Alderson 1999

Explaining Deindustrialization: Globalization, Failure, or Success?
Author(s): Arthur S. Alderson

Source: American Sociological Review, Vol. 64, No. 5 (Oct., 1999), pp. 701-721
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2657372
Accessed: 28-09-2015 13:26 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to American Sociological
Review.
http://www.jstor.org
This content downloaded from 142.132.1.147 on Mon, 28 Sep 2015 13:26:56 UTC
All use subject to JSTOR Terms and Conditions
EXPLAINING DEINDUSTRIALIZATION:
GLOBALIZATION, FAILURE, OR SUCCESS?
Arthur S. Alderson
Indiana University
Although sociologists have expressed growing interest in globalization, they

have devoted little sustained empirical attention to the many claims made in
its name. Ifocus on the link that has been drawn between globalization and
the deindustrialization of the advanced industrial societies. To examine this,
I employ a pooled time-series of cross-sections data set that combines ob-
servations on 18 OECD nations across the 1968-1992 period. Fixed-effects
regression models that control for unmeasured country-specific effects re-
veal support for arguments that implicate foreign direct investment and
North-South trade in the declining percentage of the laborforce employed in
manufacturing in the OECD countries. Regression results also show that
deindustrialization across this period is largely explained by a model that
combines an attention to the post-Golden Age "troubles" of northern manu-
facturing with classic generalizations of the process of development. Inter-
pretation of the empirical findings is tempered by an exercise in
counterfactual history, which reveals that deindustrialization would have
been considerable in these countries even if the upswings in direct invest-
ment and southern imports had not occurred or if the performance of the
manufacturing sector had been stronger
For well over a decade there has been typified by the growing mobility of capital
growing interest among sociologists in and heightened competition in international
the topic of "globalization" (Waters 1995). markets-on traditionally high-wage manu-
This interest has been driven by the sense facturing employment in the advanced indus-
that globalization raises interesting new trial societies.
questions and has profound implications for The facts of declining manufacturing em-
a range of sociological subfields. Although ployment are well-known. As the postwar
sociologists have been at the forefront of ef- "Golden Age" (Marglin and Schor 1990)
forts to give "global-babble" a rigorous and drew to a close in the late 1960s, the average
consistent theoretical status, they have de- Organization for Economic Cooperation and
voted little empirical attention to often- Development (OECD) country had over a
voiced concerns for the substantive implica- quarter of its labor force in manufacturing.
tions of globalization (Hoogvelt 1997:116). Across the 1970s and 1980s, the share of the
One such concern surrounds the impact of an labor force in manufacturing declined, in
increasingly integrated world economy- some cases precipitously. By the early 1990s,
employment in manufacturing typically
* Direct all correspondence to Arthur S. amounted to under one-fifth of total employ-
Alderson, Departmentof Sociology, Indiana Uni- ment. In the interim, many cities and regions
versity, Ballantine Hall 744, Bloomington, IN in the advanced industrial world had endured
47405-6628 (aralders@indiana.edu). I thank two decades of severe social dislocation and
Franqois Nielsen, Rachel Rosenfeld, Ken Bollen, upheaval (Hall 1993; Wallace and Rothschild
Howard Aldrich, Craig Calhoun, Indermohan
Virk, David Brady, David James, Kent Redding, 1988). Among other important develop-
participants in the Political Economy Seminar at ments, this decline of manufacturing em-
Indiana University, and the ASR Editor and re- ployment has been linked to the recent up-
viewers for their comments on early drafts. Data swing in inequality in a number of the indus-
are available from the author on request. trial countries (Lorence and Nelson 1993;
American Sociological Review, 1999, Vol. 64 (October:701-721) 701
702 AMERICAN SOCIOLOGICAL REVIEW
Nielsen and Alderson 1997; Ryscavage, one.2 Researchers have focused in particular
Green, and Welniak 1992). on the roles of rising foreign direct invest-
The phenomenon of "deindustrialization" ment and "North-Southtrade"(i.e., trade be-
has given rise to a lively debate as to which tween developed and developing societies).
culprits are responsible.' This debate has fo- Both factors have become increasingly
cused mainly on two questions: First, has the prominent in the advanced industrial societ-
dramaticdeindustrializationof former indus- ies in recent years. Between 1982 and 1992,
trial giants such as the United Kingdom been for instance, the value of outflows of direct
the result of poor industrial performance investment from the OECD countries grew
(rooted, perhaps, in features of postwar po- from 20 billion U.S. dollars to over 200 bil-
litical economy), or has it arisen from a more lion (International Monetary Fund [IMF]
mundane, entirely "normal"process of eco- 1987, 1994). The value of OECD manufac-
nomic development? Second, what role has tured imports from "Southern" countries
globalization played in deindustrialization? (i.e., non-OECD, non-COMECON [Council
Have the increasing mobility of capital (i.e., for Mutual Economic Aid] countries) grew
"capitalflight") and heightened international from around 87 billion U.S. dollars to 426
competition had an important impact on billion (OECD 1984, 1994a).
manufacturing's share of employment in the
advanced industrial societies, or is such talk
mere "pop internationalism" (Krugman Direct Investment and Deindustrialization
1996)? Regarding foreign direct investment, Blue-
Sociologists would, given their leading stone and Harrison (1982) have offered the
role in an emerging field of "globalization most influential account (also see Harrison
studies," seem to have a natural interest in and Bluestone 1988). They argue that direct
addressing these questions. However, they investment in the contemporaryperiod is be-
have generally not been empirically pursued ing undertaken as part of a "globalization
by sociologists. I address this oversight. I gambit." This move is an integral part of a
introduce the "globalization thesis" and re- new set of corporate strategies designed to
view arguments that link the recent up- abrogate the old postwar social contract be-
swings in foreign direct investment and tween capital, labor, and the state in order to
North-South trade to the deindustrialization restore acceptable levels of profitability in
of the advanced industrial societies. A gen- response to the "profitsqueeze" of the 1970s.
eral model of deindustrialization provides a The result of this strategy, they argue, is a
background against which to test this thesis. kind of hollowing of the economy-direct
The questions I ask are rather concrete: (1) investment is no longer undertakenby firms
Is there empirical evidence for the claim in an effort to complement domestic invest-
that globalization plays a significant role in
2 There is a large literature on deindustrial-
deindustrialization, and, if so, (2) how sub-
stantial is that role relative to other factors? ization. Much of the early literatureconcerned the
debate over its definition and whether it was in
fact occurring. Arguments linking globalization
THE GLOBALIZATION THESIS and deindustrialization appear to have originated
in the context of the broader debate in the United
The idea that globalization may have con- Kingdom over the "Englanditis" of the 1970s
tributed significantly to the phenomenon of (Coates and Hillard 1986). Many of the key terms
deindustrialization is a fairly general and concepts that emerged in this debate were
quickly taken up by American and Canadian re-
searchers. In continental Europe, the debate over
I I define deindustrializationhere as the decline globalization and deindustrialization has only re-
of manufacturingemployment relative to employ- cently come to engage the broader public outside
ment in other sectors. Given this definition, abso- of academic and trade union circles. One expla-
lute levels of output or employment in the manu- nation for this is that, in contrast to the United
facturing sector are not a concern. This is not Kingdom, United States, or Canada, phenomena
meant to imply, however, that attention to abso- such as foreign direct investment became notice-
lute levels is unimportant for our understanding able features of many industrial economies only
of the phenomenon (Singh 1977, 1989). in the 1980s (Messerlin 1995).
DEINDUSTRIALIZATION AND GLOBALIZATION 703
ment and production, but to replace it. As differentials are relatively minor (e.g., the
firm after firm in the manufacturing sector United States and Germany). Moreover em-
has gone abroad in search of lower labor pirical analyses have repeatedly shown that
costs, the end result of the growth of direct labor cost differentials have, at most, only a
investment in recent years has been deindus- modest and secondary effect on direct invest-
trialization. Although direct investment is ment (e.g., Alderson 1997; Bajo-Rubio and
not the sole cause of deindustrialization,it is Sosvilla-Rivero 1994; Erickson and
"certainly a major reason that the United Kuruvilla 1994; Klein and Rosengren 1994;
States lost a significant fraction of its manu- Pain 1993). While the spatial reorganization
facturing base" (Harrison and Bluestone of manufacturing along a North-South axis
1988:29). is real-witness the maquiladoras along the
While Bluestone and Harrisonfocus on the U.S. border with Mexico-this does not rep-
United States, other scholars paint a similar resent the general pattern of direct invest-
picture of the role of multinational enter- ment in the last two decades or so.
prises and direct investment in deindustrial- Second, these arguments also largely ne-
ization in other industrial societies. Stopford glect the fact that deindustrialization had
and Turner (1985), for instance, show that long been anticipated by social scientists.
