How have market and state shaped the long-term evolution of economic
performance and social protection during the nineteenth-century
globalization associated with British laissez-faire liberalism and the
post–World War II globalization associated with U.S.-embedded
liberalism?
The intensification of globalization has thrust this question to the
core of an emerging body of political economy research that compares
the two waves of globalization to draw lessons from the
nineteenth-century wave and apply those lessons to the current
wave.1
See James 2002; Aghion and Williamson 1998; Bordo, Eichengreen, and Irwin 1999; and O'Rourke 1999.
This research strongly reflects the
political tensions informing recent mass demonstrations against
perceived economic and social costs of unfettered globalization.
2 See Bhagwati 2002;
Broad 2002; Dollar and Kraay 2002; and Stiglitz 2002.
Critiques of globalization span the ideological spectrum: (1)
neoclassical economists who now believe that capitalism is subject to
structural political conflicts over the requirements of global
competitiveness and domestic social protection; see Stiglitz 2002; and Rodrik 1997;
(2) new Keynesians reasserting that capitalism engenders periodic
economic breakdowns; see Cornwall and Cornwall 2001; (3) third-way social philosophers and
activists advocating political control of unfettered global markets;
see Tonelson 2000; Gray 1998; and Giddens 1998;
(4) disillusioned liberals criticizing the destructiveness of
unregulated markets; see Soros 1998; and T.
Friedman 1999; (5) communitarians advocating
a world of self-sufficient closed communities; see Klein 2002; and Etzioni 1994;
and (6) populists supporting protectionism as they blame globalization
for long-term unemployment and increasing labor market inequalities;
see Mény and Surel 2002; and P.
Buchanan 1998.
It also reflects two
major themes voiced in the popular press by neoliberals and
neointerventionists: the long-term stability and viability of
post–World War II embedded liberalism in a fast globalizing
world; and fears of backlashes against globalization that are
reminiscent of the nineteenth-century wave.
3 Relying on neoclassical views of self-regulating economies,
neoliberals privilege the intensification of unfettered globalization:
they lament the inefficiency of state-steered embedded liberalism; fear
politically driven protectionist backlashes against globalization;
praise the efficiency of nineteenth-century laissez-faire in ensuring
superior economic growth; and favor the reconstitution of market
dominance as the ideal pathway to long-run growth, which best ensures
full employment and social protection.4
In contrast, neointerventionists fear the intensification of globalization:
they lament the social inequalities associated with unfettered markets;
denounce the social costs of nineteenth-century laissez-faire; also
fear protectionist backlashes; but privilege political control of
markets to ensure better economic performance—growth and
employment—as well as social protection.
5 Undoubtedly, these supposedly novel debates represent the resurgence
of the interwar “ideology debate”—pitting Keynes and
Polanyi versus Hayek and Von Mises—over the efficiency of market
and state in shaping the long-run evolution of economic performance and
social protection. The interwar debates concerned the long-term
stability and viability of nineteenth-century laissez-faire and its
replacement with embedded liberalism.6
Today, after half a century of
experiments with embedded liberalism, debates are about its long-term
stability and viability and its replacement with novel forms of
laissez-faire.
7 Nevertheless, systematic comparisons
of the two regimes that satisfactorily assess the empirical adequacy of
neoliberal and neointerventionist claims remain severely lacking.
The political economy research focuses on a variety of related
issues: for example, properties of globalization;8
Baldwin and Martin 1999.
Comparisons focus on globalization as an economic phenomenon: (1)
nineteenth-century globalization is as encompassing; see Vasquez 2000; (2) post–World War II globalization
is more encompassing; see Bordo, Eichengreen, and Irwin 1999; (3) trade and financial aggregates conceal
features that render comparisons misleading; see Krugman 1995.
economic and political determinants of
long-run growth;
9 See Cornwall and Cornwall
2001; and Maddison 1995.
wage and income convergence;
10 See Boyer and Drache 1996; and Williamson 1996.
inequality and growth;
11 See Aghion and Williamson 1998; and Williamson 1998.
and backlashes against
globalization.
12 Ultimately, however, such research
fails to capture how market and state have shaped the coevolution of
economic performance and social protection in each of the two regimes.
I argue that there is an unsettling disjunction between theoretical
claims about the long-run dynamics driving the two regimes and
empirical evidence derived from neoclassical theories of short-run
dynamics and statistical methodologies—such as means,
correlations, and ordinary least squares (OLS)-based regressions. At
best, such theories and methods can describe short-run stationary
fluctuations, not long-run dynamics.
13 For a discussion of such disjunction, especially in
economics, see Juselius 1999; Granger 1997; and Pesaran 1997.
The utter neglect of persistent,
nonstationary, dynamics of historical processes central to the
debates—for example, trade and financial flows, gross domestic
product (GDP), unemployment, government spending—triggers such
disjunction. It also results in spurious findings that inform erroneous
inferences about the long-run stability of regimes and unwarranted
policy lessons.
14 See Nelson and Plosser
1982; and Granger and Newbold 1974. I discuss stationary and nonstationary
dynamics below.
Nonstationarity, however, suggests alternative long-run stable
dynamics if the persistent processes were cointegrated, or coevolving,
around common stochastic trends driving the evolution of the two
regimes.15
See Johansen 1995; and Engle and Granger 1991.
This article adopts the theoretical
and methodological framework of cointegration to evaluate the rival
neoliberal and neointerventionist claims about how market and state
shape the long-term coevolution of economic performance and social
protection in the two regimes. Cointegration allows for the
investigation of three crucial dimensions of long-run dynamics: (1)
common stochastic trends, which capture system stability and rigidity;
(2) cointegrated equilibrium relations coevolving stably around the
common trends; and (3) adjustment mechanisms that maintain the
relations in long-run equilibrium.
