0.12 0.122 0.122
 Notes: The estimation procedure is ordinary least squares. All specications contain year-specic and country-
 specic dummy variables. For the regressions reported, the left-hand-side variable is the change in the natural log 
of employment over the full sample. LIBERALIZE is a dummy variable that takes on the value of 1 in the year 
that country i liberalizes (year [0]) and each of the subsequent three years ([1], [2], and [3]). TRADE, STABILIZE, 
PRIVATIZE, and BRADY are dummy variables that take on the value 1 whenever a trade liberalization, ination sta-
 bilization, privatization or Brady Plan program takes place during country i?s capital account liberalization episode. 
Standard errors appear in parentheses.
 *** Signicant at the 1 percent level.
  ** Signicant at the 5 percent level.
   * Signicant at the 10 percent level.
VOL. 4 NO. 2 125CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 reasoning has any empirical bite, then, during liberalization episodes, labor produc-
 tivity should rise in concert with wages.
 To formally test the relation between liberalization and the growth rate of labor 
productivity, we estimate the following difference-in-differences regression:
 (8) ?ln( Y _ L) it dif = a 0 + COUNTR Y i +  a 1 ? LIBERALIZ E it +  a 2 ? TRAD E it 
 +  a 3  ? STABILIZ E it +  a 4  ? PRIVATIZ E it 
 +  a 5 ? BRAD Y it +  ? it .
 Equation (8) is identical to equation (6) except that instead of the change in the 
natural log of the annual wage, the left-hand-side variable is now the change in the 
natural log of the annual real value added per worker in local currency terms minus 
the average change in the natural log of real value added per worker in the con-
 trol group. Again, to be consistent with the wage estimates, we cluster the standard 
errors by year.
 Table 4, column 6 shows that liberalization has a positive and signicant impact 
on productivity growth. Accounting for the potential impact of other economic 
reforms, the estimate of the coefcient on the liberalization dummy is 0.0972 when 
the sample is restricted to the three years prior to and three years following the lib-
 eralization year.19 This means that the average growth rate of productivity is 9.72 
percentage points higher during the three-year liberalization window than in non-
 liberalization years. The 9.72 percentage point increase in productivity growth asso-
 ciated with liberalization is larger than the 3.9 percentage point increase in wage 
growth. Because the increase in productivity outstrips the increase in wage growth, 
manufacturing sector protability actually rises during liberalizations.20 Column 7 
of Table 4 shows that the results for productivity growth and liberalization are robust 
to all of the statistical concerns raised about the estimation of wage growth and lib-
 eralization examined in Subsection IIIA.
 IV. Discussion
 While the size of the increase in productivity growth more than matches the size 
of the increase in real wage growth, the important unanswered question is whether 
the magnitude of either increase is consistent with the model that drives the esti-
 mates. To answer the question, begin with the standard assumption that liberal-
 ization has no impact on total factor prodcutivity growth and recall equation (5): 
( ? w /w) = ( ? A /A) + (1/?) ? (  f ? (k)k)/(f (k)) ? ( ? k/k). With no change in the 
growth rate of total factor productivity, equation (5) implies that the change in the 
growth rate of the real wage equals the product of three numbers: the reciprocal of 
19 In the online Appendix, every estimate of the coefcient on the liberalization dummy is statistically signi-
 cant in regression specications that include the reform dummies individually. These results suggest that liberaliza-
 tion, not an external shock or domestic economic reforms, is responsible for the increase in productivity growth.
 20 Chari and Henry (2008) also nd that the return to capital in the manufacturing sector rises during 
liberalizations.
126 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 the elasticity of substitution, capital?s share in national income, and the change in 
the growth rate of capital per effective worker.
 Specically, we have
  (9)   ? w  _ w =  1 _ ? ?  
 f ? (k)k
  
_(k)  ?  
 
?
  k
  
_
  k
 .