roughly one-third of all manufacturingjobs Social scientists had predicted, well before
lost in the United Kingdom between 1972 Bell (1973), the "coming of post-industrial
and 1983 were the result of the actions of 58 society." For instance, Clark ([1940] 1957)
U.K. multinational enterprises (which added attributed the shift from manufacturing to
200,000 such jobs outside Britain across the services to two processes: First, with eco-
same period). This echoes the data presented nomic development, "as real income per
by Bluestone and Harrison for the U.S. and head increases, it is quite clear that the rela-
the earlier research of Frank and Freeman tive demand for agriculturalproducts falls all
(1978) that links substantial U.S. domestic the time, and that the relative demand for
manufacturing job losses to direct invest- manufacture first rises, and then falls in fa-
ment by U.S. multinationalenterprises.More vor of services"; second, given higher rela-
broadly, Beenstock (1984) and the United tive productivity in the industrial sector, "a
Nations Industrial Development Organiza- stationary relative demand for manufactures
tion (1983) were, among others, early voices would lead to a decreasing proportion of the
attributing a significant portion of the gen- labor force employed therein" (pp. 493-94,
eral North-to-Southreallocation of manufac- emphasis in original). Clark adds, "[E]ven
turing production to the direct investment when the relative demand for manufactures
activity of multinational firms. is increasing, we still generally expect, in the
While these arguments have achieved a long run, a decreasing proportion of the la-
fair degree of currency, they have definite bor force to be employed therein" (p. 494).
limitations. For one, the image of globaliza- Thus as productivity grows and as the indus-
tion that emerges is one of North-to-South trial economies mature, one should expect
capital flight, in which direct investment de- that in the normal course of economic devel-
cisions are guided mainly by simple labor opment secondary-sector employment will
cost differentials. This is a highly inaccurate contract while the tertiary sector will expand
picture of the contemporarypattern of direct in the face of rising demand for services.4
investment.3 Most direct investment flows These criticisms raise two major issues for
between nations with roughly comparablela- an empirical analysis of the globalization the-
bor market conditions in which labor cost sis. First, the mechanism through which di-
rect investment is held to produce deindustri-
3 In Harrison and Bluestone's (1988) defense,
alization (i.e., North-to-South capital flight)
they do note that direct investment is influenced is more problematic than it appears at first
by an array of factors in addition to labor costs.
Nonetheless, in discussing the hollowing of the
glance. The vast majority of all direct invest-
manufacturing sector, they stress that labor cost
differentials were a key determinant of the up- 4 This generalization, Clark ([1940] 1957:492)
swing in direct investment that occurred over the relates, was first made as long ago as 1691 by Sir
past two decades or so. William Petty (1899).
ment flows occur between industrial societ- their manufacturing sectors. If profits from
ies. This suggests that a focus on the direct abroad grow at a faster rate than the domes-
labor-displacing effects of direct investment tic economy-an entirely conceivable phe-
may be too narrow.To explore the effects of nomenon-the investing economy, given that
total flows of direct investment, this argu- such profits eventually translate into in-
ment must be supplemented with some alter- comes, may begin to import more manufac-
native account of the effects of direct invest- tured goods. This, in turn, may cause a rise
ment on manufacturingemployment. Second, in the exchange rate of its currency leading
when testing for an effect of direct investment to a deterioration of its manufacturingtrade
on deindustrialization,researchers should be position and, ultimately, its manufacturing
aware of the important long-run generaliza- sector. As has been argued in regard to other
tions offered by earlier analysts. aspects of British and American hegemony
Alternative accounts (e.g., Cowling 1986; (Chirot 1986), there may be long-term "costs
Cowling and Sugden 1987; Hymer 1979; of dominance" in this realm as well.
letto-Gillies 1992; Tanaka 1991) suggest In addition to the "wealth trap"posited by
that, beyond the direct labor-displacing ef- Rowthorn and Wells, letto-Gillies (1992) ar-
fects of direct investment stressed in the gues that direct investment may also contrib-
capital-flight literature, direct investment ute to deindustrialization by lowering the
may have a range of (dynamically evolving) rate of domestic capital formation. Multina-
indirect effects on the relative size of the tional enterprises typically enjoy higher rates
manufacturinglabor force. of return on investment than do comparable
First, outflows of direct investment may domestic firms. When the activities of such
over time move a nation's economy into what firms are substantial,this should increase the
Rowthorn and Wells (1987) term, by analogy required marginal rate of returnon domestic
to the more familiar "debt trap"from devel- investment and influence investment deci-
opment economics, the "wealth trap"-"the sions accordingly. This places a nation in a
automatic process by which a country which disadvantaged position relative to nations
is intrinsically a capital exportermay become that are less dominated by the activities of
a rentier nation" (p. 353, emphasis in origi- multinational firms and may contribute to a
nal). In short, what starts out on the national vicious circle resulting in deindustrialization.
accounting ledger as an outflow (i.e., the di- letto-Gillies (1992) adds:
rect investment) may turn into a real inflow Global scanning combined with electronic
as profits from abroadoutrunoutflows of for- technology in communications and ease of
eign investment. letto-Gillies (1992) has built movements of funds across frontiers by [multi-
on this insight and argues that the end result national enterprises] may have contributed to
of this process is that raising the rate of returnon purely financial in-
vestment. This will have raised the marginal
... [nations] with a long tradition of outward
rate of return required on real capital forma-
foreign investment are likely to experience tion. Similarly, high rates of return on the ser-
overallnet "positive"effects on the balanceof vices sector (particularlythe financial services)
payments.These may cause a rise in the ex- may have raised the marginal rate of return on
change rate; in a situation in which the [manufacturingsector investments]. (P. 188)
economycannot-or is not allowedto by mon-
etaryand fiscal policies-expand to meet the Direct investment, then, may contribute to
extra demandgeneratedby the inflow of in- deindustrializationthrough a number of inti-
comes, the overall long-termeffect will be a mately intertwined mechanisms. Direct in-
weakeningof the manufacturingsector with
loss of jobs and"negative"deindustrialization. vestment may raise the required marginal
(P. 185) rate of return on domestic investment, shift
investment from manufacturing to services,
Thus for nations like the United Kingdom and reorient investment away from real in-
and United States-nations with long expe- vestments toward financial investments.
riences with outward direct investment-di- While these arguments regarding the
rect investment may not only displace manu- "wealth trap"and the effects of direct invest-
facturing labor, but may lead these countries ment on capital formation remain specula-
to "live on their past" to the detriment of tive, when combined with an attention to the
direct labor-displacing effects of direct in- ture have enabled developing countries to re-
vestment, they provide a rough framework alize a comparative advantage in unskilled,
for the interpretationof any observed effect labor-intensive manufacturing, thereby in-
of the total outflow of direct investment on creasing sharply the flow of manufactured
manufacturing's share of employment. Each goods from the South to the North over the
predicts a negative relationship and high- past two to three decades. This reduced the
lights a distinct moment in a nation's history profitabilityof labor-intensivemanufacturing
of direct investment. In the short term, there in the North and of manufacturingas a whole
is the hollowing effect posited by analysts relative to other sectors. Subsequently, new
such as Harrison and Bluestone. Over a investment in these industries was either dis-
longer term, there are the effects of direct in-couraged (leading to the outright scrapping
vestment on the rate and characterof capital of existing capacity in labor-intensive sectors
formation tracedout by letto-Gillies. Finally, and the direct displacement of labor) or di-
given a sufficient history of direct invest- rected toward defensive labor-saving innova-
ment, there is the possibility of the "wealth tions designed to keep southern imports at
trap." bay (further depressing the demand for un-
skilled labor). All told, the upswing in im-
ports of southern manufacturedgoods is ar-
North-South Trade and Deindustrialization gued to have had a large impact on manufac-
The search for the causes of deindustrial- turing employment in the North.
ization also has focused on the role of trade, The presentation of a new set of estimates
particularlyNorth-South trade. The common of the impact of North-South trade in manu-
perception that imports from developing factures on employment is the centerpiece of
countries are a major cause of deindustrial- Wood's (1994) analysis. Wood finds that
ization was given early expression by such trade has had a far larger effect on em-
Beenstock (1984). Beenstock argues that ployment than had been indicated in previ-
comparativeadvantagein a wide range of in- ous studies:
dustries has, for a variety of reasons, shifted The estimatedimpactin the Southis roughly
to developing societies, and that this has led five times larger,andin the Northroughlyten
to a reallocation of manufacturinginvestment times larger,thanstandardcalculationswould
from North to South. Deindustrialization in suggest.... In the North,the cumulativeeffect
the North is thus the result of the rise of up to 1990 of the changes in trade with the
southern manufacturing (and the liberaliza- South was to reducethe demandfor laborin
tion of world trade in manufacturedgoods). manufacturing by about9 million (plus or mi-
Students of internationaltradehave tended nus 3 million) person-years-or by about 12
to reject this argument, citing studies that percent of actual employment in Northern
manufacturing. (P. 166)
suggest that the net impact on northern em-
ployment of trade with the developing world With these estimates in hand, Wood (1994)
is negligible.5 Recently, however, Wood concludes that manufactured imports from
(1994) has reinvigorated this debate, mar- the South contributedto
shalling an impressive case in favor of the ... a reductionof at least 4 percentagepoints
hypothesis that trade-induced shifts in the in the shareof manufacturing in total employ-
demand for skilled versus unskilled labor ment for the Northas a whole [and that the]
have been an important cause of the fall in expansionof tradewith the Southwas an im-
manufacturing'sshare of total employment in portantcause of the deindustrialization of em-
the advanced industrial world. Wood argues ploymentin the North over the past few de-
that trade liberalization and improvements in cades. (P. 13)
transportationand communicationinfrastruc-
A MODEL OF
5 See the much-discussed OECD Jobs Study DEINDUSTRIALIZATION
(OECD 1994b), for example. A review of the
available evidence concludes that the impact of To test the globalization thesis, I employ the
trade on employment in OECD countries is "very frameworksuggested by Rowthorn and Wells
small ... relative to other influences" (p. 108). (1987). Deindustrialization (i.e., the relative
decline of manufacturing employment) can in the early 1980s), high labor costs, poor
occur in at least three ways. First, there is the product quality, and the failure or inability
"positive deindustrialization,"noted by early of firms to respond to changing market con-
analysts such as Clark ([1940] 1957), which ditions have all been identified as factors in
represents much of conventional economic the "decline of competitiveness" experienced
thinking on deindustrialization (e.g., Singh by a numberof industrialsocieties in the past
1989). Here deindustrialization is viewed as two to three decades (Ferguson and Ferguson
a "normal"structuralfeature of all economies 1994).