To investigate these long-run dynamics, this article develops a core
multidimensional model rooted in the constructivist approach to
international political economy.16
The model links four historical
processes that are central to political economy debates on the
coevolution of economic performance and social protection in the two
regimes: (1) trade openness, (2) real GDP, (3) unemployment, and (4)
government spending.
17 With regard to
globalization, the analytic focus is on trade openness. Research in
progress that builds on this core model investigates the coevolution of
monetary regimes—the nineteenth-century gold standard and the
postwar sequences of fixed, flexible exchange rates—and trade
openness; see Eichengreen 1996; and Verdier
1998.
In the context of
constructivism, the research design relies on the ideal-typical
comparison of the normative structures of social purpose of the two
regimes.
18 Empirically, the comparison focuses
on the normative structures of the hegemonic powers that spearheaded
and influenced the evolution of the two regimes: Great Britain for
nineteenth-century laissez-faire during the 1865–1913 period; and
the United States for post–World War II embedded liberalism
during the 1955–2000 period.
Cointegration analyses of the long-run dynamics of the hegemons
challenge core neoliberal claims. First, unit root tests shatter claims
about the greater flexibility of nineteenth-century laissez-faire: all
four processes in both regimes exhibit nonstationary persistence, which
implies similar rigidity. Second, cointegration tests challenge
scenarios of regime instability foreboding the implosion of
globalization: the presence of one common trend driving the long-run
evolution of both regimes establishes the long-run stability of the two
regimes. The one common trend also challenges claims about the greater
rigidity of embedded liberalism: both regimes feature similar degrees
of rigidity. Third, equilibrium relations refute claims about the
economic inefficiency of the state in embedded liberalism and the
efficiency of laissez-faire markets: state interventions in postwar
United States are associated with better economic performance and
social protection. Fourth, adjustment dynamics question beliefs that
markets shape the dynamics of laissez-faire liberalism whereas the
state shapes the dynamics of embedded liberalism: state interventions
are necessary to maintain the equilibria both of nineteenth-century
British laissez-faire and of postwar–U.S.-embedded liberalism.
Ultimately, these findings question the wisdom of the neoliberal
political project to reconstitute market dominance: novel laissez-faire
worlds may yield inferior economic performance and social protection.
Rather, as neointerventionists claim, political control of markets
ensures superior economic performance together with better social
protection.
The first section below traces the resurgence of the “ideology
debate.” The second section presents the four-dimensional model
rooted in the constructivist theory of regimes. The third section
justifies the comparative research design. The fourth section
demonstrates the ubiquity of nonstationary dynamics. The fifth section
outlines the VECM cointegration model. The next three sections provide
tests of regime stability and rigidity (sixth); of equilibrium
relations (seventh); and of adjustment dynamics (eighth). The final
section discusses the implications of the research.
The ‘Resurgence of the Ideology
Debate’: Efficiency of Market and State
The ideology debate between neoliberals and neointerventionists over
the relative efficiency of market and state in shaping the coevolution
of economic performance and social protection in embedded liberalism
finds its roots in the highly unstable interwar period.19
Between World War I and the Great
Depression, the highly interconnected laissez-faire international
regime that had been dominant in the second half of the nineteenth
century imploded, and countries reverted to autarchic and
near-autarchic nationalistic forms of economic management.
20 Interventionists
such as Keynes and Polanyi, in contrast to economic liberals such as
Von Mises and Hayek, challenged the long-run viability of
nineteenth-century laissez-faire liberalism and laid the theoretical
and political grounds for postwar embedded liberalism.
21 See Keynes 1936,
1982; Polanyi 1957;
Von Mises 1977; and Hayek 1944, 1966. A rich body
of recent research has retraced the theoretical and policy relevance of
the interwar debates; see Hoover 2003;
Cochran and Glahe 1999; Allen 1998; Shearmur 1996;
Mendell and Salee 1991; and Hayek and
Caldwell 1995. However, the roots of the
debates extend well into nineteenth-century British liberalism; see
Coats 1971. Classical liberals—Smith,
Malthus, Ricardo, Bentham, and Mill—shared a faith in the
economic and welfare efficiency of unfettered markets and rejected that
the state could ameliorate socioeconomic conditions. Social
liberals—Green, Hobhouse, Hobson, and Bonsaquet—advocated
legislation that provided protection against market
dislocations.
The interwar debate hinged on diametrically opposite hypotheses that
are central to contemporary debates: the trade-off or complementarity
between market and state and between economic efficiency and social
protection. The trade-off hypothesis, dominant in nineteenth-century
laissez-faire, pitted efficient markets against inefficient state, and
economic efficiency against inefficient social protection. Best
captured by neoclassical formulations of Walrasian equilibria,
perfectly competitive and self-regulating markets converged
endogenously toward intertemporal equilibria of full-employment
growth.22
State interventions in the economy
and society were unnecessary and inefficient: unnecessary because
markets endogenously converged toward optimal economic and welfare
equilibria; inefficient because they hampered spontaneous convergence
toward equilibria.
Confronting the massive economic breakdowns, social unrest, and the
final implosion of the global economy, Keynes and Polanyi questioned
the trade-off hypothesis in favor of the complementarity of market and
state: state interventions were necessary to ensure economic stability
and social protection. Far from being self-regulating, laissez-faire
markets were inherently self-destructive and generated severe economic
and social instabilities that sparked backlashes against unfettered
globalization, finally leading to its implosion.23
This “apocalyptic” view of capitalism was
widely shared at the time; see Schumpeter 1928.
Keynes traced instabilities to the
“stickiness” of goods and labor markets and to the
volatility of global financial markets. Capital controls and aggregate
demand policies, via government spending, would contribute to economic
stability in an open world economy. Polanyi traced instabilities to the
social and political responses against the unprecedented historical
experiment, spearheaded by nineteenth-century Britain, of disembedding
capitalist markets from social relations. Re-embedding
markets—global and domestic—required governments to assume
broader economic and social responsibilities.
