Globalization and Deindustrialization: Direct lnvestment and the Decline of Manufacturing Employment in 17 OECD Nations*

Recent years have witnessed a fairly dramatic upswing in the level of foreign direct investment, a phenomenon which ha..:; played an integral part in a larger process of globalization. While sociologists have devoted a good deal of attention to the consequences of direct investment for the developing hosts of foreign direct investment, much less attention ha..:; been paid to the implications of direct investment for the advanced industrial socictics. ln this paper, I focus on one of the more interesting links that ha..:; been drawn between direct investment and its effects: that between the outflow of direct investment often ca..:;t a..:; "capital flight" and dcindustrialization . To examine this link I employ a pooled time-series of cross-sections data'let which combines observations on 17 OECD nations across the 1967-1990 period (N=408) . Random effects regression models, which control for unmea..:;urcd country-specific effects, reveal strong support for arguments which link direct investment to the relative decline of the labor force in manufacturing in core socictics. ln addition, results show that dcindustriali zation across this period is largely explained by a model that combines cla..:;sic generalizations of the process of economic development with an attention to a range of more immediat e factors identifi ed by contemporary students of dcindustriali zation. *Direct all correspondence to Arthur S. Alderson, Department of Sociology, CB# 3210, University of North Carolina, Chapel Hill, NC 27599.Email:Art_Aldcrson @unc.edu . An earlier version of this paper wa..:; presented at the 1996 meeting of the American Sociological A..:;sociation in New York, NY. I thank Franc;ois Nielsen , Rachel Rosenfeld, Ken Bollen, and the JWSR editor and reviewer for helpful advice. [Page 1] Journal of World-Systems Research


INTRODUCTION
Recent years have witnessed a fairly substantial upswing in the level of direct investment . 1Following the global recession of the early 1980s (and a consequent downturn in direct investment), total outflows from seventeen OECD nations grew from 27 billion US dollars in 1982 to over 219 billion by 1990.2.The average annual rate of growth between 1982 and 1990 wa<; roughly 31 percent.This contra<;ts with an earlier period of slower, yet still impressive, growth in the outflow of dir ect investm ent from the OECD nations.Following the global recession of the early 1970s, for instanc e, outflows of direct investment from these same seventeen nations grew from 2 1 billion US dollars in 1974 to 53 billion by 1980, at an average annual rate of about 17 percent.
The growth of direct investment ha<; played an integral part in a larger process of globalization or internationalization, a process which ha<; captured the attention of analysts of diverse perspective and discipline (e.g.Ictto-Gillics, 1992;Robertson, 1992;King, 1991;Featherstone, 1990;Giddens, 1990;Albrow and King, 1990;Cha<;c -Dunn, 1989;Harvey, 1989;Lash and Urry, 1987).This interest ha<; been fueled by th e sense that the most recent round of globalization, which finds its origins -according to a variety of authors -in the late 1960s, ha<; exhibited a number of unique features and raised a number of profound questions, questions concerning everything from the repr esentation of identity to the sovereignty of the nation-state .Although "globali zation" (a<; noun) only emerged a<; a significant concept in academic circles a decade or so ago, it ha<; become in that short period of time a subject of intense scholarly and public interest (Robertson, 1992: 8).
While sociolo gists have devoted a great deal of attention to the consequences of foreign direct investment for the periphery (e.g.Bornschicr and Cha<;c-Dunn, 1985;London, 1988;Bosw ell and Dixon, 1990;Wimb erly, 1990), they have devoted much less empirical attention to often voiced concerns over the impact of the growth of dir ect investment on core societies.One such concern surrounds the impact of the heightened geographic mobility of capital on traditionally high-wage manufacturin g emplo ym ent .Across the 1970s and 1980s, all of the major industrial nation s experienced a decline in the relative size of their manufacturin g labor forces .At one extreme stand nations such as the UK, which moving into the 1970s had over 33 percent of its labor force in manufacturing and saw this decline to under 20 percent by 1990.At the other stands Japan, which saw its manufacturin g labor force decline by only 3 percent since the early 1970s .On average, the share of manufacturing employm ent in the seventeen OECD nations noted above declined from 27 percent in 1967 to 19 percent by 1990 (Figure 1)."Dcindustri alization" ha<; thus been general, if not uniform, across the core in the la<;t twenty-five ycars.   .Percent labor force in manufacttrring in 1he UK and Japan and tl1e average size of the manufacttiring labor force in l7 OECDnations.

UJRE(T I~"VESTMENT: TRENDS AND l'A ITERNS
Data on total inflows and ontflows of direct invcstmc.-nt(hc.TCaikT DI) for the 17 OECD nations noted above arc prcsc.-ntcd in l'igure 2. In addition to noting the stcc.-pg'l\1Wth in outflows of DI following the recession of the ciirly eighties.one cim also note the parallel g'l\1Wth in inflows !l(,'l\1SS the same pc.Tiod.The share of inflows going to ihc developing world dce,TCascd frnm an annual avc..!!l 0 50 1961 1910 1913 1916 1919 1982 1985 1988 Year ---DI OUT DIIN For the sociologist of international development, whose papers arc usuall y heavily weighted down with citations to a literature stressing the central, and often negative, role of the multinational enterprise (hereafter MNE) in everything from economic growth to the demographic transition in the developing world, it may come a<; something of a surprise that the relative importance of the developing world for core MNEs has been declining over time.§ It is important to note, however, that this shift in the global distribution of DI that occurred across the 1970s and 1980s is part of a longer-t erm process (Dunning, 1988;Magdoff, 1992).While estimates ofthc stock of accumulated foreign direct investment in earlier periods must be approached with a degree of caution, Table 1 indicates that the developing world's share of DI ha<; been generally declining across the twentieth century.Lying behind this shift away from the developin g world has been a shift in the sectoral composition of direct investment across the twentieth century.The bulk of DI befor e th e Second World War was devoted primarily to agriculture and raw mat erials.Following the Second World War there was a shift towar d manufacturin g DI, a form of DI which has always been disproport ionatel y sited in the developed world (Dunning, 1988).Since the 1960s, manufacturing has become the dominant sector for DI.It is important to note, however, that the 1980s have witnessed a significant increase in DI in services, particularly finance and trade related services (UNCTC , 1988).Rather than representin g a displacement of manufactur ing DI, though, the increased transnationali zation of services "has led to increased [DI] in both manufacturin g and services" (Ictto-Gillies, 1992: 26); that is, it has further facilitated the servicing of foreign markets by manufacturing MNEs.Service sector DI, as with manufacturing DI, has tended to be disproportionately sited in developed societies.