during the course of economic development. Finally, looking at the international
With development, as per capita income in- economy, Rowthorn and Wells (1987) argue
creases the share of employment in agricul- for the existence of a generic "trade-related
ture falls and the share of employment in deindustrialization."Trade affects manufac-
manufacturingrises until a high level of de- turing employment through macroeconomic
velopment is attained. However, beyond channels and throughits influence on special-
some thresholdof per capita income the share ization. First, in a mature economy dein-
of employment in services begins to increase dustrializationmay be associated with either
at the expense of manufacturingemployment. strong or weak trade performance. When
This occurs as a consequence of the typically manufacturingtradebalances are positive and
higher rate of productivity growth in the large and the strength of the manufacturing
manufacturing sector relative to the service sector contributes to sustained economic
sector and of the systematic changes in con- growth, the manufacturingsector may begin
sumption patterns that occur over the course to shed labor (via positive deindustrial-
of development (specifically, differences in ization) at a higher rate than it would in the
the income elasticity of demand across sec- absence of trade. When manufacturingtrade
tors). Such deindustrialization is "positive" positions are deterioratingand investment in
because it is viewed not as a pathological manufacturing falls, the manufacturing sec-
phenomenon but as a symptom of economic tor may begin to shed labor (via negative
success. Labor shed in the course of positive deindustrialization) into a stagnating econo-
deindustrializationis viewed as being quickly my in which it cannot be absorbedby the ser-
absorbed by the growing service sector. vice sector. Underlying these macroeconomic
Rowthorn and Wells (1987) argue that effects of trade on the relative size of the
deindustrialization can also occur through a manufacturing labor force are the structural
process of "negative deindustrialization." effects. Nations that run manufacturingtrade
Here deindustrialization is the result of a surpluses will, all else being equal, devote
pathological phenomenon-a structural dis- more resources and labor to this sector than
equilibrium in the economy that prevents a will nations that run deficits. Thus trade may
nation from reaching its growth potential or lead to further specialization in manufactur-
full employment of its resources. It is mani- ing among successful nations, and accelerate
fested in poor performance in the manufac- the move away from specialization in manu-
turing sector and is accompanied by a slow- facturing among unsuccessful nations.
down in manufacturing output and produc- These differing forms of deindustrial-
tivity. This sequence of events leads to poor ization can operate concurrently; that is, the
performance for the economy generally and deindustrializationexperienced by any given
a decline in competitiveness (in a cumulative nation is unlikely to be the result of any one
vicious circle). The labor shed by negative "form" of deindustrialization. Indeed, the
deindustrializationis, given the general state deindustrializationexperienced in the indus-
of the economy, not absorbed by the service trial societies is probably the partial result of
sector. Thus, positive deindustrialization is a mix of "positive" and "negative" factors
associated with rising real incomes and full that has varied across time and space. When
employment, while negative deindustriali- testing the link between globalization and
zation is associated with stagnating real in- deindustrialization, I control for these alter-
comes and rising unemployment. The appre- native sources of deindustrialization and
ciation of currencies (as in the United King- build on the baseline model suggested by
dom in the late 1970s and the United States Rowthorn and Wells (1987:3 1):
Percent of labor force in manufacturing = data drawn from the U.S. Bureau of Labor
f (Real GDP per capita, Statistics (1993).
Unemployment rate, Real GDP per capita (log) is measured in
Net manufactured exports/GDP). U.S. dollars. Data are from the Penn World
Tables Mark 5.6 (Summers and Heston
Real GDP per capita is a measure of devel- 1991). To reduce the collinearity generated
opment and is assumed to capture the "matu- by the inclusion of real GDP and its square
rity effect" prompting positive deindustrial- in some of the models, I deviated this vari-
ization. A curvilinear, inverted U-shaped re- able from its median (4.043). This transfor-
lationship is expected, as the share of em- mation reduces the correlation between a
ployment in manufacturing first rises and variable and its square (Koopmans 1987).
then, after a certain point, starts to fall. This Data on the unemployment rate (log) are
relationship is approximatedas a second-de- drawn from the same OECD sources as per-
gree polynomial of real GDP per capita.6The centage of the labor force in manufacturing
unemployment rate captures the "failure ef- (OECD 1989; 1995). Whenever possible, I
fect" associated with negative deindustrial- use data that are standardized by OECD
ization. A negative relationship with manu- methods.
facturing's share of employment is expected. Net manufactured exports/GDP is mea-
Finally, net manufactured exports/GDP is an sured as exports minus imports, divided by
indicator of trade-relateddeindustrialization. GDP; it is measured in current U.S. dollars.
As both positive and negative deindustriali- Data are drawn from the WorldBank's World
zation are already controlled, a positive rela- Tables (various years). CurrentGDP in U.S.
tionship is expected. Net manufactured ex- dollars is drawn from the OECD's National
ports/GDP in this instance is viewed as pri- Accounts (various years).
marily tapping into the "specialization ef- Data on direct investment outflow per
fects" of trade. capita (log) are drawn from the IMF's Bal-
The specification of the model to be esti- ance of Payments Statistics Yearbook(vari-
mated is: ous years). This represents the book value of
all investments made in a given year by na-
Percent of labor force in manufacturing =
tionals of a given country to acquire mana-
Real GDP per capita
gerial interest in enterprises sited in other
+ (Real GDP per capita)2 countries (IMF 1977). Population is from the
+ Unemployment rate + Net OECD's National Accounts (various years).
manufactured exports/GDP + Direct Finally, southern import penetration/GDP
investment outflow per capita (log) is measured as manufactured imports
+ Southern import penetration/GDP. from non-OECD, non-COMECON countries
in current U.S. dollars, divided by GDP.
Manufacturedimports are defined as SITC 5,
DATA AND METHODS
6, 7, and 8; import data are from the OECD's
Data on the dependent variable, percentage Foreign Trade by Commodities series (vari-
of the labor force in manufacturing, are ous years). Current GDP in U.S. dollars is
drawn primarily from the OECD's Labour drawn from the OECD's National Accounts
Force Statistics, 1967-1987 (1989) and (various years).
Labour Force Statistics, 1973-1993 (1995). Lagging all explanatoryvariables one year,
Cases not covered in these sources (e.g., the data set contains a maximum of 432 ob-
Netherlands, 1967-1974) are coded with servations-25 observations (1968-1992)
for each of 16 nations, 10 observations
6 Given their high level of development, I ex- (1984-1992) for Switzerland, and 23 obser-
pect that most countries under study will (in most
years) be arrayedalong a declining slope of manu-
facturing employment and development. In keep- likely scenario is that, as has happened with the
ing with the inverted-U hypothesis, I expect that relationshipbetween agriculturalemployment and
their trajectorywill be concave to the origin. This development, the relationship between manufac-
should not be construed as suggesting that manu- turing employment and development in these so-
facturing employment will reach zero. A more cieties will turn increasingly convex to the origin.
vations (1968-1990) for France.7 I use a the other explanatory variables in the model.
pooled time-series of cross-sections method- Whereas country-specific effects can corre-
ology designed to address the heterogeneity late with other regressors in the FEM, they
bias-the confounding effect of unmeasured are assumed not to in the REM.
time-invariant country-specific variables- I estimated OLS, REM, and FEM models.
that can plague ordinaryleast squares (OLS) The results of Breusch and Pagan's Legrange
using this sort of data set (see Greene 1993; Multiplier Test (1979; also see Greene 1993)
Hsiao 1986). Heterogeneity bias can seri- clearly argued in favor of REM or FEM over
ously affect OLS coefficient estimates, mak- OLS. The Hausman test (1978; also see
ing OLS an inappropriate estimation tech- Greene 1993), however, yielded inconsistent
nique. The fixed-effects (FEM) and random- results, favoring REM in some models and
effects (REM) models are two commonly FEM in others. The Hausman test results in-
used estimation strategies designed to correct dicate that I am not always justified in treat-
for unmeasuredtime-invariant factors. ing country-specific effects as uncorrelated
The FEM (fixed effects) procedure differs with other explanatory variables in the
from the OLS procedure in its treatment of model, and thus I opt for the more conserva-
the intercept. Under OLS, all countries are tive choice and present here the FEM esti-
constrained to the same intercept, whereas in mates of the regression models.8 Note, how-
the FEM, indicator variables are introduced ever, that the FEM and REM results were
for each country to represent country-spe- substantively similar.