 The capital share typically lies between 1/3 and 1/2 .21 Obtaining an estimate of the 
change in capital growth requires a little more effort. We know from previous work 
that aggregate capital stock growth increases by 1.1 percentage points following 
liberalizations (Henry 2003). We can use the aggregate number to calculate a rough 
upper bound for the change in manufacturing sector capital growth. For the countries 
in our sample, the manufacturing sector accounts for about 1?5 of GDP. Assuming 
zero net growth in capital for the agriculture and service sectors, the largest pos-
 sible increase in the growth rate of capital in manufacturing is about 5.5 percentage 
points. This back-of-the-envelope calculation nds empirical support elsewhere in 
the literature. Using a subset of the countries in this paper, Chari and Henry (2008) 
calculate that the growth rate of capital in the manufacturing sector increases by 
4.1 percentage points per year following liberalizations. Increases in capital stock 
growth between 4.1 and 5.5 percentage points are also consistent with the size of the 
fall in the cost of capital that occurs following liberalizations.22
 Suppose the capital share is 1/2 and the change in capital growth is 5.5 percentage 
points. Then for capital-deepening alone to explain the 3.9 percentage point increase 
in wage growth you need an elasticity of substitution less than or equal to 0.7. If 
the capital share is 1/3 and the change in capital growth is 4.1 percentage points, 
then the elasticity of substitution must be less than or equal to 0.35. There is little 
consensus on the size of the elasticity of substitution. Early work estimated small 
elasticities that were statistically indistinguishable from zero.23 More recent studies 
cannot reject the hypothesis that the elasticity of substitution is 1 (Caballero 1994). 
Most relevant to the countries in this paper, Coulibaly and Millar (2007) estimate 
an elasticity of substitution of about 0.8 for South Africa. With a standard error of 
0.08, the Coulibaly and Millar estimate could imply an elasticity as small as 0.64. 
With a change in capital growth of 5.5 percentage points, a capital share of ?, and 
an elasticity of substitution of 0.64, equation (9) predicts that liberalization would 
generate a 4.3 percentage point increase in wage growth.
 We do not mean to push any particular value for the capital share. And it is not 
clear that we have a consensus estimate of the size of the elasticity of substitution in 
developed countries, let alone emerging economies. What is clear, however, is that 
if you want to maintain the assumption that liberalization has no impact on total 
factor productivity, then the observed increases in wage growth are consistent with 
the model only if you are willing to concede that the elasticity of substitution is sub-
 stantially less than one (i.e., the world is not Cobb-Douglas). But if the elasticity of 
21 A few studies nd capital shares in developing countries as high as two-thirds. (See, for example, Rodr?guez 
and Ortega 2006). On the other hand, Gollin (2002) documents capital shares closer to a third.
 22 See Henry (2007, 897?900).
 23 For a survey of this literature see Chirinko (1993).
VOL. 4 NO. 2 127CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 substitution is signicantly less than one, then it is hard to understand how capital?s 
income share remains constant (or increases) following the liberalization-induced 
fall in the cost of capital.
 On the other hand, if you maintain that the world is Cobb-Douglas, then our wage 
results imply that liberalization has an impact on total factor productivity. In a Cobb-
 Douglas world with a capital share of one half, equation (9) implies an increase in 
real wage growth of 2.75 percentage points?leaving a gap of roughly 1 percentage 
point to be explained. A gap of 1 percentage point per year would require a 4 percent 
increase in the level of factor productivity over 4 years. If opening up brings about 
changes that raise an economy?s permanent level of efciency by 4 percent, then 
this will show up as a temporary increase in total factor productivity growth during 
the transition.24 One can imagine a number of channels that lie outside the connes 
of the Solow model through which liberalization raises efciency by this magnitude 
over the course of a 4-year window.
 For instance, liberalization may enable rms to import more efcient machines 
(e.g., tractors instead of hoes) that effectively shift the country?s production tech-
 nology closer to the world frontier. DeLong (2004) argues that after liberalizing the 
capital account ? ? developing countries ? would enjoy the benets from technol-
 ogy advances and from learning-by-doing using modern machinery.? To the extent 
that technological progress diffuses from developed to developing countries, the 
importation of new machinery provides an important conduit through which the dif-
 fusion may occur (Eaton and Kortum 2001a). Almost all of the world?s research and 
development (R&D) takes place in a small number of industrial countries (Eaton 
and Kortum 1999), and the same group of countries accounts for over 70 percent of 
the world?s machine exports in a given year (Eaton and Kortum 2001b; Alfaro and 
Hammel 2007).