[Page 5] Journal of World-Systems Research
In addition to these changes in the global distribution and sectoral composition of direct investment across the twentieth century, there have been major changes in the lineup of actors on the international investment scene.Before the Second World War, Britain wa.:; by far the largest direct investor.It is estimated, for instance, that it held roughly 46% of the world's accumulated stock of DI in 1914(Dunning, 1988: 46).Following WWII, the US quickly rose to a position of dominance and had by 1960 attained the sort of hegemonic position which Britain had enjoyed in the decades surrounding the turn of the twentieth century, holding 48% of the world's stock of DI.In the la.:;t two decades or so, the situation ha.:; become much more diverse.Britain remains an important home and host to DI.Most of the other industrial societies have experienced an increase in their relative share of total world outflows of DI, notably Japan, France, Germany, and Sweden (OECD, 1987;UNCTC, 1988).Most of this incrca.:;ccame at the "expense" of the US's position a.:; it experienced a simultaneous incrca.:;c in inflows and dccr ca.:;c in outflows, becoming a net DI importer in 1981 and remaining one thereafter.Table 2 gives one a sense of the situation that currently prevails in the industrialized world relative to an earlier period of American hegemony.As one can note, the US's share of the total DI flowing out of the 17 nations compared in Table 2 dccrca.:;cdrather substantially over the period from 1967-1990 while its share of the total DI inflows to these same nations increa.:;cdmarkedly.Thus outflows of DI from th e major investing nation s become more evenly distributed while inflows have become skewed toward the US.

Journal o.l World-Systems R esearch
Weighting the flows to and from the individual nations in Table 2 by gross domestic fixed capital formation (GDFCF) gives one a sense of the relati ve importance of direct investment for their cconomics. 1Once normalized, a somewhat different picture emerges.
Column five in Table 3 combines inflows and outflows as a percentage ofGDFCF and averages these over the 1967 -1972 period.The point of combining inflows and out flows is to get a general picture of the importance of international production for any give n soci ety.A.., one can not e, although the US wa.., host to 15 % and hom e to 60% of th e DI flowing into or out of the 17 nations under consideration in 1967 -1972 (Table 2), this amounted to a comparatively modest proportion of US dom estic fixed investm ent activity, roughly 5%.Ind eed, relativ e to the size of their economics, DI wa.., m ore important for Australia, Belgium, Canada, Holland, and the UK than it wa.., for the US.Column 5 is disaggr egate d in columns 1 and 3. Herc we sec that only the Dut ch and the British were proportionally larger expo rters or homes to DI th an the US; the Australi ans, Belgians, and Cana dians being net import ers of DI.Column 6 combines infl ows and outflows over the 1985-1990 period.When compared with column 5, the most striking fact that emerges here is the substantial and often dram atic incrca..,c in the imp ortance of international production for all of the nations under consideration save one (Ireland).DI inflow and outflow is now equivalent to over 5% ofGDFCF in 13 of the 17 nations consid ered in this period, and equivalent to 10% or more in 7 of the se 13.Britain, Holl and, and Belgium again appear to be particularly heavil y invo lved in international production , DI now being equivalent to nearly a third ofGDFCF.Column 6 is disaggre gated in columns 2 and 4. Compari ng columns 2 and 4 to columns 1 and 3, one can note that while only Finland, Japan , the Netherlands, Sweden, Brit ain, and the US were net exporters of DI in the earlier per iod, the majori ty arc in the lat er period.In addition to the US, whose somewhat peculiar behavior wa.., noted ab ovc, onl y Australia, Austria , Belgium, Irel and, and New Zealand remain net import ers.But even these nations , with the exception of Ireland, have seen subst antial growth in their outflow of DI.International Financial Statistics (I 979, 1989); World Bank World Tables (1994).
There arc, then, a number of longer and shorter-term patterns and trends which come together to make the contemporary period both interesting and unique.Direct investment is of growing importance for almost all of the developed societies.Most DI originates in the developed world and an incrca~ingly large proportion of it is sited there a~ well.In terms of its sectoral composition, most DI in the contemporary period is directed toward the manufacturing sector, but in recent years a growing proportion ha~ been directed toward services.Most of the industrial societies have moved over the pa~t two decades to become net exporters of DI, while the US ha~ moved to become a net importer, receiving nearly half of the DI flowing into the 17 nations in Table 2 over the 1985-1990 period.All of these factors point, a~ the NIDL and GOP accounts of the contemporary worldcconomy have stressed, to the incrca~ing importance of the MNE and its activities for core societies.

THE DIRECT INVESTMENT-DEINDUSTRIALIZATION THESIS
The idea that direct investment has contributed in an important way to the phenomenon of deindustrialization is, at least in the English-speaking world, a fairly general one.~In two widely cited pieces, for instance, Bluestone and Harrison have set out one of the more popular versions of this argument (Bluestone and Harrison, 1982;Harrison and Bluestone, 1988).They argue that DI in the contemporary era is being undertaken as part of a "globalization gambit."This move constitutes an integral part of a new set of corporate strategies designed to abrogate the old post-war social contract between capital, labor, and the state and, in doing so, to restore acceptable levels of profitability in response to the "profit squeeze" of the 1970s.The result of this strategy, they argue is a kind of hollowing of the economy.By hollowing, they mean in part that DI is no longer being undertaken by firms in an effort to complement domestic investment and production, but to replace it.As firm after firm in the manufacturing sector has gone abroad in search of lower labor costs, the end result of the growth of DI in the last two decades has been deindustrialization.While careful to note that DI is not the sole cause of deindustrialization, they nonetheless maintain that it is "certainly a major reason that the United States lost a significant fraction of its manufacturing base" (1988: 29).