cific intercepts. Doing this "simulates" un- To correct for serially correlated distur-
measured time-invariantcountry-specific ef- bances, I applied an adjustment for first-or-
fects and thereby addresses the problem of der autocorrelation (AR1) using the
heterogeneity bias. The FEM takes the fol- Cochrane-Orcutttransformation(see Greene
lowing general form: 1993). In the FEM context, this yields
(Greene 1992:324):
Yit= a1dIit++ a2d2 +t+ +P Xit+ 'Fit
where the subscript i denotes the country, t Yit - PYi= I (Xit - PXi t- ) + *i(1- P) + Nit-
denotes the time point of observation, a rep- As the first observation in each country is
resents the individual-specific constants, and lost in the course of applying p, the maximum
dj denotes country-specific indicator vari- number of observations is reduced to 414. I
ables that equal 1 when j = i. The errorterm, estimated the FEM and REM models with an
Eit,is the classical disturbance term. autocorrelated error structure using the
The REM (random-effects) procedure LIMDEP statistical program (Greene 1992).
takes the following form:
8 In using the FEM, I assume that parameters
Yit=aa + P'Xit+ ui + Eit are homogeneous across years and that any coun-
Thus rather than treating country-specific try heterogeneity can be effectively captured in
effects as fixed effects to be estimated, as in differences in the constant term. To assess the re-
the FEM, the REM treats them as a random alism of these assumptions, I estimated a series
of random coefficient models (RCM) that allow
component of the errorterm. Comparedwith country- and time-varying parameters (Hildreth
OLS, the REM involves the estimation of an and Houch 1968; Hsiao 1986; Swamy 1971). Re-
additional country-specific component of the garding the FEM's assumption of temporal pa-
error variance-ui. Like the classical error rameter constancy, I could not reject the hypoth-
term sit, this country-specific error compo- esis that parameters are constant across time on
nent ui is assumed to be uncorrelated with the basis of Swamy's (1971; also see Greene
1993:461-62) chi-square test of RCM versus the
constant coefficients model. Regarding the
7 The Swiss series is "short"because Switzer- FEM's treatment of country heterogeneity, I
land has only recently begun to report direct in- found that it made no substantive difference
vestment data. The French series is limited by whether I allowed intercepts to vary across coun-
missing data on manufacturing employment for tries (FEM) or allowed intercepts and slopes to
1991 and 1992. The fact that the panel is "unbal- vary across countries (RCM). These results sug-
anced" does not raise any special problems for the gest that FEM assumptions are reasonable for
methodology I employ. these data.
Table 1. Unstandardized Coefficients from the Fixed-Effects Regression of the Percentage of the La-
bor Force in Manufacturing on Selected Independent Variables: 18 OECD Nations, 1968-
1992
Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Real GDP per capita (log)a 532.75** -- 250.53**

(128.65) (92.56)
(Real GDP per capita)2 -67.27*** -32.96**
(15.83) (11.49)
Unemployment rate (log) -5.99*** -6.59***
(.57) (.55)
Net manufactured exports/GDP -.03 .06*
(.03) (.03)
Direct investment outflow per capita (log) -2.17*** -.51*
(.28) (.27)
Southern import penetration/GDP (log) -8.43*** -4.13***
(1.06) (1.10)
Adjusted R2 .432 .553 .252 .517 .542 .788

Rho .814 .762 .878 .740 .780 .689
Number of observations 405 405 405 405 391 391
Note: Numbers in parentheses are standard errors. First-order autocorrelation coefficients (ARI) are rep-
resented by rho.
a Deviated from median
*p < .05 ** < .01 < .001 (one-tailedtests)
An analysis of outliers and influential cases real GDP per capita, the unemployment rate,
was performed using the various diagnostics and net manufactured exports/GDP. Models
available in the SYSTATand SYGRAPH sta- 4 and 5 present results, respectively, for the
tistical programs (Belsley, Kuh, and Welsch measures of direct investment outflow and
1980; Bollen and Jackman 1985; Wilkinson southern import penetration. Model 6 in-
1990a, 1990b). These diagnostics revealed cludes all these variables.
several cases that appeared unduly influen- Model 1 introduces real GDP per capita
tial; I excluded these outliers.9 (log) and its square to capture the maturity
effect-the hypothesized curvilinear, in-
verted U-shaped relationship between devel-
REGRESSION RESULTS
opment and manufacturing's share of em-
The fixed-effects (FEM) regression results ployment (positive deindustrialization).Both
are presented in Table 1. Models 1, 2, and 3 coefficients are significant and correctly
present results for the three variables that signed, indicating that manufacturing em-
form the baseline deindustrialization model ployment first rises and then declines with
suggested by Rowthorn and Wells (1987): development. Finding the maximum of the
parabola, the regression coefficients in
9 In total, nine observations were identified as Model 6 indicate that the value of real GDP
deviant on the basis of multiple criteria (i.e., per capita beyond which manufacturing's
Studentized residual, Cook's D, influence in par- share of employment begins to fall is $6,311
tial regression plots): Australia 1990; Belgium (in 1985 prices). All but two of the countries
1968, 1981, 1982; Canada 1982; Denmark 1979; (Japan and Ireland) had reached this level of
Japan 1991; New Zealand 1977, 1984. When
these cases are excluded, the coefficients of net
maturity by 1968. Table 2 shows that some
manufactured exports/GDP and southern import countries reached this threshold substantially
penetration/GDP were appreciably larger. The earlier than others: Australia, Canada, New
exclusion of outliers had no other notable sub- Zealand, Switzerland, and the United States
stantive effect. had reached the estimated turning point at or
Table 2. The Maturity Effect: Year in Which plain why deindustrialization was so severe
Real GDP per Capita Reached Esti- in countries such as Australia, Britain, New
mated Turning Point Value of $6,311 Zealand, and Sweden-these countries had
Country Date all reached maturity before the close of the
1950s. Thus employment was bound to shift
Australia before 1950 away from manufacturingat a comparatively
Austria 1966 rapid pace. By contrast, Figure 1 shows that
Belgium 1964 a handful of other countries (Austria, Italy,
Canada before 1950
Finland, Japan, Ireland) reached maturity at
a relatively late date. Indeed, Ireland and Ja-
Denmark 1959 pan only reached the stage of development
Finland 1965 beyond which further growth is associated
France 1962 with deindustrialization during the period
Germany 1960 under study. From the perspective of the ma-
Ireland 1978
turity thesis, it is not surprising that these
countries have experienced only modest de-
Italy 1967 clines in manufacturing's share of employ-
Japan 1969 ment.
Netherlands 1962 The maturity thesis appears to offer a con-
Norway 1963 vincing explanation for the deindustriali-
zation of the advanced industrial societies-
New Zealand before 1950
the adjustedR2of Model 1 is .432. Of course,
Sweden 1954 some of the fit in Model 1 is contributed by
Switzerland before 1950 the fixed effects for country. To isolate the
GreatBritain 1957 marginal contribution of the real GDP per
United States before 1950 capita quadratic, I calculated the coefficient
of partialdeterminationusing the "extrasums
Note: The thresholdvalue of $6,311 is expressed of squares" method (Neter, Wasserman, and
in 1985 U.S. dollars. Kutner 1989).12 I found that, net of the fixed
effects, the maturity effect explains roughly
before 1950; Austria, Belgium, Finland, and 9 percent of the variation in the percentage of
Italy did not reach it until well into the labor force in manufacturing (partial R2 =
1960s.10 .093), indicating that the maturity thesis is
These differences help explain cross-na- poorer at explaining within-countryvariation
tional variation in both the timing and extent in deindustrialization than between-country
of deindustrialization (Rowthorn and Wells variation (Figure 1).
1987). Figure 1 plots the change in manu- Model 2 assesses the role of negative dein-
facturing's share of employment from 1967 dustrialization. The above discussion sug-
through 1992 against the approximatedate at gests that unemployment can proxy for the
which "maturity" was reached.'1 Although sort of structural imbalance that is stressed
the left-censoring clouds the issue a bit, there in more critical accounts of deindustrial-
is a strong relationship between maturityand ization. Under this view, the deindustriali-
subsequent deindustrialization (r = .722): zation experienced by the advanced indus-
Countries that reached maturity earliest ex- trial societies over the past few decades
perienced the largest declines in manufac- should not be viewed as the result of a "natu-
turing's share of employment. This helps ex- ral" and "self-correcting" phenomenon, but
12 With two independent variables, the coeffi-
10 Summers and Heston's (1991) real GDP cient of partial determination between Y and X1,
time-series begins in 1950. Thus the cutoff date given that X2is already in the model, is calculated
here is entirely arbitrary. as (Neter et al. 1989: 285):
11All regression lines presented in Figures 1,
2, 4, and 5 are estimated by ordinary least SSR(X1X2 )/SSE(X2),
squares. Due to missing data, Switzerland does where SSR denotes the regression sum-of-squares
not appear in Figures 1 through 5. and SSE denotes the error sum-of-squares.