 Evidence from the literature supports the conjecture that developing countries can 
import technological progress by liberalizing the capital account. In the immediate 
wake of liberalizations, rms in the manufacturing sector of developing countries 
accumulate capital at a faster rate than they did before the liberalization (Chari and 
Henry 2008). Furthermore, these countries raise their rate of capital accumulation 
by importing more capital goods. As a result of liberalization, the share of capital 
goods imports to total imports rises by 9 percent, and the share of total machine 
imports as a fraction of GDP rises by 13 percent (Alfaro and Hammel 2007).
 The observation that both imports of capital goods and total factor productiv-
 ity rise in concert with liberalization lends credence to the notion that new capital 
goods embody technological progress and that developing countries can raise their 
growth rates of total factor productivity by liberalizing the capital account.25 The 
observation is also consistent with work showing that cross-country variation in the 
24 While endogenous growth models such as Romer (1986) certainly allow for the possibility of permanent 
growth effects due to innovation, this view of the world strikes us as most relevant for countries that actually do 
the innovating. Emerging economies are more likely to be adopters of technology, engaged in a periodic process of 
capital upgrading and technological catch up that bears closer resemblance to periodic jumps in the level of total 
factor productivity.
 25 Also, in the spirit of Rajan and Zingales (1998), liberalization may improve domestic rms? access to external 
nance, which might, in turn, increase the rate at which rms import capital goods.
128 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 composition of capital investment explains much of the cross-country variation in 
total factor productivity (Caselli and Wilson 2004).
 On the other hand, some argue that the simplest explanation of capital-account-
 liberalization-induced total factor productivity growth lies with the economic 
reforms that accompany liberalization. Economic reforms improve resource alloca-
 tion, essentially producing a one-time shift in the production function that temporar-
 ily raises the growth rate of total factor productivity, without inducing technological 
progress per se (Henry 2003). Others posit that liberalization facilitates increased 
risk sharing, which might encourage investment in riskier, higher-growth technolo-
 gies (Levine 1997; Levine and Zervos 1998a). Yet another explanation is that capital 
account liberalization generates unspecied ?collateral benets? that increase pro-
 ductivity (Kose et al. 2006).
 Sorting through competing explanations for the increase in total factor produc-
 tivity following liberalizations is an important research challenge that lies beyond 
the scope of this paper. The bottom line of the discussion here is that the size of the 
increases in wage growth and productivity we report are consistent with the model 
that drives our estimation. Whether the primary source of those increases lies with 
capital deepening or increased total factor productivity depends on reasonable dif-
 ferences in views about the elasticity of substitution that have yet to be resolved in 
the literature.
 V. Conclusion
 In the process of debating the impact of trade on wages, international economists 
pay relatively little attention to the impact of trade in capital. Debating the costs and 
benets of capital market liberalization on economic growth, macro and nancial 
economists largely ignore the implications of increased capital market integration 
for wages. Yet labor income typically accounts for about two-thirds of GDP. Almost 
two decades after the advent of capital account liberalization in the developing 
world, our paper provides the rst systematic analysis of the impact of liberalization 
on the level of real wages.
 Increased capital market integration in the 1980s and 1990s sharply reduced the 
cost of capital for manufacturing rms in emerging economies. In response to the 
fall in their cost of capital, these rms installed new machinery, much of which was 
imported from abroad, and may have embodied substantial technological progress. 
The combination of capital deepening and technological progress drove up the pro-
 ductivity of workers in the manufacturing sector. Accordingly, the demand for those 
workers increased, along with their real wage.26
 While the focus of this paper is on the level of real wages, our ndings also pro-
 vide important clues about the rise in wage inequality in developing countries doc-
 umented by Goldberg and Pavcnik (2007). If the liberalization-induced increase in 
wages was evenly distributed across skilled and unskilled workers in the manufactur-
 ing sector, then no increase in the skill premium would have occurred. However, two 
26 Cragg and Epelbaum (1996) and Behrman, Birdsall, and Szekely (2000) make a similar argument for Latin 
America.