Journa I of World-Systems Research
While Bluestone and Harrison focus most intently on the US, other studies paint a similar picture of the role ofMNEs and DI in deindustriali zation in other core societies.Stopford and Turner (1985), for instance, show that roughly a third of all manufacturing jobs lost in the UK between l 972 and 1983 were the result of the actions of 5 8 UK multinationals (who added 200,000 such jobs outside Britain across the same period).This echoes the data presented by Blu estone and Harrison for the US and the earlier research of Frank and Freeman (1978) which links substantial domestic manufacturing job losses to DI by US multinational..,.More broadly, Beenstock (1984) and UNIDO (1983) have, among others, attributed a significan t portion of the general North to South reallocation of manufacturing production and trade to the direct investment activity of core MNEs.
While the above treatment of Bluestone and Harrison's argument represents a bare bones statem ent of the popular version of the DIDT, it is an accurate one.Th e preceding discussion raises a number of questions regarding such argum ent..,.For one, the image of "globalization" that Bluestone and Harrison have in mind is first and foremost one of North to South capital flight in which direct inves tment decisions are guided in large st part by simple labor cost differ entials .A.., I have already indicat ed, this is an inaccurat e picture of the contemporary pattern of DI. 2 Most DI flows between nations with roughly comparable labor market conditions in which labor cost differential.., arc relatively minor (e.g. the US and Germany).While the spatial reorganization of manufacturing along a north-south axis is a real phenomenon -witness, famously, the malquiladoras along the U.S. border with Mexico -this simply docs not represent the general pattern of DI in the last two decades or so.Moreover, there may be other avenues through which DI might prompt dcindustrialization (discussed below).However, the North to South capital flight image is so strong in popular statements of the DIDT that these more subtle mechanisms arc sometimes overlooked.

[Page 9] Journa I of World-Systems Research
Moving away from the issue of direct investment, the DIDT, a.., usually stated, also largely neglects to account for the fact, a.., Singh (1977) has noted, that "de industrialization" ( or, at lea..,t, the relative decline of industrial employment) had long been anticipated by social scientists.Much a.., the Industrial R evolution initiated a movement oflabor out of the primary sector and into the secondary sector, social scientists had, well before Daniel Bell (1973), predicted development.., which would yield the future "coming of post-industrial society."For instance, Colin Clark in his The Conditions of Economic Progress ( originally published in 1940) laid out an early version of the shift from manufacturing to services argument that roots this phenomenon in two processes (1960: 493-494, cmpha..,is in original): first, with economic development, "a" real income per head incrca..,cs, it is quite clear that the relative demand for agricultural product.., falls all the time, and that the relative demand for manufactur e first rises, and then falls in favor of services;" second, given higher relative productivity in th e industrial sector, "a stationary relative demand for manufactures would lead to a decreasing proportion of the labor force employed therein."And, a.., Clark goes on to add (1960: 494), "even when the relative demand for manufactur es is increasing, we still generally expect, in the long run, a dccrca..,ing proportion of the labor force to be emplo yed therein."Thus a.., productivity grows and a.., the industrial economics matur e, one should expect that in the "normal" course of economic development secondary sector employment will contract while the tertiary sector will expand in the face of risin g demand for scrviccs.l.Q

[Page 10] Journa I of World-Systems R esearch
For an empirical analysis of the direct invcstmcnt -dcindustrialization thesis, th ese criticisms raise two major issues .First, the mechanism through which DI is often held to produce dcindus trializa tion (i.e.North to South capital flight) is mor e problematic than it might appear at first glance.As the data on DI indicat e, the va..,t majority of all DI flows occur betwe en core societies.While this docs not invalidate the DIDT, it docs suggest that a rea-;oned interpretation of any observed effect of DI on de industrialization will require some ( careful) post hoc theorizing and analysis.Second, when testing for an effect of DI on dcindustrialization, one should also be attentive to the important long-run generalizations offered by earlier analysts.

A MODEL OF DEINDUSTRIALIZATION
A-; background against which to test the DIDT, I employ the framework developed by Rowthorn and Wells (1987).For Rowthorn and Wells dcindustrialization (a-; the relative decline of manufacturing employment) can occur in at lca.:;t three ways.First there is the "positive dcindustrialization" that wa-; noted by earlier analysts such a-; Clark (1960) and which represents much of conventional economic thinking on deindustrialization (sec, for instance, Singh, 1989).Here dcindustrialization is viewed as a structural feature of all economics during the course of economic development.With development, a-; per capita income incrca-;cs, the share of employment in agriculture falls and the share of employment in manufacturing rises until a high level of development is attained.However, beyond some threshold of per capita income the share of services in employment begins to expand at the expense of manufacturing.This will occur a-; a consequence of the typically higher rate of productivity growth in the manufacturing sector relative to the service sector and of the systematic changes in consumption patterns that occur over the course of development (specifically, differences in the income cla-;ticity of demand across sectors).Such dcindustrialization is "positive" because it is viewed, not a-; a pathological phenomenon, but a-; a symptom of economic success.And labor shed in the course of positive dcindustrialization is viewed a-; being more or less quickly absorbed by the growing service sector.
[Page ll] Journal of World-Systems Research 'Negative dcindustrialization" is the second form of dcindustrialization that is posited by Rowthorn and Wells (1987).Here deindustrialization is the result of a pathological phenomenon, a structural disequilibrium in the economy, which prevents a nation from reaching its growth potential or a full employment of its resources.It manifests itself in poor performance in the manufacturing sector and is accompanied by a slow-down in manufacturing output and productivity.This leads to poor performance for the economy generally and a decline in competitiveness (in a cumulative vicious circle).Th e labor shed by negative dcindustrialization is, given the general state of the econom y, not absorbed by the service sector.Thus where positive deindustrialization is associat ed with rising real incomes and full employment, negative dcindustrialization is associated with stagnating real incomes and rising unemployment.The appreciation of currencies ( as in the UK in the late 1970s and the US in the early 1980s), high labor costs, poor product quality, and the failure or inability of firms to respond to changing market conditions have all been identified as factors in the "decline of competitiveness II experienced by a number of core societies in the past two decades or so (Ferguson and Ferguson, 1994 ).