0
l) Japan Ireland
-2
Finland
-4
0
Italy
-6 France
~~~~~~~~~Austria
O ~~~~~~~Denmark
0 -8
LIL
0 Canada Germany+.
M ~ ~ *Netherlands
-10
0
oUnited States
Norway
-12
0 ~~~~~~Sweden
New Zealand
0-14
* *~~~~~~~Belgium
Australia United Kingdom
0
-16
1945 1950 1955 1960 1965 1970 1975 1980
Year Country's Real GDP Reached $6,311
Figure 1. Scattergram Showing the Relationship between Change in the Percentage of the Labor Force
in Manufacturing and Year Country's Real GDP Reached the Threshold Value of $6,311:
17 OECD Countries
as evidence of profound economic distress instance, both "heavy" (e.g., Britain) and
and a symptom of economic failure. The "light" (e.g., Ireland) deindustrializing na-
highly significant negative effect of the un- tions experienced substantialincreases in un-
employment rate (log) lends support to the employment rates across the period under
failure thesis. Failure, defined as rising un- study. Nonetheless, clear supportis found for
employment, appears to have a large impact the failure thesis. On the basis of the regres-
on manufacturing's share of employment: sion results, it seems highly unlikely that the
The partial coefficient of determination of deindustrialization witnessed across the ad-
the unemployment rate in Model 2 is .225. vanced industrial world over the past two to
Figure 2 explores the association between three decades has been the result of positive
change in the unemployment rate and change deindustrialization alone.
in manufacturing'sshare of employment. Al- Model 3 introduces net manufactured ex-
though the regression results indicate that the ports/GDP as an indicator of a nation's
failure effect accounts for more within-coun- manufacturing trade position. This variable
try variation in manufacturing'sshare of em- captures "trade-related deindustrializa-
ployment than does the maturity effect, Fig- tion"-the effect of autonomous changes in
ure 2 suggests that the failure effect is less the structure of a nation's foreign trade on
successful in explaining cross-national varia- manufacturing employment. The prediction
tion in deindustrialization trajectories."3For of a positive relationship between these two
variables is premised on controls for positive
13 The correlation between change in the per- and negative deindustrialization.In this con-
centage of the labor force in manufacturing and text, the specialization effects of trade out-
change in the unemployment rate is -.521. weigh its macroeconomic, accelerating ef-
C0o -1 Japan Ireland
- Finland
-
_-
-5 - Italy
C
Austria ~~~~~~~~~France
7
-7
o Denmark
.0
X -9 - Germany*
NO *Netherlands
0 ~ + Canada
United States
0)
l11 Norway
.2 ~~~~Sweden
?, -13 - Sweden
Zealand
E.New ii-
Belgium
0 United Kingdom -* *Australia
-15 -
0 .2 .4 .6 .8 1.0 1.2
Change in Unemployment Rate (log), 1967-1992
in Manufacturing and Change in the Unemployement Rate: 17 OECD Countries, 1967 to
1992
fects. Thus, for example, trade surpluses are terns of trade specialization that the devel-
(net of positive deindustrialization)expected oped nations exhibited moving into the
to lead to the employment of additional re- 1970s have largely been maintained.'4 This
sources (labor) in manufacturing. Nations suggests that net manufacturedexports/GDP
that run large manufacturingtrade surpluses is likely to be more powerful as a predictor
should experience less deindustrialization of cross-national variation in deindustrial-
than nations that do not. The nonsignificant ization trajectories than it is as a predictor of
coefficient in Model 3 is interpretablein this within-country (intertemporal) variation. It
light. Absent controls for positive dein- also suggests that there should be a fairly
dustrialization, the positive effects of spe- substantial positive relationship between a
cialization in manufacturingare likely to be country's pattern of trade specialization
confounded by the accelerating effects of moving into the period under study and its
strong manufacturingtradeperformance.The subsequent experience with deindustrial-
full model, Model 6, supports this interpre- ization. Figure 3 confirms this suspicion:
tation: When both positive and negative Countries that were large net importers of
deindustrialization are accounted for, net manufactured goods moving into the 1970s
manufacturedexports/GDP has a significant (i.e., Australia, Norway, and New Zealand)
positive effect on manufacturing's share of experienced substantialdeindustrializationin
employment. the following two decades. The exceptional
Cross-nationally, there are large and re-
markablystable differences in the patternsof 14 The average change in net manufacturedex-
trade specialization exhibited by the OECD ports over the 25-year period studied here was
countries (Rowthorn and Wells 1987). Pat- only ?1.502 percent of GDP.
N - Ireland Japan
Finland
-3
' -5 - Italy
W . Austria
C France
o -7
o Denmark
.0
W 9 Germany*
* * Netherlands
0) Canada
X United States
a -11 - Norway
.S . Sweden
?, -13 -New Zealand
Belgium
O *Australia United Kingdom- *
-15- I I I
-20 -15 -10 -5 0 5 10
Net Manufactured Exports/GDP, 1967-1969
in Manufacturing and Average Net Manufactured Exports/GDP (1967 to 1969): 17 OECD
Countries
case here is Ireland, but it is an exception specialization thesis. Thus it appears that
that proves the rule. Ireland experienced a positive, negative, and trade-related dein-
substantial shift in its pattern of trade spe- dustrialization all have played important
cialization, from net importerto net exporter, roles in changes in manufacturing's share of
during the period under study, and thus its employment in the advanced industrial soci-
outlying status in Figure 3 is explainable. By eties over the past two to three decades.
contrast, countries that were net exporters of Models 4 and 5 explore the globalization
manufacturedgoods moving into the 1970s thesis. Model 4 examines the role of direct
(i.e., Japan,Germany,and Italy) experienced investment in deindustrialization. A signifi-
comparatively modest deindustrialization. cant negative relationship between direct in-
The United Kingdom moved from being a vestment outflow per capita (log) and the
net exporter of manufacturedgoods to a net percentage of the labor force in manufactur-
importer.Belgium, however, presents some- ing is observed. The highly significant nature
thing of a mystery as its patternof trade spe- of this relationship is perhaps somewhat sur-
cialization changed little over the 1967-1992 prising, given the criticisms of the popular
period-it was a net exporter of manufac- literatureon capital flight and the admittedly
tured goods moving into the 1970s and a net speculative status of the alternative theories.
exporter in the early 1990s. Nonetheless, it Nonetheless, the adjusted R2 of .517 and co-
too experienced severe deindustrialization. efficient of partial determinationof .136 sug-
Models 1, 2, and 3 confirm each of the gests that direct investment played an impor-
three theses regarding deindustrialization tant role in the deindustrialization experi-
suggested by Rowthorn and Wells (1987): enced by the 18 nations under study. This
the maturitythesis, the failure thesis, and the finding reflects the labor-displacing effect of
g9 -1 Ireland Japan
to
Finland
-
.3
Ad -5 Italy
Austria France
7
-7 .
2 - Denmark
0
-9 * Germany Netherlands
Canada
United States
0 -11 Norway
Cl * - - .Sweden
~-13-
New Zealand
Belgium
0
Australia* * United Kingdom
-15- I I I I I I I I
-10 10 30 50 70 90
Percentage-Point Change in Direct Investment Outflow per Capita, 1967-1992
in Manufacturing and Change in Direct Investment Outflow per Capita: 17 OECD Coun-
tries, 1967 to 1992
direct investment stressed in the "new inter- trade-inducedshifts in the demandfor skilled
national division of labor"and "globalization versus unskilled labor have been a major
of production"literatures,the effects of com- cause of deindustrialization. Wood (1994)
paratively high (and growing) levels of di- proposes that the substantial upswing in the
rect investment on domestic capital forma- flow of manufacturedgoods from the South
tion and investment, and the "wealth trap," to the North over the past two to three de-
which in the long term may turn capital ex- cades either discouraged new investment in
porters into rentier nations. manufacturing or directed it toward defen-
Direct investment is also related to cross- sive, labor-saving innovations that offset the
national variation in deindustrialization tra- cost advantage of southern imports. Both re-
jectories. Figure 4 plots the change in the sponses should reduce the demand for manu-
percentage of the labor force in manufactur- facturing labor. Model 5 lends strong support
ing against the change in direct investment to Wood's account (and to the common per-
outflow per capita across the 1967-1992 pe- ception that imports from developing coun-
riod. Countries that experienced growth in tries played an importantrole in the deindus-
direct investment outflow also experienced trialization of the developed world): south-
extensive deindustrialization. In sum, argu- ern import penetration/GDPhas a highly sig-
ments that link deindustrializationto the sub- nificant negative effect on the percentage of
stantial upswing in direct investment that the labor force in manufacturing (partial R2
many OECD countries have experienced in = .145). 15
recent years appear convincing.
Model 5 introduces southern import pen- 15 The number of observations in Models 5 and
etration/GDP (log) to test the argument that 6 falls to 391 because of missing data on the im-
c -1 Japan Ireland
Finland
-3
.
' -5 - Italy
E .