VOL. 4 NO. 2 129CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 observations suggest that the increase in manufacturing sector wages was probably 
concentrated among highly skilled workers. First, countries? imports of machinery 
and equipment rise substantially in the aftermath of capital account liberalizations, 
and rms that import machinery and equipment generally employ a larger share of 
high-skilled workers than rms that do not import such capital (Harrison and Hanson 
1999). Second, work that is characterized as unskilled from a developed country?s 
perspective may be skilled-labor intensive when compared with typical domestic pro-
 duction activities in a developing country (Feenstra and Hanson 1997, 2003).
 While it seems plausible to consider that capital market liberalization contributed to 
increased wage inequality in developing countries during the 1980s and 1990s, it is no 
more than a conjecture because the economic model we employ makes no distinction 
between skilled and unskilled workers. Moving from conjecture to testable predic-
 tion would require a model with skilled labor, unskilled labor, and capital as three 
distinct factors of production. In a three-factor model, capital deepening could exacer-
 bate wage inequality through a form of skill-biased technological change if capital is 
more substitutable for unskilled workers and more complementary to skilled workers. 
However, since we do not use such a model, and our data do not provide information 
about the skill composition of the labor force, we cannot test this prediction.
 Be that as it may, the bottom line of this paper is that increased capital market 
integration has raised the average standard of living for a signicant fraction of the 
workforce in developing countries. If labor is mobile across sectors, then over time 
we would expect the productivity-driven wage gains in manufacturing to translate 
into higher incomes for workers elsewhere in the economy. The extent to which the 
labor market in these countries functions well enough to allow workers to respond to 
wage differentials across sectors is an important issue that lies beyond the scope of 
this paper.27 As the quality and breadth of data on labor markets in developing coun-
 tries continues to improve, future work may produce more denitive conclusions.
 A
 A
 Countries that had not liberalized as of 1997: Algeria, Bangladesh, Barbados, 
Benin, Burkina Faso, Cameroon, Central African Republic, Chad, Congo, Costa 
Rica, Cote D?Ivoire, Dominican Republic, Ecuador, El Salvador, Fiji, Gabon, 
Gambia, Ghana, Guatemala, Guyana, Haiti, Honduras, Iceland, Iran, Jamaica, 
Kenya, Kuwait, Madagascar, Malawi, Mali, Malta, Mauritius, Nepal, Nicaragua, 
Niger, Oman, Paraguay, Peru, Rwanda, Saudi Arabia, Senegal, Sierra Leone, Syrian 
Arab Republic, Togo, Trinidad and Tobago, Tunisia, Uruguay, Zambia.
 Countries that liberalized before 1980: Australia, Austria, Denmark, Finland, 
Ireland, Norway.
 Countries that liberalized between 1980 and 1997: Argentina, Brazil, Chile, 
Colombia, Egypt, Greece, India, Indonesia, Israel, Jordan, Malaysia, Mexico, 
Morocco, New Zealand, Nigeria, Pakistan, Philippines, Portugal, South Africa, 
Spain, Sri Lanka, Thailand, Turkey, Venezuela, Zimbabwe.
 27 See Wacziarg and Seddon Wallack (2004) for an analysis of intersectoral mobility of labor in response to 
trade reforms.
130 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 REFERENCES
 Aitken, Brian, Ann Harrison, and Robert E. Lipsey. 1996. ?Wages and Foreign Ownership: A Com-
 parative Study of Mexico, Venezuela, and the United States.? Journal of International Economics 
40 (3?4): 345?71.
 Alfaro, Laura, and Eliza Hammel. 2007. ?Capital Flows and Capital Goods.? Journal of International 
Economics 72 (1): 128?50.
 Almeida, Rita. 2007. ?The Labor Market Effects of Foreign Owned Firms.? Journal of International 
Economics, 72(1): 75?96.