Hayek and Von Mises, instead, reasserted the validity of the
trade-off hypothesis. Beyond the negative effects of the singular event
of World War I, they traced the socioeconomic instabilities and the
ultimate implosion of the laissez-faire regime to inefficient positive
economic and social state interventions. Such interventions, which were
associated with the emergence of “collectivism,” undermined
the spontaneous functioning of laissez-faire markets. In particular, as
later research attempted to demonstrate empirically, inefficient
monetary policies were the principal cause of the Great Depression.24
Friedman and Schwartz 1963.
Hence, only the unfettering of
markets, by means of negative state interventions that bolstered market
dominance, would ensure long-run, full-employment growth and the wealth
necessary for social protection.
By the early 1960s, the “end of ideology debate” appeared
to irreversibly sanction the complementarity hypothesis. In the context
of the Keynesian-based Bretton Woods multilateral system, the
historical compromise informing embedded liberalism seemed to
successfully ensure global growth, full employment, and social
protection.25
See Marglin and Schor 1990; Ruggie 1982; and
Shonfield 1969.
The “end of
ideology” captured a new balance between market and state: the
political left had accepted the market as a central instrument of
growth; the political right, the state as an instrument of growth and
redistribution.
26 See Bell 1962; and Lipset 1960. In
economics, the “end of the business cycle” captured the
success; see Okun 1975.
Political
conflict was confined to valence issues, away from the interwar
fundamental clashes over the proper relations between market and
state.
Instead, by the early 1980s, following the breakdown of the Bretton
Woods system, stagflations, and the intensification of globalization,
the “resurgence of the ideological debate” under fierce
neoliberal critiques weakened the postwar consensus on the long-term
viability of embedded liberalism.27
A first wave of
attacks—especially from monetarist and new classical economists,
social conservatives, and the Virginia school of public
choice—challenged the complementarity hypothesis in the domestic
arena: the inherently inefficient positive economic and welfare
interventions, best captured by the historically unprecedented growth
of the public sector, hampered long-term economic growth.
28 Monetarist and new classical critiques
challenged the efficiency of economic stabilization; see J. Stein 1982; Sargent and Wallace 1975; and M. Friedman 1968. Demand-push neoconservative theorists argued
that malignant interest groups formation triggered demands that
overloaded the state budget; see Olson 1982;
and Crozier, Watanuki, and Huntington 1975.
Supply-pull public choice theorists pointed to the fiscal
irresponsibility of self-regarding politicians; see Mueller 2003; and J. Buchanan 1977.
Negative interventions, à la
Hayek, were necessary to achieve market deregulation and the withdrawal
of the state from the economy and society.
29 Campbell and Pedersen 2001.
Social inequality was deemed
necessary to provide incentives for economic growth.
30 Aghion, Caroli, and Garcia-Penalosa 1999.
A second wave of critiques
challenged the complementarity hypothesis in a fast globalizing world:
international markets, especially financial, inexorably weakened the
ability of the state to stabilize the economy and to provide social
protection;
31 the welfare state, in
particular, undermined international competitiveness.
32 Alesina and Perotti 1998.
Neo-Hayekean evolutionary theories
predicted the irreversibility of unfettered globalization as well as
the inevitability of convergence toward laissez-faire.
33 More recently, however, several countercritiques have reasserted the
complementarity between market and state. Endogenous growth theories
claim that selected types of government spending are economically
efficient.34
See Aghion and Howitt 1998; and Barro and Sala-i-Martin 1999.
In addition to investments on
research and development, spending on human capital is particularly
relevant. New political economy theories link social protection,
especially reduction of poverty and inequality, to productive
efficiency.
35 Aghion and Williamson 1998.
Claims that globalization inevitably
leads to the retrenchment of the welfare state appear to be dubious,
particularly in the context of postindustrialism.
36 The state
continues to shape national adaptations to globalization.
37 The strategies and instruments of
interventions used in embedded liberalism have changed but have not
eroded the normative structure of social purpose that defines
market-state-society relations.
38 Ruggie
aptly distinguishes between “norm transforming” and
“norm governed” change; Ruggie
1998.
These critiques have informed
neointerventionist defenses of the capacity of embedded liberalism to
ensure long-term economic stability and social protection.
39 Even self-professed neoclassical
economists now maintain that laissez-faire liberalism is a threat to
democracy and social stability: social protection is necessary to
forestall backlashes against unfettered globalization.
40 In historical context, the resurgence of the ideology debate
represents a rupture with the postwar consensus on embedded liberalism.
Neoliberals now explicitly admire the once discredited
nineteenth-century laissez-faire and question the long-term viability
of embedded liberalism established at Bretton Woods to replace the
then-discredited laissez-faire liberalism.41
See Lal 2000; Hayek and Caldwell
1995; and Friedman and Friedman 1980.
Simultaneously, they praise the
efficiency of nineteenth-century laissez-faire: it generated superior
growth through more encompassing globalization, greater flexibility of
global and domestic markets, and minimal state intervention. The policy
lesson they advocate—“to go back to the future” by
politically reconstituting global laissez-faire—rests on the
promise that unfettered markets, as in the nineteenth century, will
deliver sustained long-term growth.
42 The
metaphor “to go back to the future” is borrowed from Bordo,
Eichengreen, and Irwin 1999.
In contrast, mindful of Keynes's and Polanyi's critiques,
neointerventionists fear the reconstitution of unfettered
globalization: nineteenth-century laissez-faire might have created
unprecedented wealth but the social costs associated with disembedded
markets also triggered harsh backlashes against globalization.43
Concerns about renewed backlashes
against unfettered globalization inform an alternative policy
lesson—“to go forward to the future” by enhancing
political control over global and domestic markets to better ensure
long-term economic growth, full employment, and social protection.