. +~~Frnc
France Austria
-7
o Denmark
0
.0
X " Germany*
0 ... Netherlands
R ,,,,,, +
a)
0)
Canada United States*
-11 -
o Norway
E Sweden
-13 -
An New Zealand
(U
United Kingdom* Australia* *Belgium

-15 III
.5 1.0 1.5 2.0 2.5 3.0 3.5
Percentage-Point Change in Southern Import Penetration, 1967-1992
in Manufacturing and Change in Southern Import Penetration/GDP: 17 OECD Countries,
1967 to 1992
Figure 5 relates change in southern import countries that experienced the largest in-
penetration to change in the percentage of creases in southern import penetration.
the labor force in manufacturing. Southern Model 6 includes all the variables exam-
import penetration/GDPis less impressive in ined thus far. The development quadraticre-
its accounting for cross-national variation in mains highly significant, as does the unem-
deindustrializationtrajectories.Although the ployment rate. Net manufactured exports/
expected inverse relationship is in evidence, GDP has a significant positive effect on the
there is substantial scatter around the main percentage of the labor force in manufactur-
trend.'6 Both the light deindustrializers (ex- ing. The fit of the full model (adjusted R2 =
cepting Ireland) and a number of countries .788) and the partial coefficient of determi-
that experienced substantial deindustrial- nation of the explanatory variables net of the
ization (e.g., the UK, New Zealand, Sweden) fixed effects (partial R2 = .558) are impres-
experienced only modest increases in their sive, indicating that the variables in Model 6
penetration by southern imports. Nonethe- account for a large part of the deindustrial-
less, both the evidence in Figure 5 and the ization in these 18 OECD nations.
regression results suggest that deindustrial- Models 4 and 5 support two main subsid-
ization has generally been most severe in iaries of the globalization thesis. Do these
results hold, net of positive, negative, and
port penetration measure. Specifically, trade data trade-related deindustrialization? Model 6
were missing for the years 1968-1971 for Aus- indicates that they do: Both direct investment
tralia and 1968-1975 for New Zealand. outflow per capita and southern import pen-
16 Compare the much stronger relationship in- etration/GDP have significant negative ef-
dicated in Wood (1994:206, fig. 5.10). fects on the percentage of the labor force in
manufacturing.Although the introduction of all persuasions. Only the final item presents
other regressors substantiallyreduces the co- a problem: It does not necessarily follow
efficients of the globalization variables, the that, absent failure, manufacturing'sshare of
results suggest an important, independent employment would be larger. Suppose that
role for globalization in deindustrialization, stronger manufacturing performance had
net of the effects of maturity, failure, and been accompanied by large increases in la-
specialization. bor productivity. While output might be
higher, the percentage of the labor force in
manufacturingmight have fallen even faster.
DISCUSSION
One can, of course, imagine a situation in
The regression results support each of the which stronger performance in manufactur-
hypotheses advanced, including the global- ing would be accompanied by only a small
ization thesis. Maturity, failure, specializa- increase in labor productivity, in which case
tion, and globalization all play a role in ex- manufacturing's share of total employment
plaining the deindustrialization experienced would be greater than it is at present. The
by developed societies in recent decades. point, however, is that it is difficult to choose
Somewhat more delicate, though, is the a priori between these two accounts of what
question of the broader meaning and inter- would have happened in the absence of fail-
pretation of the results. Critical accounts of ure (Rowthorn and Wells 1987).
deindustrialization that stress the role of de- The same considerations permeate discus-
clining international competitiveness and sions of the impact of direct investment and
foreign investment and trade tend to assume trade on manufacturing employment. Such
that, absent these factors, deindustrialization discussions often assume that, absent capital
would have occurred far more slowly. flight or North-South trade, the percentage
Although the regression results are consis- of the labor force in manufacturing would
tent with this assumption, they are not con- be considerably larger than it is at present in
clusive. the advanced industrial societies. However,
Consider the failure thesis. In their discus- this conclusion is sensitive to the assump-
sion of the deindustrialization of Britain, tions made regarding what would happen in
Rowthorn and Wells (1987) summarize the the absence of direct investment or North-
failure thesis in the following set of proposi- South trade. For instance, does direct in-
tions: vestment substitute for, or add to, other
(I)The U.K.'s economic record in the realm of forms of investment (Hufbauer and Adler
incomes and employment has been poor; 1976; letto-Gillies 1992)? Grant that direct
investment represents a real loss of produc-
(2) This is largely due to the weak performance tive capacity in the investing country that
of U.K. manufacturingindustry; has direct implications for manufacturing
(3)If the performance of U.K. manufacturing employment and its share of total employ-
industry had been much stronger, U.K. ment. But on what basis can one make the a
manufacturing output would have been priori claim that the decline of manufactur-
much greater; ing employment would have been smaller
(4)This would have stimulated the non-manu- absent this loss of productive capacity?
facturing side of the economy and led to the The effects of the absence of heightened
creation of more employment in services international competition or of the foreign
and other non-manufacturingactivities; investment and trade upswings are not obvi-
(5)Finally, if U.K. manufacturing output had ous. Deindustrialization might have pro-
been higher, neither the absolute number of ceeded at a slower pace, or it might have
people employed in manufacturing, nor this proceeded even more rapidly. One way to
sector's share in total employment, would address this issue is to engage the appropri-
have fallen anything like as fast as they ate historical counterfactuals.'7 Consider,
have done. (P. 221)
Most of these propositions, Rowthorn and 17 Sociologists generally have paid little atten-
Wells (1987:221) note, are uncontroversial tion to counterfactuals (Lieberson 1985). Al-
and would be accepted by social scientists of though counterfactuals are arguably inherent to
first, the finding that the deindustrialization gression results might seem to lend support
of the industrial countries is in large part an to calls for greater protection in response to
example of negative deindustrialization. globalization, the counterfactuals suggest
What would have happened to manufactur- that the erection of barriers to southern im-
ing employment if growth had not slowed ports or the imposition of controls on direct
(and unemployment risen) at the close of investment are unlikely to have "saved" the
the postwar Golden Age? Absent failure, vast majority of "lost" manufacturing jobs
would manufacturing's share of employ- (and this before the potential costs of protec-
ment have fallen much more slowly? Like- tionism are factored in).
wise, consider the finding that globalization
has made a measurable contribution to dein-
dustrialization. What would have happened CONCLUSIONS
to manufacturing employment if both the A major goal has been to assess the role of
outflow of direct investment and inflow of globalization in deindustrialization. While a
southern imports had not exploded over the large literature suggests direct links between
past few decades? Absent globalization, globalization and the decline of industrial
would deindustrialization have been consid- employment in the developed world, such ar-
erably more modest? guments have not inspired much empirical
Appendix A provides some preliminaryan- research by sociologists. This is unfortunate
swers to these questions by using the regres- because the issues of globalization and dein-
sion results to approximate the above condi- dustrialization impinge directly on an array
tions. Such exercises are, of course, specula- of core sociological concerns, including so-
tive. Nonetheless, they can quantify, within cial stratification, the sociology of the labor
wide bounds, the relative influence and im- force, and political sociology.
portance of the factors in question. The re- I find support for arguments that link
sults indicate that the net impact of failure deindustrialization to the outflow of direct
and of globalization on manufacturing em- investment-the upswing in direct invest-
ployment-defined as the difference between ment experienced across the advanced indus-
the actual change in percent manufacturing trial world has significantly affected manu-
employment and the change that would have facturing employment. The results lend sup-
occurred absent failure and globalization- port to arguments that stress: (1) that direct
is significantly more modest than is assumed investment displaces those employed in
in critical treatments of deindustrialization. manufacturing; (2) that direct investment
This points out the utility of explicitly engag- may raise the required marginal rate of re-
ing the counterfactuals deeply embedded in turn on domestic investment, shift invest-
so much of the thinking on globalization, es- ment from manufacturingto services, and re-
pecially given the policy implications of orient investment away from real investment
such research. For instance, although the re- toward purely financial investment; and (3)
that direct investment may, over the long
sociological research, discussion of them tends to
term, move a nation's economy into a
be confined to programmatic statements in his- "wealth trap."Given the speculative status of
torical sociology (Tilly 1996) and exercises in many of these arguments,additional work on
formal theory (Martin and Quinn 1996; see these issues is clearly needed.
Brooks 1999 for an exception to this generaliza- I also find support for arguments that im-
tion). Among historians, however, there has re- plicate North-South trade in manu-
cently been a resurgence of interest in counter- facturing's declining share of total employ-
factuals (Hawthorn 1991; Tetlock and Belkin ment in the North. The significant negative
1996). In other social sciences, counterfactuals effect of southern import penetration con-
are seen as an importantaccompaniment to quan-
titative research on a variety of policy-relevant
firms Wood's (1994) argumentregarding the
topics, including income distribution (Dinardo impact of trade-induced shifts in the de-
and Lemieux 1997), economic development (Bar- mand for skilled versus unskilled labor (as
row and Hall 1995), and poverty (Squire 1995). well as the common perception that imports
My own approach (see Appendix A) follows in from developing countries have played an
this tradition. important role in the deindustrialization of
the developed world). The findings are con- the critical thinking on globalization. Many
sistent with Wood's characterization of treatments of deindustrialization assume
North-South trade as (1) reducing the profit- that, absent globalization and the "troubles"
ability of northern manufacturing; (2) de- that began to plague Northern industry by
pressing the rate of new investment in la- the close of the 1960s, manufacturing em-
bor-intensive sectors; and (3) directing new ployment in OECD countries would be much
investment in manufacturing toward defen- higher than it is at present. The counter-
sive labor-saving innovations. While the factuals explored in Appendix A reveal that
analysis reveals that the role of North-South a significant reduction in the percentage of
trade has been substantially overestimated, the labor force in manufacturing in the ad-
the results clearly suggest that such trade vanced industrial societies was practically
has had a growing impact on manufacturing unavoidable, as almost all of the OECD
employment in the North. countries had reached "maturity" by the
In sum, the results indicate that globaliza- close of the 1960s. Owing to this economic
tion has played an important, independent success, deindustrialization in the post-1970
role in the deindustrialization of advanced period would have been considerable even if
industrial countries. The results further es- the performance of manufacturing had been
tablish that: (1) manufacturing's share of strongeror the upswings in direct investment
employment follows a curvilinear, inverted and southern imports had not occurred.