[Page 12] Journal of World-Systems Research
Finally, shifting focus from the domestic to the international economy, Rowthorn and Wells (1987) argue for the existence of "trade-related deindustrialization." Trade is seen as affecting manufacturing employment through macroeconomic channels and through its influence on specialization.First, in contrast to many discussions in the NIDL and GOP literatures of the effect of international trade on manufacturing employment in the core, Rowthorn and Wells stress that in a mature economy de industrialization may be associated with either strong or weak trade performance.Wh ere manufacturing trade balances are positive and large, and the strength of the manufacturing sector contributes to sustained economic growth in the economy at large, the manufacturing sector may begin to shed labor (via positive dcindustrialization) at a higher rate than it would in th e absence of trade.Where manufacturing trade positions are deteriorating, and investment in manufacturing falls as a result, the manufacturing sector may begin to shed labor (via negative deindustrialization) into a stagnating economy in which it is not absorbed by the service sector.Underlying these potential macroeconomic effects of trade on the relative size of the manufacturing labor force arc the structural effects.Quite simply, nations that run manufacturing trade surpluses will, all else being equal, devote more resources and labor to this sector than will nations that run deficits.Trade may thus lead to further specialization in manufacturing among successful nations, and accelerate the move away from specialization in manufacturing among unsuccessful nations.
These differing forms of deindustrialization can be understood to operate concurrently; that is, the deindustrialization experienced by any given nation need not be the exclusive result of any one "form" of dcindustrialization.Indeed, it is probabl y most likely that the dcindustrialization experienced in core societies has been in part the result of a mix of • posltlvc an negative 1actors -continue s ow growt m re per capita incomes coupled with weak manufacturing performance -that has varied across time and place.In testing for a link between DI and deindustrialization, I will thus simultaneously control for these alternative sources of deindustrialization and employ as a baselin e the simple model suggcstcdbyRowthorn and Wells (1987: 31): where PCTMAN is the percent labor force in manufacturing, LRGDP is the logarithm (base 10) of real gross domestic product per capita, LUNEMP is the logarithm (base 10) of the unem ploymen t rate, and NMX is net manufacturing exports as a percentag e of GDP.LRGDP is employed to capture positive dcindus trialization.A curvilinear, inverted U-shapcd relationship is expected, as the share of employment in manufacturin g should first rise and then, after a certain point, start to fall.This relationship will be approximated as a second-degree polynomial of gross domestic product per capita.LUNEMP is employed to capture negative dcindustrialization and a negative relationship with PCTMAN is expected.Finally, N!vIX is employed a-; an indicator of trade-related dcindustrialization.As both positive and negative dcindustriali zation arc already controlled for, a positive relationship is expected.N!vIX in thi s instance is viewe d a-; primarily tapping into the structural or specialization effects of trade.

The direct inyestment-deindustrialization thesis revisited
In the empirical analysis that follows, I examine the effects of the total outflow of DI on the employment share of manufacturing.The criticism voiced above regarding the tendency of proponents of the DIDT to overstate the magnitude of North to South capital flight in their characterization of DI suggests that, in this context , reasoning such as Harrison and Blucstonc's must necessarily be supplcmcn tcd with some more ~eneral account of the relationship between direct investment and dcindustrialization. -1 A range of alternatives have been developed (e.g.Hymer, 1979;Cowling, 1986 ;Tanaka, 1991;Ictto-Gillics, 1992).These suggest that, beyond the direct labor-displacing effects of foreign investment stressed in the NIDL and GOP literatures, direct investment may have a range of ( dynamically evolving) indirect effects on the relative size of the manufacturing labor force.

[Page 14] Journal of World-Systems Research
First, outflows of DI may over time move a nation's economy into what Rowthorn and Wells term, by analogy to the more familiar "debt trap" from development economics, the "wealth trap."By this they mean "the automatic process by which a country whi ch is intrinsically a capital exporte r may become a renticr nation." (1987: 353, emphasis in original).In short , what starts out on the national accounting ledger as an outflow (i.e. the direct investment) may effect ively turn into a real inflow as profits from abroad outrun outflows of foreign investment.Ictto-Gillies has built on this insight and argues that the end result of this process is that (1992: 185): [nation s J with a long tradition of outward foreign investment arc lik ely to experience overall net 'positive' effects on the balance of payments.Th ese may cause a rise in the exchan ge rate; in a situation in which the econom y cannot -or is not allowed to by monetary and fiscal policies -expand to meet the extra demand generated by the inflow of incomes, the overall long -term effect will be a weakening of the manufacturing sector with loss of jobs and 'negat ive' deindustrialization.

[Page 15] Journal of World-Systems Research
Thus for nations such as the UK and US -nations with a long experience with outward direct investment -DI may not only lead directly to the displacement of manufacturing labor, but lead also to their "living on their pa'lt" in such a fa'lhion a'l to prompt a deterioration of their manufacturing sectors.If profits from abroad grow at a fa'ltcr rate than the domestic economy -an entirely conceivable phenomenon -the investing economy may, given that such profits will eventually translate into incomes, begin to import more manufactured goods and experience a rise in the exchange rate of its currency leading to a deterioration of its manufacturing trade position and, ultimatel y, its manufacturing sector.A'l ha'l been argued in regards to other a'lpccts of Briti sh and American hegemony ( e.g.Chirot, 1986), there may be long-term "costs of dominance" in this realm a'l well.
In addition to the "wealth trap" posited by Rowthorn and Wcll'l, Ictto-Gillics (1992) argues that DI may also contribute to dcindustrialization by lowering the rate of domestic capital formation.MNEs typically enjoy higher rates of return on investment than do comparable domestic firms.Where the activities of such firms arc substantial, this should tend to incrca'lc the required marginal rate of return on domestic investment and influence investment decisions accordingly.This will place a nation in a disadvantaged position relative to nations which arc less dominated by the activities ofMNEs.All of this may contribute to a cumulative vicious circle resultin g in dcindustrialization.Furthermore, Ictto-Gilli cs adds that (1992: 188): Global scanning combined with electronic technology in communications and ca'lc of movements of funds across :fronti ers by [MNEs] may have contribut ed to raising the rate of return on purely financial investment.This will have raised the mar ginal rate of return required on real capital formation.Similarly, high rates of return on the services sector (particularly the financial services) may have raised the marginal rate of return on [ manufacturing sector investments].