The Model: Centrality of Long-Run Dynamics
At the core, the ideological debate between neoliberals and
neointerventionists is about the long-run dynamics—stability and
viability—of the two international regimes. A model ought to
capture the rival claims about long-run dynamics. The theoretical roots
of such a model are best traced to constructivist theories of
international regimes. A basic multidimensional model usefully captures
core mechanisms through which market and state shape economic
performance and social protection in the two regimes.
International Regimes: Structure and Dynamics
Tracing the model to constructivist regime theory is problematic: all
major schools of international relations—realism, neoliberal
institutionalism, and constructivism—disagree fundamentally on
the causal relevance of regimes.44
See
Simmons and Martin 2002; and Hasenclever,
Mayer, and Rittberger 1997.
Realism
attributes little importance to international regimes: they are
instruments for hegemons to enhance their power in the anarchic
international system. Neoliberal institutionalism attributes greater
instrumental significance to regimes: given anarchy, they can reduce
transaction costs in the interaction of states. Only in constructivism
are regimes ontologically independent from configurations of interstate
power and indeed can autonomously affect such power. By privileging a
social ontology that rests on holism, norm-based behavior, socially
constructed institutions, and historical contingency, constructivism
explicitly considers how the normative structure of social purpose
shapes the identity of regimes.
45 Such identity rests ultimately on
authority relations defining the relationship of market and state to
society and their contributions to economic performance and social
protection.
46 In the constructivist framework, international regimes emerge as
stable and viable social institutions around which the long-run
expectations of actors participating in the regime converge. Organizing
principles of legitimate order and meaning shape the historical
evolution—formation and transformation—of regimes. Market,
state, and society represent three fundamental macro-structures of
regimes. Authority relations define the legitimate orderings of market,
state, and society. The two regimes—nineteenth-century
laissez-faire and post–World War II embedded
liberalism—represent distinct and historically contingent
institutional configurations of authority relations.
Neoliberals and neointerventionists largely agree that diametrically
opposite principles and authority relations inform the normative
structures of the two regimes. Laissez-faire liberalism rests on market
dominance over the state and on disembedding markets from society. Its
central objective is economic growth through markets unfettered by
positive state interventions. Growth determines the dynamics of
unemployment. Growth also determines social protection via lower
unemployment and higher living standards. Negative interventions of the
laissez-faire state, à la Hayek, bolster market dominance. In
contrast, embedded liberalism rests on the complementarity of market
and state and on the social embedding of markets. Full-employment
growth and social protection are both central objectives. Positive
state interventions—stabilizing and welfare—in the
international and domestic arenas, as in the multilateral Bretton Woods
system, serve to achieve these twin objectives.
Neoliberals and neointerventionists, however, clash over how the
distinctive organizing principles of the two regimes ensure their
long-term stability and viability.47
According to neoliberals, the
institutional framework of laissez-faire liberalism generates stable
long-run dynamics. Unfettered—that is, politically
deregulated—markets endogenously adjust to shocks: they quickly
clear and converge toward a long-run stable equilibrium characterized
by full-employment growth. Instead, the institutions of embedded
liberalism inevitably generate unstable, sticky, dynamics: inefficient
state interventions in the economy and society hamper spontaneous
convergence of markets toward long-run equilibrium. Mindful of Keynes,
neointerventionists challenge the economic and welfare efficiency of
laissez-faire liberalism: unfettered markets fail to clear and to
converge toward the long-run equilibrium. Markets rather tend toward
disequilibrium characterized by persistent and socially costly
underemployment. Positive state interventions are necessary to ensure
convergence toward full-employment growth and social protection.
The Multidimensional Model: Testable Hypotheses
Such disagreements over long-run dynamics inform the debates on the
specific mechanisms by which market and state shape the long-run
coevolution of economic performance and social protection in the two
regimes. A simple multidimensional model, which focuses on the demand
side of the economy, captures the basic relations at the core of such
debates:
where g is government spending as a ratio of GDP; y
is the level of real GDP; u is the unemployment rate;
x, trade, describes the openness of the economy; and
t is a linear time trend capturing productivity shocks.48
These processes, albeit in single-equation
format, are central to the research on the post–World War II era;
see Swank 2002; Iversen 2001; Huber and Stephens 2001; and Garrett 1998.
They are also central to research on nineteenth-century laissez-faire;
see O'Rourke and Williamson 1999; and
Williamson 1998. However, the single-equation
modeling obscures the complexity of dynamics stemming from feedbacks
traversing the three equations.
The model thus links the
coevolution of economic growth, unemployment, and government spending
in the context of trade regimes. As in established research, trade is
analytically exogenous to the model.
49 The dimensionality, or complexity, of the model can be
increased by introducing other historical processes: for example, (1)
real wage, (2) investment, and (3) political institutions such as the
organizational and political power of labor and parliamentary
institutions. The increased dimensionality would not change the
theoretical argument and the empirical predictions of this core
model.
Equation (1) captures the evolution of the public sector. It links
the dynamics of government spending (g) to shocks in GDP and
unemployment—two crucial mechanisms of social protection. The
positive spending-GDP relation operates via Wagner, or wealth,
redistributive effects: spending increases as a proportion of national
income.50
In the historical context of
Bismarckian Germany, Wagner's original formulation of the growth
of the public sector was rooted in the requirements of statism and
militarism; see Wagner 1958; and Von Mises 1969.
The positive
spending-unemployment relation captures welfare interventions that protect
social strata exposed to labor market shocks, especially those linked to
trade.
51 For the postwar era, several theories
have linked public sector expansion to the unemployment effects of economic
openness: for example, for modernization, see Cameron 1978; for political-class conflict, see Garrett
1998; Esping-Andersen 1990; and Flora and Heidenheimer 1981; for the compromise of embedded liberalism,
see Ruggie 1982; for adaptive responses of
small but highly open economies, see Katzenstein 1985; for deindustrialization, see Iversen 2001; Rodrik 1997; and
Wood 1994.