U-shaped path over the course of develop-
ment; (2) "negative deindustrialization"ow- Arthur S. Alderson is Assistant Professor of So-
ing to poor industrial performance has ciology at Indiana University. He continues his
played a large role in the deindustrialization research on the globalization of production, ex-
of the OECD countries; and (3) a country's ploring both its causes and consequences. He is
also studying income inequality, the structure of
overall pattern of trade specialization has an
the world city system, labor market institutions
importanteffect on its employment structure. and unemployment, and modernity and time. He
Interpretation of the regression results is the author, with Francois Nielsen, of "Income
should be tempered by the recognition that Inequality, Development, and Dependence: A Re-
the findings are silent on the counterfactual consideration" (American Sociological Review,
assumptions that are embedded in much of 1999, vol. 64, pp. 606-31).
Appendix A. Two Key Counterfactuals
Accounts of deindustrialization that stress the role growth (and rise in unemployment) that began in the
of declining international competitiveness and for- OECD countries in the early 1970s never happened.
eign investment and trade tend to assume that, ab- Under this model, real GDP per capita (D) begins at
sent these factors, deindustrialization would have its actual level in 1967 and grows at a steady rate of
proceeded at a considerably more modest pace in the 3.5 percent per year (the actual average rate of
OECD countries. To explore this assumption, I en- growth in the 18 OECD countries was 2.1 percent
gage two key counterfactuals. I begin with the fol- over the 1967-1991 period). Under this success
lowing equation: model, D would have been more than $3,400 higher
M = 29.846 + 250.527D - 32.963D2 - 6.591U by 1991 than it was ($17,269 versus $13,856 in 1985
prices). Given a growth rate typical of the immedi-
+.05 8E -.505I - 4. 132S,
ate postwar period, the unemployment rate U does
where M is the percentage of the labor force in man- not rise, but remains at its 1967 level of 3.1 percent
ufacturing, D is the logarithm (base 10) of real GDP throughout the 1967-1991 period. All other vari-
per capita, U is the unemployment rate (log), E is ables take their actual values over time.a In sum, the
net manufactured exports/GDP, I is the direct invest- "success model" is estimated as follows:
ment outflow per capita (log), and S is southern im-
port penetration/GDP (log). This is the same equa- a See Scenario I in Rowthorn and Wells (1987:228).
tion estimated in Model 6, but I average across the Trade would likely be affected under the "success
country-specific intercepts to produce an overall model." Unfortunately, it is unclear what a reasonable
constant. estimate of the impact of high growth and low unem-
Estimating what would have happened absent ployment would look like. Inasmuch as "success" im-
failure involves estimating M under conditions of a plies a positive feedback resulting in an improved
"success model." I assume that the slowdown in manufactured trade balance, factoring in this possibil-
035(1-1967) 2
Mt = 29.846 + 250.527Dt - 32.963D; - 6.59 lUt
Mt = 29.846 + 250.527(D1967
+.058Et - .5051967 -4 132SI967 .
-32.963(Df; )2 659
e.035(t-1967)2 197
-66.591(U1967) Although the percentage of the labor force in
-
+.058(Et ) -.505(It ) 4. 132(St ). manufacturing fell by 8.7 points, in the absence of
In reality, the average percentage employed in globalization it would have fallen by 5.8 points.
manufacturing declined by 8.7 points over the 1968- Thus, deindustrialization would have been far from
1992 period. In the absence of failure (i.e., under the negligible in the absence of globalization.d To dis-
success model), the percentage in manufacturing tinguish between the effects of direct investment and
would have declined by 7.3 percentage points. Thus trade, I also explore two subsidiary counterfactu-
the percentage of the labor force in manufacturing als-the "no direct investment" and "no southern
would have fallen fairly far even if the performance import penetration" models.
of the manufacturing sector in the average OECD Under the "no direct investment model," south-
country had been much stronger. So while the de- ern import penetration/GDP takes its actual value
cline of manufacturing in the OECD countries is in while direct investment outflow per capita is held at
large part attributable to negative deindustrializa- its 1967 value. Absent the direct investment up-
tion, the assumption that there would have been ap- swing, the percentage of the labor force in manufac-
preciably less deindustrialization had industrial per- turing would have declined by 7.1 percentage points
formance been stronger is clearly called into ques- from 1968 to 1992. Those concerned with the effects
tion. A substantial decline in the relative size of the of direct investment clearly assume that deindustri-
manufacturing labor force in the post-1970 period alization would not have proceeded at anything like
was practically unavoidable, as all of the OECD the pace that it has in the past two to three decades
countries except Ireland had reached maturity by the in the absence of capital flight (Bluestone and Har-
end of the 1960s (Table 2) b rison 1982). While the regression results suggest
To explore what would have happened to manu- that direct investment has an effect on manufactur-
facturing employment if both the outflow of direct ing's share of employment, the counterfactual sug-
investment and inflow of southern imports had not gests, again, that substantial deindustrialization was
risen so dramatically in the contemporary period, I likely to occur whether or not there was an upswing
estimate M under a simple "no globalization model" in direct investment.
in which direct investment outflow per capita and The "no southern import penetration model" sets
southern import penetration/GDP are held at their southern import penetration/GDP to its 1967 value
1967 values (2.15 percent and 1.47 percent, respec- while allowing direct investment outflow per capita
tively), while all other variables take on their actual and the other variables to take their actual values.
values.c The "no globalization model" is estimated Had manufactured imports from developing societ-
as follows: ies not experienced the notable upswing they did, the
percentage of the labor force in manufacturing would
ity would make the role of failure appear even more have declined by 6.3 percentage points from 1968 to
modest. For this exercise, I assume no effect, recogniz- 1992. This indicates a more modest role in deindus-
ing that this may be an unrealistic assumption. trialization for North-South trade than that suggest-
bThese conclusions directly parallel those drawn in ed by Wood (1994:13), who attributes more than
Rowthorn and Wells' (1987) analysis of the deindus- one-half of the decline in manufacturing employment
trialization of Britain. to trade shifts. None of the evidence examined here
I Some researchers have linked globalization to the
suggests such a large role for North-South trade. Al-
recent emergence of high and persistent unemployment though North-South trade plays an important part in
in many developed countries (OECD 1994b). Attribut- deindustrialization, the assumption that deindustri-
ing some proportion of the rise in unemployment from alization would have proceeded at a far more modest
1967 to 1991 to globalization suggests a more promi- pace absent the upswing in manufactured imports
nent role for globalization than is indicated below. Un- from the South appears to be unfounded.
fortunately, estimates of the overall employment effects
of direct investment and North-South trade vary widely, d I do not attempt to incorporate the positive feed-
but many analysts argue that their net effects are near back to growth that some critics of globalization imply
zero (e.g., Dicken 1986; Frank and Freeman 1978; Gra- would follow if globalization were reined in. Given that
ham and Krugman 1991; Krugmanand Lawrence 1993; incorporating such an effect would make the impact of
OECD 1994b). I adopt this position for the purposes of globalization appear more modest, this is a liberal test
this exercise. of the counterfactual.
REFERENCES
Alderson, Arthur S. 1997. "Globalization, lina, Chapel Hill, NC.
Deindustrialization, and the Great U-Turn: The Bajo-Rubio, Oscar and Simon Sosvilla-Rivero.
Growth of Direct Investment in 18 OECD 1994. "An Econometric Analysis of Foreign
Countries, 1967-1990." Ph.D. thesis, Depart- Direct Investment in Spain, 1964-1989."
ment of Sociology, University of North Caro- Southern Economic Journal 61:104-19.
Barrow, Michael and Mike Hall. 1995. "The Im- Industrial Economics: Issues and Perspectives.
pact of a Large Multinational Organization on 2d ed. London, England: Macmillan.
a Small Local Economy." Regional Studies Frank, Robert H. and Richard T. Freeman. 1978.
29:635-53. Distributional Consequences of Direct Foreign
Beenstock, Michael. 1984. The World Economy Investment. New York: Academic.
in Transition. 2d ed. London, England: Allen Graham,Edward M. and Paul R. Krugman. 1991.
and Unwin. Foreign Direct Investment in the United States.
Bell, Daniel. 1973. The Coming of Post-Indus- Washington, DC: Institute for International
trial Society. New York: Basic Books. Economics.
Belsley, David A., Edwin Kuh, and Roy E. Greene, William H. 1992. Limdep Version 6.0
Welsch. 1980. Regression Diagnostics: Identi- User's Manual and Reference Guide. New
fying Influential Data and Sources of Col- York: Econometric Software
linearity. New York: Wiley. . 1993. Econometric Analysis. New York:
Bluestone, Barry and Bennett Harrison. 1982. The Macmillan.
Deindustrialization of America. New York: Hall, Peter. 1993. "Forces Shaping Urban Eu-
Basic Books. rope." Urban Studies 30:883-98.
Bollen, Kenneth A. and Robert W. Jackman. Harrison,Bennett and Barry Bluestone. 1988. The
1985. "Regression Diagnostics: An Expository Great U-Turn. New York: Basic Books.