 Ashenfelter, Orley C. 1978. ?Estimating the Effect of Training Programs on Earnings.? Review of Eco-
 nomics and Statistics 6 (1): 47?57.
 Barro, Robert J., and Xavier Sala-i-Martin. 1995. Economic Growth. New York: McGraw Hill.
 Behrman, Jere R., Nancy Birdsall, and Miguel Sz?kely. 2000. ?Economic Reform and Wage Differen-
 tials in Latin America.? Inter-American Development Bank Working Paper 435. 
Bekaert, Geert, Campbell R. Harvey, and Christian Lundblad. 2001. ?Emerging Equity Markets and 
Economic Development.? Journal of Development Economics 66 (2): 465?504.
 Bertrand, Marianne, Esther Duo, and Sendhil Mullainathan. 2004. ?How Much Should We Trust 
Differences-in-Differences Estimates?? Quarterly Journal of Economics 119 (1): 249?75.
 Borjas, George J., Richard B. Freeman, and Lawrence F. Katz. 1997. ?How Much Do Immigration 
and Trade Affect Labor Market Outcomes?? Brookings Papers on Economic Activity 28(1): 1?90.
 Caballero, Ricardo J. 1994. ?Small Sample Bias and Adjustment Costs.? Review of Economics and 
Statistics 76 (1): 52?58.
 Cameron, A. Colin, Jonah B. Gelbach, and Douglas L. Miller. 2006a. ?Bootstrap-Based Improvements 
for Inference with Clustered Errors.? University of California Davis Department of Economics 
Working Paper 06-21.
 Cameron, A. Colin, Jonah B. Gelbach, and Douglas L. Miller. 2006. ?Robust Inference with Multi-way 
Clustering.? National Bureau of Economic Research Technical Working Paper 327.
 Card, David. 2009. ?Immigration and Inequality.? National Bureau of Economic Research Working 
Paper 14683.
 Caselli, Francesco, and Daniel J. Wilson. 2004. ?Importing Technology.? Journal of Monetary Eco-
 nomics 51 (1): 1?32.
 Chari, Anusha, and Peter Blair Henry. 2008. ?Firm-Specic Information and the Efciency of Invest-
 ment.? Journal of Financial Economics 87 (3): 636?55. 
Chari, Anusha, Paige Parker Ouimet, and Linda L. Tesar. 2004. ?Cross Border Mergers and Acquisi-
 tions in Emerging Markets: The Stock Market Valuation of Corporate Control.? EFA 2004 Maas-
 tricht Meetings Paper 3479.
 Chari, Anusha, Peter Blair Henry, and Diego Sasson. 2012. ?Capital Market Integration and Wages: 
Dataset.? American Economic Journal: Macroeconomics http://dx.doi.org/10.1257/mac.4.2.102.
 Chirinko, Robert S. 1993. ?Business Fixed Investment Spending: Modeling Strategies, Empirical 
Results, and Policy Implications.? Journal of Economic Literature 31 (4): 1875?1911.
 Cline, William R. 1997. Trade and Income Distribution. Washington, DC: Institute for International 
Economics.
 Coulibaly, Brahima, and Jonathan Millar. 2007. ?Estimating the Long-Run User Cost Elasticity for a 
Small Open Economy: Evidence Using Data from South Africa.? Board of Governors of the Fed-
 eral Reserve System, Finance and Economics Discussion Series 2007-25.
 Cragg, Michael Ian, and Mario Epelbaum. 1996. ?Why Has Wage Dispersion Grown in Mexico? Is It 
the Incidence of Reforms or the Growing Demand for Skills?? Journal of Development Econom-
 ics 51 (1): 99?116.
 DeLong, J. Bradford. 2004. ?Should We Still Support Untrammelled International Capital Mobility? 
Or Are Capital Controls Less Evil than We Once Believed?? Economists? Voice 1 (1): 1?7.
 Eaton, Jonathan, and Samuel Kortum. 1999. ?International Technology Diffusion: Theory and Mea-
 surement.? International Economic Review 40 (3): 537?70.