Equation (2) captures economic growth—a crucial dimension of
economic performance. It links the evolution of GDP (y) to
shocks in trade, spending, and productivity. The positive GDP-trade
relation describes the openness-growth connection.52
The theoretical literature posits the positive
trade-growth link; see Obstfeld and Rogoff 1996. New growth theories for the postwar era focus
on trade-induced positive effects of R&D, increasing returns to
scale, and technological spillovers; see Aghion and Howitt 1998. Research on the nineteenth century also
posits the positive link; see O'Rourke 2000.
The GDP-spending relation gauges the
effects of economic and social interventions on economic growth. The
inequality sign ≠ indicates the disagreements between neoliberals
and neointerventionists over the long-term efficiency of positive state
interventions: the sign is positive, or growth inducing, for
neointerventionists; negative, or growth dampening, for
neoliberals.
53 The trend
t describes
technology-driven productivity shocks.
54 Aghion and Howitt 1998.
Productivity shocks encompass shocks in investments.
Equation (3) captures a second crucial dimension of economic
performance—unemployment. It links the evolution of unemployment
(u) to shocks in trade and GDP. Trade, as in Heckscher-Ohlin
models, induces socioeconomic dislocations best captured by
unemployment.55
Greenaway and Nelson
2001.
GDP describes the business
cycle trade-off between growth and unemployment.
56 In this system, government spending plays the crucial role in
determining the dynamics of the two regimes.57
In the context of constructivism, the focus on government
spending is meant to capture two main dimensions of state intervention
in the economy and society: (1) economic stabilization, and (2) social
security/protection; Ruggie 1982. The
objective of spending aimed at macroeconomic stabilization is to bring
the economy toward full employment growth—which is itself a
crucial component of social protection. The y and, via
recursive substitutions, u equations capture this first
dimension. The objective of social spending aims at minimizing the
domestic social costs of economic openness stemming from adjustments to
structural change in the international division of labor. The wealth
(y) and welfare (u) effects in the g
equation capture this second dimension. A narrower focus on social and
welfare spending would miss the complexity of these multiple dimensions
of social protection. Moreover, new growth theory demonstrates how
difficult it is to disaggregate government spending in its economic and
social components—for example, investment in physical and human
capital, welfare, and even military; Aghion and Howitt 1998.
Regarding the laissez-faire regime,
the model captures rival claims over the efficiency of unfettered
markets in the presence of minimal state intervention. Neoliberals and
neointerventionists share a basic agreement about the spending equation
(
g). Given the minimal state, small wealth and welfare effects
do not engender a considerable long-run growth of government spending.
Accordingly,
g cannot shape significantly the dynamics of
growth and unemployment. For neoliberals, self-regulating markets
unencumbered by state interventions ensure quick convergence toward
long-term full-employment equilibrium growth. Shocks are temporary, and
any long-run dynamics are traceable to exogenous deterministic
trends.
58 Instead, for neointerventionists, in
Keynesian fashion, laissez-faire markets are sticky: they fail to clear
quickly and yield suboptimal long-run equilibria characterized by
long-term underunemployment. Thus, in the
y equation, booms
and busts in trade are persistent and generate long-term fluctuations
in economic growth. In the
u equation, persistent booms and
busts in trade and GDP generate long-term fluctuations in
unemployment.
For embedded liberalism, the model captures the sharp divisions over
the efficiency of the large public sector. Here, neoliberals and
neointerventionists also share a basic agreement about the g
equation: significant wealth and welfare effects have contributed to
the historically unprecedented long-term growth of public spending.
Wealth effects point to strong mechanisms of redistribution:
vis-à-vis nineteenth-century laissez-faire, a greater proportion
of national income goes toward government spending. Given the
centrality of social protection, unemployment also represents a major
mechanism of growth for spending. For neoliberals, the trade-off
hypothesis predicts that the growth of the inefficient public sector
dampens long-run growth and yields persistent unemployment. For
neointerventionists, given the complementarity hypothesis, spending
contributes to long-term growth, lower unemployment, and social
protection.
Comparative Research Design: Inference from
Hegemonic Powers to Regimes
How does one evaluate the empirical adequacy of the rival claims
about the long-run dynamics of the two international regimes? A first
crucial methodological decision concerns the appropriate research
design and sample for valid inference. All theories of international
political economy—realism, neoliberal institutionalism,
constructivism—tend to rely on comparisons of the hegemonic
powers to investigate the core properties of the two regimes.59
See Keohane 1984a; A. Stein 1984; and
Ruggie 1982. Formally, the comparison
involves regimes as ideal types. More precisely, it involves the class
of ideal types based on particular historical phenomena, such as
“the Protestant Ethic,” and “modern
capitalism,” which appear only in specific historical periods and
in particular cultural areas; Weber 1949. The
polar regimes of nineteenth-century laissez-faire and post–World
War II embedded liberalism represent such types. The analytic task is
to identify the core features of these two polar types, which
essentially relate to the normative structure of social purpose. The
empirical task is to determine whether and how the normative structures
of the hegemons—nineteenth-century Britain and post–World
War II United States—approximate the properties of the two
types.
A fundamental, but by no means uncontroversial,
hypothesis supports the comparative design: hegemons influence the
evolution—creation and maintenance—of regimes.
60 Controversies concern: (1) differences in
motives ascribed to hegemons in realist, liberal, and constructivist
theories; (2) whether hegemons are necessary and/or sufficient for
regime formation and maintenance; and (3) whether hegemons privilege an
open international economy; Hasenclever, Mayer, and Rittberger 1997.
Thus, theoretical inference about
the dynamics of regimes is drawn from their respective hegemons.
Moreover, constructivist social theory, which in its critical version
relies on Gramscian views of hegemony, suggests the comparison of the
normative structures of the respective
hegemons—nineteenth-century Britain and post–World War II
United States. The logic of inference is that the relative dominance of
market and state in the hegemonic countries influences the evolution of
regimes, and that the investigation of the normative structures of
hegemons provides significant insights into the dynamics of
regimes.