Treatment of Outliers and Influential Cases." Hausman, Jerry A. 1978. "Specification Tests in
Sociological Methods and Research 13:510- Econometrics." Econometrica 46:1251-71.
542. Hawthorn, Geoffrey. 1991. Plausible Worlds:
Breusch, Trevor S. and Adrian R. Pagan. 1979. Possibility and Understanding in History and
"A Simple Test for Heteroskedasticity and the Social Sciences. New York: Cambridge
Random Coefficient Variation." Econometrica University Press.
47:1287-94. Hildreth, Clifford and James P. Houck. 1968.
Brooks, Clem. 1999. "Civil Rights Liberalism "Some Estimators for a Linear Model with
and the Suppression of a Republican Political Random Coefficients." Journal of the Ameri-
Realignment in the U.S., 1972-1996." Ameri- can Statistical Association 63:584-95.
can Sociological Review. Forthcoming. Hoogvelt, Ankie. 1997. Globalization and the
Chirot, Daniel. 1986. Social Change in the Mod- Postcolonial World. Baltimore, MD: Johns
ern Era. New York: Harcourt, Brace, Hopkins University Press.
Jovanovitch. Hsiao, Cheng. 1986. Analysis of Panel Data. New
Clark, Colin. [1940] 1957. The Conditions of York: Cambridge University Press.
Economic Progress. 3d ed. London: Hufbauer, Gary C. and F. Michael Adler. 1976.
Macmillan. Overseas Manufacturing Investment and the
Coates, David and John Hillard. 1986. The Eco- Balance of Payments. New York: Arno.
nomic Decline of Modern Britain: The Debate Hymer, Steven. 1979. The Multinational Corpo-
Between Left and Right. Brighton, England: ration: A Radical Approach. Cambridge, En-
Wheatsheaf. gland: Cambridge University Press.
Cowling, Keith. 1986. "The Internationalization letto-Gillies, Grazia. 1992. International Produc-
of Production and Deindustrialization."Pp. 23- tion: Trends, Theories, Effects. Cambridge,
31 in Technological Change, Industrial Re- England: Polity.
structuring, and Regional Development, edited International Monetary Fund (IMF). 1977. Bal-
by A. Amin and J. Goddard.London, England: ance of Payments Manual. 4th ed. Washington,
Allen and Unwin. DC: InternationalMonetary Fund.
Cowling, Keith and Roger Sugden. 1987. . Various years. Balance of Payments Sta-
Transnational Monopoly Capitalism. New tistics Yearbook. Washington, DC: Interna-
York: St. Martin's. tional Monetary Fund.
Dicken, Peter. 1986. Global Shift. London, En- Klein, Michael W. and Eric Rosengren. 1994.
gland: Harper and Row. "The Real Exchange Rate and Foreign Direct
Dinardo, John and Thomas Lemieux. 1997. "Di- Investment in the United States." Journal of In-
verging Male Wage Inequality in the United ternational Economics 36:373-89.
States and Canada, 1981-1988: Do Institutions Koopmans, Lambert H. 1987. Introduction to
Explain the Difference?" Industrial and Labor Contemporary Statistical Methods. Boston,
Relations Review 50:629-51. MA: Duxbury Press.
Erickson, Christopher L. and Sarosh Kuruvilla. Krugman,Paul R. and Robert Z. Lawrence. 1993.
1994. "Labor Costs and The Social Dumping "Trade,Jobs, and Wages." National Bureau of
Debate in the European Union." Industrial and Economic Research Working paper no. 4478,
Labor Relations Review 48:28-47. National Bureau of Economic Research, Cam-
Ferguson, Paul R. and Glenys J. Ferguson. 1994. bridge, MA.
Krugman, Paul R. 1996. Pop Internationalism. DC: U.S. Bureau of the Census.
Cambridge, MA: MIT Press. Singh, Ajit. 1977. "UK Industry and the World
Lieberson, Stanley. 1985. Making It Count: The Economy: A Case of Deindustrialization?"
Improvement of Social Research and Theory. Cambridge Journal of Economics 1:113-36.
Berkeley, CA: University of California Press. . 1989. "ThirdWorld Competition and De-
Lorence, Jon and Joel Nelson. 1993. "Industrial Industrialization in Advanced Countries."
Restructuring and Metropolitan Earnings In- Cambridge Journal of Economics 13:103-20.
equality, 1970-1980." Pp. 145-84 in Research Squire, Lyn. 1995. "Evaluating the Effectiveness
in Social Stratification and Mobility 12, edited of Poverty Alleviation Programs."New Direc-
by R. Althauser and M. Wallace. Greenwich, tions for Program Evaluation 67:27-37.
CT: JAI. Stopford, John M. and Louis Turner. 1985. Brit-
Marglin, Stephen A. and Juliet B. Schor. 1990. ain and the Multinationals. Chichester, En-
The Golden Age of Capitalism. New York: Ox- gland: Wiley.
ford University Press. Summers, Robert and Alan Heston. 1991. "The
Martin, Andrew D. and Kevin M. Quinn. 1996. Penn World Table (Mark 5): An Expanded Set
"Using Computational Methods to Perform of International Comparisons, 1950-1988."
Counterfactual Analysis of Formal Theories." Quarterly Journal Of Economics 106:327-68.
Rationality and Society 8:295-323. Swamy, P. A. V. B. 1971. Statistical Inference in
Messerlin, Patrick A. 1995. "The Impact of Trade Random Coefficient Regression Models. New
and Capital Movements on Labor:Evidence on York: Springer-Verlag.
the French Case." OECD Economic Studies Tanaka, Kunikazu. 1991. "On the Effects of For-
24:89-124. eign Investment: A Consideration of the Pro-
Neter, John, William Wasserman, and Michael H. cess of Deindustrialization in Connection with
Kutner. 1989. Applied Linear Regression Mod- the Trade Balance and the Patterns of DFI Fi-
els. Burr Ridge, IL: Irwin nancing." Kobe University Economic Review
Nielsen, Franqois and Arthur S. Alderson. 1997. 37:61-84.
"The Kuznets Curve and the Great U-Turn: In- Tetlock, Philip E. and Aaron Belkin. 1996.
come Inequality in US Counties, 1970 to 1990." Counterfactual Thought Experiments in World
American Sociological Review 62:12-33. Politics. Princeton, NJ: Princeton University
Organization for Economic Cooperation and De- Press.
velopment (OECD). 1984, 1994a, and various Tilly, Charles. 1996. "Invisible Elbow." Socio-
years. Foreign Trade by Commodities. Paris, logical Forum 11:589-601.
France: OECD. United Nations Industrial Development Organi-
. 1989. Labour Force Statistics, 1967- zation (UNIDO). 1983. Industry in a Changing
1987. Paris, France: OECD. World. New York: United Nations.
. 1994b. The OECD Jobs Study: Facts, U.S. Bureau of Labor Statistics. 1993. "Interna-
Analysis, Strategies. Paris, France: OECD. tional Comparisons of Manufacturing Produc-
. 1995. Labour Force Statistics, 1973- tivity and Unit Labor Cost Trends, 1993."
1993. Paris, France: OECD. Washington, DC: U.S. Bureau of Labor Statis-
. Various years. National Accounts. Paris, tics.
France: OECD. Wallace, Michael and Joyce Rothschild. 1988.
Pain, Nigel. 1993. "An Econometric Analysis of "Plant Closings, Capital Flight, and Worker
Foreign Direct Investment in the United King- Dislocation: The Long Shadow of Deindus-
dom." Scottish Journal of Political Economy trialization." Pp. 1-35 in Research in Politics
40:1-23. and Society, vol. 3, edited by M. Wallace and
Petty, Sir William. 1899. The Economic Writings J. Rothschild. Greenwich, CT: JAI.
of Sir William Petty. Vol. 1, edited by C. H. Waters, Malcolm. 1995. Globalization. New
Hull. Cambridge, England: Cambridge Univer- York: Routledge.
sity Press. Wilkinson, Leland. 1990a. SYSTATT: The System
Rowthorn, Robert E. and John R. Wells. 1987. for Statistics. Evanston, IL: SYSTAT, Inc.
De-Industrialization and Foreign Trade. Cam- 1990b. SYGRAPH: The System for
bridge, England: Cambridge University Press. Graphics. Evanston, IL: SYSTAT, Inc.
Ryscavage, Paul, Gordon Green, and Edward Wood, Adrian. 1994. North-South Trade, Em-
Welniak. 1992. "The Impact of Demographic, ployment, and Inequality. Oxford, England:
Social, and Economic Change on the Distribu- Oxford University Press.
tion of Income." Pp. 11-30 in U.S. Bureau of World Bank. Various years. World Tables. Balti-
the Census, Series P 60, No. 183. Washington, more, MD: Johns Hopkins University Press.

Alderson 1999

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Alderson 1999

Uploaded by

Copyright:

Available Formats

Explaining Deindustrialization: Globalization, Failure, or Success?

Author(s): Arthur S. Alderson

Although sociologists have expressed growing interest in globalization, they

American Sociological Review, 1999, Vol. 64 (October:701-721) 701

Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Real GDP per capita (log)a 532.75 -- 250.53

Adjusted R2 .432 .553 .252 .517 .542 .788

C0o -1 Japan Ireland

United Kingdom* Australia* *Belgium

Appendix A. Two Key Counterfactuals

You might also like