 Eaton, Jonathan, and Samuel Kortum. 2001a. ?Technology, Trade, and Growth: A Unied Frame-
 work.? European Economic Review 45 (4?6): 742?55.
 Eaton, Jonathan, and Samuel Kortum. 2001b. ?Trade in Capital Goods.? European Economic Review 
45 (7): 1195?1235.
 Eichengreen, Barry. 2001. ?Capital Account Liberalization: What Do Cross-Country Studies Tell Us?? 
World Bank Economic Review 15 (3): 341?65.
 Feenstra, Robert C., and Gordon H. Hanson. 1997. ?Foreign Direct Investment and Relative Wages: 
Evidence from Mexico?s Maquiladoras.? Journal of International Economics 42 (3?4): 371?93.
VOL. 4 NO. 2 131CHARI ET AL.: CAPITAL MARKET INTEGRATION AND WAGES
 Feenstra, Robert C., and Gordon H. Hanson. 2003. ?Global Production Sharing and Rising Inequality: 
A Survey of Trade and Wages.? In Handbook of International Trade, edited by E. Kwan Choi and 
James Harrigan, 146?185. Malden, MA: Blackwell. 
Goldberg, Pinelopi Koujianou, and Nina Pavcnik. 2007. ?Distributional Effects of Globalization in 
Developing Countries.? Journal of Economic Literature 45 (1): 39?82.
 Gollin, Douglas. 2002. ?Getting Income Shares Right.? Journal of Political Economy 110 (2): 458?74.
 Gourinchas, Pierre-Olivier, and Olivier Jeanne. 2006. ?The Elusive Gains from International Financial 
Integration.? Review of Economic Studies 73 (3): 715?41.
 Gozzi, Juan Carlos, Ross Levine, and Sergio L. Schmukler. 2010. ?Patterns of International Capital 
Raisings.? Journal of International Economics 80 (1): 45?57.
 Groningen Growth and Development Centre (GGDC). 2011. ?Total Economy Database.? http://www.
 ggdc.net/databases/ted.htm (accessed April 10, 2011).
 Hale, Galina, and Cheryl Long. 2008. ?Did Foreign Direct Investment Put an Upward Pressure on 
Wages in China?? Federal Reserve Bank of San Francisco Working Paper 2006-25.
 Harrison, Ann, and Gordon Hanson. 1999. ?Who Gains From Trade Reform? Some Remaining Puz-
 zles.? Journal of Development Economics 59 (1): 125?54.
 Henry, Peter Blair. 2002. ?Is Disination Good for the Stock Market?? Journal of Finance 57 (4): 
1617?48.
 Henry, Peter Blair. 2003. ?Capital-Account Liberalization, the Cost of Capital, and Economic Growth.? 
American Economic Review 93 (2): 91?96.
 Henry, Peter Blair. 2007. ?Capital Account Liberalization: Theory, Evidence, and Speculation.? Jour-
 nal of Economic Literature 45 (4): 887?935.
 Karolyi, G. Andrew. 2004. ?The Role of American Depositary Receipts in the Development of Emerg-
 ing Equity Markets.? Review of Economics and Statistics 86 (3): 670?90.
 Kose, M. Ayhan, Eswar Prasad, Kenneth S. Rogoff, and Shang-Jin Wei. 2006. ?Financial Globaliza-
 tion: A Reappraisal.? National Bureau of Economic Research Working Paper 12484.
 Krugman, Paul R. 1995. ?Growing World Trade: Causes and Consequences.? Brookings Papers on 
Economic Activity 26 (1): 327?62.
 Lawrence, Robert Z. 2008. Blue-Collar Blues: Is Trade to Blame for Rising US Income Inequality? 
Washington, DC: Petersen Institute for International Economics.
 Lawrence, Robert Z., and Matthew J. Slaughter. 1993. ?International Trade and American Wages 
in the 1980s: Giant Sucking Sound or Small Hiccup?? Brookings Papers on Economic Activity: 
Microeconomics (2): 161?210.