Journa I of World-Systems Research
Direct investment, then, may in this vein contribute to dcindustrialization through a number of intimately intertwined mechanisms.DI may raise the required mar ginal rate of return on domestic investment, shift investment from manufacturing to services, and reorient investment away from real investments toward financial investments.
While these arguments regarding the "wealth trap " and the effects of DI on capital formation remain speculative, when combined with an attention to the direct labordisplacing effects of DI stressed in the NIDL and GOP literatures, they do provid e a rough :framework for the interpretation of any observe d effect of th e total outflow of DI on the employment share of manufacturing.All argue for a negative relationship and each highlights a distinct moment in a nation's history of direct investment.In the short term there is the hollowing effect posited by analysts such as Harrison and Bluestone.Over a longer term there arc the effects of DI on the rate and character of capital formation traced out by Ietto-Gillies.Finally, given a sufficient history of DI, there is the possibility of Rowthorn and W ell's "wealth trap."While it is not my aim in this paper to develop a synthetic theory of the relationship between DI and de industrialization, I would maintain that any observed effect of DI can reasonably be interpreted in light of these mechanisms.

DATA AND METHODS
Data on the dependent variable, Percent labor force in manufacturing (PCTMAN), arc drawn primarily from the OECD's Labour Force Statistics, 1967-1987and Labour Force Statistics, 1973-1993(OECD, 1989;1995).Cases not covered in this source (i.e.Netherlands, 1967Netherlands, -1974) ) arc coded with data drawn from the U.S. Bureau of Labor Statistics (BLS, 1993).

Journa I of World-Systems Research
Real gross domestic product per capita (LRGDP) is mca<;urcd in US dollars.Va lues for 1967 through 1985 come from Summers and Heston (1988).Indices of real gross domestic product per capita published by the World Bank arc used to extend the Summers and Heston data to 1990.
Data on the unemployment rate (LUNEMP) arc drawn from the OECD's Labour Force Statistics, 1967-1987and Labour Force Statistics, 1973-1993(OECD, 1989;1995).To the degree possible, I use data that arc standardi zed by the OECD methodol ogy.
Net manufactur ed exports a<; a percentage of GDP (NtvIX) is mca<;urcd a<; exports minus imports mca<;urc d in current US dollars.Data arc drawn from the World Bank's World Tables (various years).Curre nt GDP in US dollars is drawn from the OECD's National Accounts (various years).
Data on outflow and inflow of direct investment a<; a percentage of GDP (LDIO, LDH) arc drawn from the IMF's Balanc e of Payments Statistics Yearbook (IMF, various years) .Curr ent GDP in US dollars is drawn from the OECD's National Accounts (various years).A<; direct investment is the variable of greatest interest in the analysis, there arc a few features of thes e data that arc worth noting."Direct investment" is defined by the IMF (1977: 136) a<; "investment that is made to acquire a la<;ting interest in an enterprise in an economy other than that of the investor, the investor's purpose being to have an effect ive voice in the mana gement of the enterprise." The key clement of this definition -that which distinguishes direct investment from portfolio investment -is its requir ement of "management interest" or control.Control is operat ionali zed in term s of a certain level of ownership.So a foreign investment which resulted, for instance, in one per cent ownership of the voting stock of a domestic firm would typicall y be cla..:;sificd a..:; portfolio investment, while a foreign investment which resulted in forty per cent ownership would be classified a..:; direct investment.In practice, the ownership thre shold is set much lower than forty per cent.As the llv1F (1977: 138) notes , this is done in the "recognition of the fact that -especially for large corporations of th e type that arc likel y to engage in multinational operations -a small, organized group of stockholders may well have an influence in management that is much more than proportionate to its share in the equity capital."

[Page 18] Journal of World-Systems Research
Nations differ, however, in the ownership thresholds (beyond which portfolio investment becomes direct investment) that they apply to international flows of investment. 12This , of course, raises the issue of comparability -an issue, surprisingl y enough, which ha..:; tended to receive less attention than it deserves in prior research on foreign direct investm ent .On this subject, however, the IMF's advice is instructive .The llv1F suggests that, whi le the problem of differing definitions should not be ignored, "borderline" ca..:;cs of foreign investm ent -cases in which the minimum thresholds becom e important -constitute a relativel y small proportion of the total universe of direct investment since most direct investment enterprises arc either wholly or majority owncd. 11Nations also differ in their reporting of DI data in a numb er of mor e idiosyncratic ways.While, for instance, the benchmark llv1F and OECD definitions argue for the inclusi on of reinvested earnings in DI flow data, nations such as Bel gium and Franc e exclude them.H A..:; th e pooled timcscrics of cross-sections methodology that I employ (sec below) enables one to control for unspecified time-invariant country-specific factors, these features of nati onal data collection and reporting system..:; will be implicitl y controlled.

[Page 19] Journal of World-Systems Research
Finally, in all of the models that I will est imat e, I includ e two peri od indicators (1974-81 and 1982-90, with the 1967-1973 period a..:; the baseline) to captur e time-specific effects.These indi cators trace, respec tively, the period from the trough of the 1973 -74 global recession to the peak of late 1970s expans ion and the period from the trough of th e 1981 -82 global recession to the end date of 1990.

Pooled time-series of cross-sections methods
The data..:; ct contains 408 observatio ns ; 24 observat ions (196 7 -1990) on each of 17 nations.In analyzing this data..:;ct, I empl oy an estimation procedure that is designed specifically to address the heterogeneity bia-; -the confounding effect ofunmea-;ured time-invariant country-specific variables -that is likely to plague the more familiar ordinary lea-;t squares (OLS) procedure in the context of the pooled time-series of crosssections data-;ct that I employ (see Hsiao, 1986;Greene, 1990).Heterogeneity bia-; can seriously affect OLS coefficient estimates, making OLS an inappropriate estimation technique.The fixed effects (FEM) and random effects (REM) models are two commonly used estimation strategies designed to correct for unmea-;urcd time-in variant factors.