61 In this constructivist
framework, the main objective of the research is to achieve a better
understanding of the normative structures and causal macro-social
mechanisms of the two political economy regimes, rather than the
intersubjective dimensions of human consciousness and action; see Adler
2002; Wendt 1999;
Gill 1993; and Ruggie 1982, 1998. Nevertheless,
the research has a strong reflexive—practical and
political—component: establishing the relative empirical
congruence of the rival neoliberal and neointerventionist claims has
significant implications for political discourse and political
action.
Three specific claims legitimize the constructivist comparative
design. The first justifies the comparison of the normative structures
of hegemons: the normative structures of the hegemons exemplify those
of regimes, thereby capturing their identity. Hegemons influence the
evolution of regimes through the fusion of two principles that project
their authority in the international system: (1) concentration of
wealth and power, and (2) normative structure of social purpose.
Resource concentration, central to realism, helps explain why a hegemon
creates and maintains an international order based on its self-interest
while enhancing global welfare. However, resource concentration does
not capture the distinctiveness of laissez-faire and embedded
liberalism. The normative structure does. The authority relations
defining state-market-society relations in each hegemonic provide the
legitimate social purpose in pursuit of which state power is employed
in both the domestic and international arenas. Hence, comparison of the
normative structures of hegemons best captures how market and state
shape economic performance and social protection of regimes.
The second claim justifies the sample selection: the normative
structures of nineteenth-century Britain and post–World War II
United States capture the distinctiveness of the regimes of
laissez-faire and embedded liberalism respectively. Resource
concentration remains crucial to explain British and U.S. hegemony.
Britain, under the impulses of the industrial revolution, colonization,
and global trade and finance, developed the largest and most open
economy.62
See Maddison 1995; Floud and McCloskey 1994; and Kindleberger 1975.
As the paramount military power in a
system of interstate hegemony, Britain also enforced its preferred
rules of free trade and the gold standard by maintaining freedom of the
seas.
63 The Pax Britannica ensured an
international order of relative peace and security.
64 As for the
United States, its large economy drives the world economy.
65 Obstfeld and Rogoff 1996.
The United States also fulfills
power criteria of hegemony: for example, acting as world police;
establishing the Bretton Woods regimes of trade and finance; and
providing investment and aid. The Pax Americana, too—debates on
the decline of hegemony notwithstanding—has ensured an
international order of relative peace and security.
66 Resource concentration alone, however, does not capture the
laissez-faire identity of nineteenth-century Britain: its fusion with
the normative structure that privileged self-regulating markets does.
The authority relations instituted in the international regime
reflected the domestic balance of British market-state relations that
determined the dominance of market over society. Laissez-faire Britain
privileged market forces both domestically and internationally: it
limited positive state interventions in the commodity, currency, and
labor markets; and it engaged in negative interventions to institute
and safeguard self-regulating markets.67
Three pieces of legislation institutionalized the
principles of laissez-faire: (1) the Poor Law Amendment Act of 1834
facilitated the flexibility of labor markets; (2) the Peel's Bank
Act of 1844 launched the gold standard monetary regime; and (3) the
Anti-Corn Law Bill of 1846 and the Cobden-Chevalier Treaty of 1860 set
in motion international trade; see Hobsbawm 1990; Taylor 1972; and
Polanyi 1957.
Likewise, for the
United States, the fusion of resource concentration and a normative
structure that privileges state steering of markets uniquely identifies
embedded liberalism. The authority relations instituted in the regime
reflect the domestic balance of U.S. state-society relations that rest
on the dominance of state over market and on the subordination of
market to society. From the New Deal to Bretton Woods and to the Great
Society, the state has taken greater responsibilities to buffer the
domestic economy and society from external shocks without sacrificing
the benefits of economic openness.
68 The
New Deal represented the U.S. reaction against the collapse of
laissez-faire and the Great Depression; the U.S.-driven Bretton Woods
regime sought to reconstruct the postwar international economic order
while minimizing cost of social disruption generated by international
integration; the Great Society programs added several layers of social
protection; Ruggie 1982.
The third claim establishes the validity of inference from the
comparison of hegemonic Britain and United States. Britain is the
extreme representative of laissez-faire. This ideal status had long
been noted in political economy: Britain exemplified the evolution of
nineteenth-century global capitalism.69
Other powers, such as
France and Germany, differed in socioeconomic and political structures
but were integrated in the global capitalism that Britain
spearheaded.
70 This ideal status continues to
inform the contemporary ideological debates on market and state: for
neoliberals, laissez-faire Britain exemplifies the world to which to
return; for Polanyi's “neointerventionist children,”
the world from which to escape.
71 In contrast, vis-à-vis other industrial democracies, the
United States is the weakest representative of embedded liberalism. It
never had a significant socialist movement or labor party.72
The social and economic reforms of
the New Deal lacked a strong ideological basis, were modest in scope,
and confronted strong opposition.
73 The Great
Society saw its limitations in the small public sector and residualistic
welfare state.
74 Labor markets have remained the most
flexible among the advanced industrial democracies.
75
The ideological attacks on the normative structure of embedded liberalism
since the early 1980s have been strongest in the United States.
76 Indeed, the
weakening consensus in the United States over the future of embedded
liberalism is about the new order that should replace the New Deal
state domestically and the Bretton Woods system internationally.
77 Yet, despite weak
representativeness, the central hypothesis—the normative
structure of hegemonic United States influences the identity of
embedded liberalism—holds.
78Historical Persistence: From
Nonstationarity to Cointegration
Testing the rival neoliberal and neointerventionist claims about the
long-run dynamics of the two regimes requires a second crucial
methodological decision: the choice of the appropriate statistical
framework. Graphical analyses and unit root tests of the four processes
in the model reveal that persistence, in the form of nonstationary
random walk dynamics, drives the evolution of both Britain and the
United States. Nonstationary persistence undermines established claims
about the long-run dynamics of the two regimes. It also suggests that
the framework of cointegration is necessary to investigate such
dynamics.