 Levine, Ross. 1997. ?Financial Development and Economic Growth: Views and Agenda.? Journal of 
Economic Literature 35 (2): 688?726.
 Levine, Ross, and Sara Zervos. 1998a. ?Stock Markets, Banks, and Economic Growth.? American Eco-
 nomic Review 88 (3): 537?58.
 Maddison, Angus. 1980. ?Monitoring the Labour Market: A Proposal for a Comprehensive Approach 
in Ofcial Statistics (Illustrated by Recent Developments in France, Germany and the U.K.).? 
Review of Income and Wealth 26 (2): 175?217.
 Obstfeld, Maurice. 2009. ?International Finance and Growth in Developing Countries: What Have We 
Learned?? National Bureau of Economic Research Working Paper 14691. 
Organisation for Economic Co-operation and Development (OECD). 2008. OECD Benchmark De-
 nition of Foreign Direct Investment. 4th ed. Paris: OECD Publishing. http://www.oecd.org/datao-
 ecd/26/50/40193734.pdf. 
Ottaviano, Gianmarco I. P., and Giovanni Peri. 2008. ?Immigration and Wages: Clarifying the Theory 
and the Empirics.? National Bureau of Economic Research Working Paper 14188.
 Park, Keith K. H., and Antoine W. Van Agtmael. 1993. The World?s Emerging Stock Markets: Struc-
 ture, Development, Regulations and Opportunities. Chicago: Probus Publishing Company.
 Pencavel, John H. 1986. ?Labor Supply of Men: A Survey.? In Handbook of Labor Economics. Vol. 1, 
edited by Orley Ashenfelter and Richard Layard, 3?102. New York: Elsevier Science.
 Petersen, Mitchell A. 2009. ?Estimating Standard Errors in Finance Panel Data Sets: Comparing 
Approaches.? Review of Financial Studies 22 (1): 435?80.
 Price, Margaret M. 1994. Emerging Stock Markets. New York: McGraw Hill.
 Rajan, Raghuram G., and Luigi Zingales. 1998. ?Financial Dependence and Growth.? American Eco-
 nomic Review 88 (3): 559?86.
 Rodr?guez, Francisco, and Daniel Ortega. 2006. ?Are Capital Shares Higher in Poor Countries? Evidence 
from Industrial Surveys.? Wesleyan University Department of Economics Working Paper 2006-023. 
Rogoff, Kenneth. 1999. ?International Institutions for Reducing Global Financial Instability.? Journal 
of Economic Perspectives 13 (4): 21?42.
132 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS APRIL 2012
 Romer, Paul M. 1986. ?Increasing Returns and Long-run Growth.? Journal of Political Economy 94 
(5): 1002?37.
 Solow, Robert M. 1956. ?A Contribution to the Theory of Economic Growth.? Quarterly Journal of 
Economics, 70(1): 65?94.
 Stolper, Wolfgang F., and Paul A. Samuelson. 1941. ?Protection and Real Wages.? Review of Economic 
Studies 9 (1): 58?73.
 Stulz, Ren? M. 2005. ?The Limits of Financial Globalization.? Journal of Finance 60 (4): 1595?1638.
 Thompson, Samuel B. 2006. ?Simple Formulas for Standard Errors that Cluster by Both Firm and 
Time.? Unpublished. http://papers.ssrn.com/sa13/papers.cfm?abstract_id=914002
 United Nations Industrial Development Organization, 1963?2004. ?Industrial Statistics Database: 
INDSTAT3 2006 ISIC Revision 2? United Nations, http://www.esds.ac.uk/international/support/
 user_guides/unido/indstat.asp (accessed February 4, 2012).
 Wacziarg, Romain, and Jessica Seddon Wallack. 2004. ?Trade Liberalization and Intersectoral Labor 
Movements.? Journal of International Economics 64 (2): 411?39.
 Wilcoxon, Frank. 1945. ?Individual Comparisons by Ranking Methods.? Biometrics Bulletin 1 (6): 
80?83. 
Wilson, Ian M. 1992. The Wilson Directory of Emerging Market Funds. Saskatoon: Wilson Emerging 
Market Funds Research.