[Page 20] Journal of World-Systems Research
These techniques ba.:;ically differ from each other in the fa.:;hion in which they treat the intercept and the disturbance term.FEM, like OLS, a.:;sumes the cla-,sical disturbance term but differs in it-, treatment of the intercept.Where, under OLS, all countries arc constrained to the same intercept, under FEM, indicator variables are introduc ed for each country and act a-; country specific intercept.:;.By doing so one "simulat es" unmca-,ur cd time-invariant country-specific effects and thus resolves the problem of heterogene ity bia-;.The FEM takes the following general form: where the subscript i denot es the country and t the time point of observation.In this equation, a.a represents the general intercept, ai the country specific intercept , and t it is the cla-;sical disturbance term (with E,:it = 0 and Var,: it = crt2).REM differ s from OLS mainly in the fa-,hion in which it treats the disturbance term.The REM takes the following general form: Yi1= a + ~•xit + Ui + tit Thus rather than treat country specific interc epts a-; fixed effects to be estimat ed, a-; und er FEM, the REM treats them a-; a random component of the error term.Compared to OLS, the REM involves the estimation of an additional component of the error varianc e: ui (country specific).

[Page 21] Journal of World-Systems Researc h
It can be shown that FEM is equivalent to applying OLS to data tran sform ed by subtracting the country -specific means from the data, while the equivalent REM transforma tion invo lves subtract ing only a portion of the country-specific means (Rosenfeld and Nielsen, 1984;Hsiao 1986).For methodo logical reasons, I present the REM estimates of the regression models.While I est imated both REM and FEM models, Hausman 's (1978 ; sec also Green, 1990) chi-square test of REM versus FEM uniformly favored REM ..li Analysis of outliers and influential ca~cs wa~ performed using the various diagnostics available in the SYSTAT and SYGRAPH statistical programs (Belsky, Kuh, and Welsch, 1980;Bollen and Jackman, 1985;Wilkinson 1990aWilkinson , 1990b)).These revealed the presence of a number of outliers.Their exclusion, however, had no dramatic effect.Standard errors were lowered and the significance levels of the various coefficients (and R 2 s) were consequently raised, but the exclusion of outliers had no substantive impact.For this rca~on, I include all 408 observations.I estimated the REM models with the LIMDEP statistical program (Greene, 1992).

Table 4. Correlations and basic statistics for variables in the analysis (N=408)
Variable )PCTMAN 1.000 (2) LRGDP -0.311 1.000 (3)LUNEMP -0.604 0.218 1.000 (4)NMX 0.441 0.118 0.091 1.000 (5) LDIO -0.250 0.456 0.211 0.089 1.000 (6) LDII -0.208 0.009 0.132 -0.307 0.364 1.000 (7) 1974-1981 0.098 -0.002 -0.035 0.000 -0.164 -0.054 1.000 (8) 1982-1990 -0.520Net manufactured exports a~ percentage of GDP NMX LDIO LDII Outflow of direct investment a~ percentage of GDP (log ba~c 10) Inflow of direct investment a~ percentage of GDP (log ba~c 10) 1974-1981Period indicator, 1=1974-19811982-1990Period indicator, 1= 1982-1990 Table 5. Unstandardized coefficients for the regression of percent labor force in manufacturing on selected independent variables: random effect model estimates for 17 OECD nations, 1967OECD nations, -1990   Model 1 introduces LRGDP and its square to capture the hypothesized curvilin ear relationship between development and the employment share of manufacturing in mature industrial societies (positive deindustrialization).Both terms are highl y significant and correctly signed, indicating that manufacturing employment first rises and then turns to decline with development.The R2 of .397 is only moderatel y strong.As some of the fit in Model 1 is contributed by the significant negative effect of the two period indicators, this suggests that the deindustrialization experienced by these seventeen nations has been the result of more than simple positive deindustrialization.
The role of negative deindustrialization is assessed in Model 2. The above discussion suggests that, in addition to indicating a nation's stage in the business cycle, unemployment may also proxy for the sort of structural imbalance that is stress ed in more critical accounts of deindustrialization.Under this view, the deindustriali zation experienced by core societies over the past two decades should not be viewed as the result of a "natural" and "self-correcting" phenomenon, but as evidence of profound economic distress.The strong negative effect ofLUNEMP, net of period indicat ors which crudely trace the business cycle, lends support to this argument.