Historical Processes: I(0) Stationary and I(1)
Nonstationary Dynamics
Historical persistence hinges on the concepts of stationary and
nonstationary series.79
The properties of a series,
zt, are described usefully by an
autoregressive process of order one—AR(1):
where c is a constant; βt, a deterministic time
trend; ρ1, the autoregressive parameter describing
persistence; εt, a random disturbance term.
Table 1 summarizes the properties of
stationary and nonstationary series. zt is
I(0) stationary, or integrated of order zero, if
|ρ1| < 1. This property establishes the
time-independence of the (1) mean (μ = c/1 −
ρ1), which captures the intertemporal equilibrium; (2)
finite variance (σ2/(1 − ρ2))
describing the dispersion around the intertemporal equilibrium; and (3)
quickly decaying autocovariances (γs) and
autocorrelations (ρs), which reveal the memory
of historical processes. Accordingly, past shocks
εi do not cumulate historically but
converge quickly toward the intertemporal equilibrium. The trend
βt describes deterministic patterns of growth and decay,
but its detrending yields stationary, temporary fluctuations.
Properties of AR(1) processes
Instead, zt is a nonstationary random
walk integrated of order one, or I(1), if ρ1 = 1. This
property establishes the time-dependence of the variance
(tσ2), which tends to infinity, and of
covariances ((t − s)σ2) and
autoregressions ρ1 = 1, which do not decay over the long
run. With nonstationary dynamics, past shocks
εi cumulate permanently, or integrate,
in the memory of historical series. Therefore, such series cannot have
an intertemporal equilibrium to which they can converge. The absence of
an intertemporal equilibrium is obvious for a drift-driven random walk,
which exhibits a time-varying mean (μ = z0 +
c0 t): the drift c0
t captures cumulating patterns of growth/decay from the
historical origin z0.80
The intertemporal equilibrium is mathematically undefined:
given ρ1 = 1, μ = c/(1 −
ρ1) yields μ = c/0.
For a
drift-less random walk (
c = 0), the mean is μ =
z0, which is time invariant and corresponds to its
origin
z0. Thus only stochastic shocks
εi shape the evolution of
zt, which unfolds with sticky waves around
z0. However, given stochastic waves, the mean of
zt does not describe a meaningful
intertemporal equilibrium to which historical processes converge
systematically and quickly.
81 Such
equilibrium is also mathematically undefined: given ρ1 =
1, μ = c/(1 − ρ1) yields μ =
0/0.
Nonstationary Persistence: Britain and the United
States
Nonstationary persistence suggests that neoliberals commonly rely on
neoclassical equilibrium models that associate market flexibility with
stationary I(0) converging processes and sticky markets with
nonstationary I(1) nonconverging processes.82
See Aghion and Howitt 1998;
Juselius 1999; and Sargent 1987.
Thus, given highly flexible markets,
nineteenth-century laissez-faire would be a stationary world: shocks
are temporary, and historical processes quickly converge toward stable
intertemporal equilibria. By contrast, given sticky markets, postwar
embedded liberalism would be a nonstationary world: shocks are
persistent, and sticky historical processes fail to converge quickly
and systematically to equilibrium. By contrast, reliance on Keynesian
disequilibrium models informs neointerventionist beliefs that markets
in nineteenth-century laissez-faire are sticky and therefore generate
unstable—that is, persistent, nonconverging, and hence
nonstationary—dynamics.
Preliminary empirical evidence from Britain and the United States,
shown in Figure 1, challenges any claims of
converging stationary dynamics, particularly the neoliberal claims
about nineteenth-century laissez-faire.83
For Britain, data are for the 1865–1913 period. The
onset at 1865 eliminates the confounding effects of the Crimean War
while capturing the free trade drives of the 1860s. Trade and
government spending are from Mitchell 1998;
real GDP, from Backus and Kehoe 1992;
unemployment, from Mitchell 1988. For the
United States, data for 1955–99 are from Economagic
(www.economagic.com). The start at 1955 eliminates the effects of the
Korean War. Trade is exports (expgs) plus imports
(impgs) as a ratio of nominal GDP (gdpn). Government
spending is federal expenditures (afexpnd) net of federal
grants in aid to state and local governments (afgl) as a
percentage of GDP. Real GDP is gdpc.
Simply,
persistent random walk dynamics appear to be ubiquitous.
British trade exhibits long-memory stochastic waves of openness and
implosion.84
The implosion is associated
with the Great Depression of 1873–96; Floud and McCloskey 1994.
Instead, a drift, which may have
fueled recent concerns about the intensification of globalization,
drives U.S. trade. GDP exhibits clear drift-driven growth patterns in
both countries. U.S. unemployment exhibits substantial persistence,
which is associated with rigid labor markets. The higher frequency
fluctuations in British unemployment may suggest more flexible labor
markets, as neoliberals maintain. Yet long-memory swings typical of
random walks may characterize its evolution, as neointerventionists
contend. Lastly, upward sloping drifts (
c0
t > 0) also drive the growth of government spending in both
countries, albeit with different strength and from significantly
different initial historical conditions (
z0).
85 Research has recognized such growth; see
Lindert 2004; Tanzi and Schuknecht 2000; and Flora and Heidenheimer 1981. Public spending, as argued (see note 56
above), includes military spending for both Britain and the United
States; see Hobsbawm 1990; and Hooks 1991.
For Britain,
z0
in 1865 was about 6 percent of GDP; for the United States,
z0 in 1955 was about 16 percent. Drifts indicate
that for the sample period the absolute growth of British public
spending was roughly about 3 percent; in the United States, spending
grew by about 15 percent.