Journa I of World-Systems Research
Model 3 introduces Nl'vlX as an indicator of the nation' s manufactu rin g trade position.The highly significant positive relationship observed (and its stab ility across Models 5 and 6) suggests that specialization effects outweigh the "accelerat ing" effects of trade; that is, rather than suggesting that trade surpluses have contributed to positive deindustrialization, the results indicate that they have, to date, led to the employment of additional resources (labor) in manufacturing.In largest part, then, th e patterns of trad e specialization that these nations exhibited moving into the 1970s have been repli cated down to the present.There arc, however, three important exceptions to this rule.The early eighties saw the formerly large manufacturing trade surpluses of th e UK, US, and France turn negative.While there may be a variety of reason s why this occurred (Rowthorn and Wells, 1987;Wood, 1994), this indicates that the results are also partially consistent with the NIDL and GOP interpretations of the effects of trade in the contemporary period: in an environment of heightened internat ional competition, particul arly from scmipcripheral and peripheral nations, tra diti onally high-wag e manufacturing operations in core nations have become vulnerable.This vulnerability has expr essed itself in at least some nations in deteriorating trade balances, disinvestment, and, ultimat ely, declining employment in the manufacturing sector.
The DIDT is teste d in Model 4. A neg ative rel ationship between LDIO and the employment share of manufacturin g is observed .The highly significant nature of thi s relationship (and its relative stability across Models 5 and 6) is somewhat surpri sing.Given the criticisms voiced above of popular statements of the DIDT and the speculative nature of the alternative theories that I touched on, one mi ght expect to find only modest support for the DIDT in the context of an examination of the effect of total outflows of direct investment.While the R2 of .303indicates that the three variables already discussed each provide a better fit than LDIO, the results do suggest an important role for DI in the deindustrialization experienced in the 17 nations under study.As suggested above, this result may be consistent with a combination of factors: the direct labordisplacing effect of DI stressed in the NIDL and GOP literatures, the effect of comparatively high ( and growing) levels of DI on domestic capital formation, and the "wealth trap" which turns capital exporters into rcnticr nations.
[Page 24] Journal of World-Systems Research Model 5 collects all of the variables examined thus far in isolation.LRGDP and it.:;square remain highly significant and correctly signed, a.:; do LUNEMP, NMX, and LDIO .The outflow of direct investment thus remains an important determinant of the employment share of manufacturing net of positive, negative, and trade-related dcindustrialization.The period indicators also remain significant, while the size of their coefficients declines noticeably.This suggest.:; that while the full model docs a better job at capturing tim especific effects than either of the preceding four models, some unm ca.:; ur ed effects remain (a.:; one might expect given it.:;relative simplicity).The fit of the full mod el (R2 = .650)is impressive for a model of this type, indicating that the variables collected in it account for a good part of the phenomenon of deindustrialization in these 17 nations .
Finally, Model 6 introduces a mea.:;ure of direct investment inflow (LDH).If outflows of DI are negatively related to the relative size of the manufacturing labor forc e, might not inflows of DI, given that their composition should tend to be similar, have a positive effect on manufacturing employment?In some conventional economic treatments of the employment effects of DI, it is argued that any job loss due to DI outflows may b e mad e up by DI inflows (in addition to being offaet by domestic employment growth prompt ed by incrca.:;eddemand for the input.:; of oversea.:;subsidiari es ) ( e.g.Frank and Fr eeman, 1978;Dick.en, 1986).In others, it is simply maintained a priori that the n et impact of DI on employment is near zero ( e.g.Graham and Krugman, 1991 ).As regards manufacturing employment, howev er, the results do not support these conclusions.LDII is not significantly related to the relative size of the manufacturing labor force.And controlling for LDII does no t apprecia bly influence the coefficient of LDIO or its significanc e level.This surprising finding lends additional gravity to the role of direct investm ent outflows.
[Page 25] Journal of World-Systems Research

CONCLUSIONS
The results presented here arc based on a data-;ct that pools ob scrvations on 17 OECD nations over the 1967-1990 period.They show that the dcindustrialization experienced across this period is largely explained by a model that combin es cla-;sic generalizations of the process of economic development with a range of more immediat e factors identified by contemporary students of dcindustrialization.The findings support a numb er of conclusions.First, dcindustrialization in the contemporary period ha-; not been the result of a "natural" process of "positive de industrialization" alone.While the results suggest that it is of continuing importance and should not be ignored, they also indicate a role for the sort of "negative dcindustrialization" discussed in more critical and specific treatments of de industrialization in the contemporary period.Support is also found for arguments which stress the role of international trade in dcindustrialization.Success, a-; indicated by a manufacturing trade surplus, has tended to lead to the devotion of additional labor to manufacturing.And where nations have historically specialized in other sectors, or, a-; with the UK, US, and France, have faltered, international trad e ha-; accelerated the move away from manufacturing.
The main goal of this paper wa-; to a-;scss the role of direct investm ent outflows in dcindustrialization.While the NIDL and GOP literatures offer a variety of arguments for the existence of a link between globalization and core dcindustriali zation, such arguments have generally not inspired empirical research by sociologists.This is unfortun ate since the issues of globalization and dcindustrialization imping e directly on a numb er of core sociological concerns, includin g social stratification, the sociology of th e labor force, and political sociology.I have tried to address this omission by explorin g one clement of th e broader discussion surrounding globalization and its effects.

[Page 26]
Journa I of World-Systems Research I find surprisingl y strong support for arguments that link deindustrializ ation to the outflow of direct investment.As I suggested above, a rca-;oncd interpretation of this link will require additional theoretica l work.While North to South capital flight would seem a clear enough (if not uncontro versial) mechanism through which direct investment mi ght contribute to dcindustriali zation, most direct investm ent flows between core societies.This suggests that a general account of the effect s of direct investment must necessarily incorporat e additional, indirect mechan isms through which direct investment mi ght operate.I have offered two such mechan isms.While I would stress again that arguments regarding the effect of direct investment on capital formation and the "wealth trap" remain speculative, the results of my analysis lend additional import to the pursuit of such hypothe ses.I am currently exploring these issues, and it is my hop e that this research might encourage other sociologists to continue to rigorously engage the variety of profound issues that surround globalization.Sociological discussions of globalization have too often taken on a very general and polarized character -alternating between a stance which suggests that globalization means that in some sense "everythin g ha-; changed" and one in which plus r.;a change, plus c 'est la meme chose.While it may indeed be the case that grander claims for the significance of globalization arc overblown, this docs not mean that the most recent round of globalization ha<; not been intimatel y involved in many of the most important social changes that the past two to three decades have witnessed.

Endnotes
.1.A direct investment is one which involves an ongoing, managerial interest on the part of the investor in the firm or operation invested in.The IMF (1977: 136) defines dir ect investment as "investment that is made to acquire a la<;ting interest in an enterprise operating in an economy other than that of the investor, the investor's purpos e being to have an effective voice in the management of the enterprise."As such, direct investm ent is distinct from portfolio investment : "long-term bonds and corporate equities other than those included in the categories for direct investment and reserves" (IMF, 1977:142).
The key distinction between direct and portfolio investment is one of control.Contro l is usually defined in terms of a certain level of ownership (sec OECD, 1987).This subject is taken up again below.
2. These seventeen nations arc Australia, Austria, Belgium, Canada, Denmark , Finland, France, Germany, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Sweden, Great Britain, and the United States.Data come from the IMF's Balanc e of Payments Statistics Yearbook (various years) .
.1."De industriali zat ion" ha<; of course been defined in a variety of ways.For the purposes of this paper, I define de industrializat ion a<; the relative decline of the manufacturin g labor force.Under this definition, absolute levels of output or employment in the manufacturing sector arc not a concern.This is not meant to imply, however, that attention to absolute levels is unimportant for our understand ing of the phenomenon (e.g.Singh, 1977;1989).4. The NIDL and GOP literature is simply too large to attempt to review it here.I a<;sumc that the participants in this session arc familiar with it.Sec Gordon (1988) for an appreciative, yet highly critical, review.In terms of "foundational" statement<;, the NIDL account is most strongly link ed with Frobel, Heinrichs, and Krcyc (1980), while the GOP school is probably best represented by Blucstonc and Harr ison (1982) and Piorc and Sable (1984).
2:.It is intere sting to note, for instance, that in a relative sense the liberalization of controls on international capital flows -one of the hallmarks of th e latest round of globalization (Harvey, 1989) -ha<; proceeded at a fa<;tcr pace in the developing wo rld than it ha.., in the developed world (UNCTC, 1991).The developing countries did, however, start the general liberalization process with a higher average level of control (Oxelhcim, 1993: 22).6.Which is not to suggest that importance of the MNE and its activities for the developing world ha.., declined.Indeed, a.., commercial bank.lending ha.., come to be seen, in the wake of the debt crisis of the early eighties, as a highly unpredictable source of development finance, direct investment ha.., found new favor of late in the eyes of reformers and advocates (Hcllcincr, 1991).
7. Direct investment is weighted by gross domestic fixed capital formation because it is the domestic indicator which is most similar to DI (Ictto-Gillics, 1992).GDP and GNP are also commonly employed a.., normalizers in empirical analyses of direct investment.There arc high correlations (r > 0.97) between the results of these different weighting schemes in my data..,ct.
~ There is, of course, an extremely large literature on deindustrialization.A good bit of the early discussion of dcindustrialization wa.., taken up with a debate over its definition and whether it wa.., in fact occurring.As I noted above, I adopt the common current definition under which deindustrialization is defined a.., the relative decline of employment in manufacturing.The argument linking DI and dcindustrialization appears to have found its contemporary form in the context of the broader debate in Britain over the "British Disea..,c" or "Englanditis" of the early 1970s (see, for example, the views collected in Coates and Hillard's (1986) volume on the economic decline of Britain).Many of the key terms and concepts that emerged in this debate were quickly taken up, largely wholesale, by American and Canadian researchers.Outside of the Anglo world, the direct investmcnt/dcindustrialization issue -if electronic searches of the rel evant literatures can be used a.., an indicator -is one that ha.., only relatively recently come to concern the broader public out..,idc oflabor movement circles in nations such a.., France, Germany , and Japan.
2.:.In Harrison and Blucstone's (1988) defense, they do note that dir ect investment is influ enced by an array of factors in addition to labor costs.Nonetheless, in discussing the hollowing of the manufacturing sector, they stress that labor cost differentials were the prime determinant of the upswin g in direct investment that the pa..,t two decades or so have witnessed.Even more finely drawn statements of the DIDT by authors who work more directl y with the specialized literature on DI sometimes fall into argum ents like this (e.g.Cowling and Sugden, 1987).Scholars unfamiliar with the large (primarily economic) theoretical and empirical literature on direct investment can find excellent reviews in Ictto-Gilli cs (1992), Dunning (1988), andUNCTC (1992).
lL One could, of course, more directly test Blue stone and Harrison's versio n of th e DIDT with data on north to south flows of manufacturing direct investment.Given my inter est in the globalization of production, however, data on total DI flows arc more appropriate.A-; noted above, North to South flows do not represent the general pattern of DI in the pa-;t few decades.
12. The OECD (1987) provides a detailed account of the collection techniques and definitions applied by various nations.As regards ownership thresholds, this source provides information on a number of the nations under consideration in the present study.In these nations, the following minimums were (circa 1985) set for an investment outflow or inflow to qualify a-; direct investment: Australia (25% I 25%), Austria (no acknowledged minimum I 5%), Belgium (no minimum for outflows or inflows), Canada (10% I 10%), Denmark (10% I 10%), Finland (20% I 20%), Franc e (20% I 20%), Germany (25% I 25%), Japan (10 % I no minimum), Netherlands (no minimum for outflows or inflows), New Zealand (25% I 25%), Great Britain (20% I 20%), Unit ed States (10% I 10%).These minimums have occa-,ionally changed over the time period under consideration.For instance, prior to 1980 Japan applied a 25% minimum to outflows (Julius, 1990).These changes appear to have been undertak en in an effort to achieve harmonization with the benchmark IMF/OECD Common Reporting System for Balance of Payments Statistics which suggests a minimum of 10% for outfl ows and inflows.
11.In total, the IMF offers the following on this issue (1977: 138): Much stress is often laid on the difficulty of defining dir ect investment precisely and of applying the concept in practice.It may be pointed out, however, that these problems, serious though they may seem, do not necessarily hav e a corresponding import ance for the validity and intcrcountry comparability of the statistics on dire ct investment.Most direct investment enterprises, in fact, either arc branches or arc subsidiari es that arc wholly owned by foreigners or in which a clear majority of the voting stock is held by a single foreign investor or group.The real borderline ca-;cs arc thus likely to form a rather small proportion of the universe.Moreover, since an enterpr ise is most apt to be inconsist ently cla.:; sificd when the share of the investor in question is quite small, the weight of the doubtful ca-;cs tends in principle to be further reduced by adherence to the [prescribed benchmark cla-;sification system].
.1.4...These deviations from the stand ard cla.:;sification system arc a result of the national data collection and reporting system-; in use in such countries.Two systems arc currentl y employ ed: surveys of invest ing companies and reports ofrclatcd ca-,h-flows through the banking system.Most OECD nations employ either a survey methodology or combin e survey techniques with ca-;h-flow data .Some nations (e.g.France), however, empl oy only a ca-,h-flow system and thus forms of equity other than ca-,h, such a-; reinvested earnings, arc not covered.Na tions also differ in how they approach taxes and in how they treat short-term loans, trade credit-,, interest pay ments and dividends (sec OECD, 1987).
li Substantively, the REM and FEM result-; were identical.The FEM estimates did produce much higher R 2 s (as all between-country variation is perfectly fitted with indicator variables). 1 L Employment share ofmanufacttrring.

Table 1 .
Estimated stock of accumulated direct investment by area of origin and

Table 3 .
Inflow and outflow of direct investment in 17 OECD nations as a pe rcentage of gross domestic fixed capital formation, 1967-1972 and 1985-1990