Augmented Dickey Fuller (ADF) unit root tests lend support to the
qualitative insights that all four processes are long-memory random
walks.86
ADF tests take the form
where zt is a nonstationary I(1) process
and Δzt is its stationary I(0) change;
δi capture higher-order correlations in lagged
Δzt that whiten
εt; and βt describes the time
trend. Given γ = (ρ − 1), ADF tests evaluate two
hypotheses: (1) the null of unit root (H0: γ =
0), by which ρ = 1; and (2) the alternative of stationarity
(H1: γ < 0), by which ρ < 1. A random
walk has a drift if c ≠ 0, and no drift if c = 0.
Table 2 shows the findings.87
Phillips-Perron tests yield similar
results; Phillips and Perron 1988. Tests are
performed in Eviews.
MacKinnon's critical values of the
t-statistic for γ,
t(γ), fail to reject the null γ = 0: for the sample period,
all four processes in both regimes appear to be I(1) random walks.
GDP is drift-driven in both countries (c > 0).88
Established research corroborates the
validity of these findings for the United States; Fatás 2000.
In postwar United States, trade is
drift-driven, which validates concerns about the
“acceleration” of globalization. In Britain, instead, trade
is a pure random walk, which suggests that by the 1860s, and for the
subsequent period up to World War I, trade had reached maximum
expansion. In both countries, unemployment appears to exhibit
drift-less random walk properties.
89 The
nonstationary persistence of unemployment, and more generally of
variables defined as shares, requires a distinction between conceptual
and statistical properties; Patterson 2000,
285–99. With regard to statistical properties, shocks are better
described as having permanent rather than transitory effects.
Conceptually, however, economic theory finds it difficult to consider
unemployment as I(1) nonstationary: its “random walk” is
bounded between 0 and 1 and therefore has finite variance. Indeed,
historically those boundaries are not reached or even approached.
Nonstationary persistence emerges as a local, or sample-specific,
property reflecting the effects of a big historical shock—such as
the oil shock of the 1970s—and not the global property of
unemployment unfolding in infinite time. In a long sample, say 200
years, there may be no significant trace of the local shock. Thus if
tests indicate that the null γ = 0 cannot be rejected over the
sample period, then the series is locally but not globally I(1).
Current research on unemployment reflects the tension between
theoretical and statistical properties. Neoclassical theory continues
to treat unemployment persistence in terms of protracted, yet
converging, deviations from natural equilibrium; Pissarides 2000. Nevertheless, unit root dynamics are
considered to be consistent with unemployment hysteresis, by which
unemployment has a permanent component and a path-dependent natural
rate; see Papell, Murray, and Ghiblawi 2000;
and Blanchard and Summers 1986. In the
context of cointegration, even if the roots were slightly lower than
unity, the unit root model is a sound approximation because it provides
a more reliable framework for statistical inference in a multivariate
setting; Juselius 1999.
Government
spending is drift driven in both countries. Both the bigger size of
government spending at the onset (
z0) and the
stronger drift-driven growth (
c = 0.3) for the United States
seemingly provide the rationale to neoliberal claims of the economic
inefficiency of the public sector. Nevertheless, despite the
significant institutional differences in the normative structures of
social purpose, both regimes share one common property: nonstationary
dynamics drive their long-run evolution. Shocks cumulate permanently to
generate persistent patterns of historical inheritance.
Implications of Nonstationarity: Disjunctions Between
Theory and Evidence
Nonstationarity in Britain challenges neoliberal claims about the
stability of nineteenth-century laissez-faire that rest on neoclassical
models of converging stationary dynamics. Nonstationarity lends support
to neointerventionist claims, inherited from Keynes, about the
stickiness of laissez-faire markets. Nonstationarity in the United
States, especially of unemployment and government spending, would
support neoliberal claims about the instability of embedded liberalism.
However, the persistence, even hysteresis, of unemployment in both
countries suggests that labor markets are sticky in both regimes.
Ultimately, these claims of regime instability based on the
nonstationarity of individual historical processes are mistaken. Random
walks may exhibit stable long-run equilibria if, as a system, they are
cointegrated—that is, coevolving historically around common
stochastic trends.90
Nonstationarity also challenges the use of standard univariate
statistics—most crucially algebraic means—to compare the
levels, or magnitudes, of historical processes in the two regimes. For
instance, it challenges neoliberal claims about the superiority of
nineteenth-century laissez-faire such as the claim that higher levels
of trade and smaller public sectors are associated with better economic
performance. Means properly depict intertemporal equilibria of
stationary processes; means of random walks are nonsensical.91
Note that the mean of stationary series
μ = c/(1 − ρ1) varies with the
values of ρ1. This property cautions against the
uncritical use of algebraic means even for stationary
processes.
Comparing means of nonstationary trade, GDP,
unemployment, and government spending in the two regimes is
mistaken.
92 The use of stationary growth
rates is problematic because stochastic detrending obliterates
long-term persistence, which is the crucial phenomenon under
investigation.
Appropriate comparisons would focus on the
historical origins
z0 and the drifts
(
c0 t) of series.
Lastly, nonstationarity invalidates claims of causal relations based
on fashionable single-equations OLS regressions.93
Such regressions at best would capture stationary relations. Given
nonstationarity, they yield spurious findings and mistaken causal inference.
While explicit comparisons of how market and state shape economic performance
and social protection in the two regimes are lacking, OLS techniques
dominate within-regimes research.
94 Dominance is clear in research on the postwar era; see
Swank 2002; Huber and Stephens 2001; and Garrett 1998.
It is also clear in research on the nineteenth century; see
O'Rourke and Williamson 1999; Verdier
1998; and Williamson 1996.
Indirect comparisons of such
findings to establish the superior empirical congruence of the
trade-off and complementarity hypotheses would lead to misleading
inference about the long-run dynamics driving the two regimes. A
cointegration model in VECM form more properly captures the complexity
of persistent historical dynamics.
95 Recent considerations of cointegration methods reveal the
concern with historical persistence; see Iversen 2001; and Freeman et al. 1998. Otherwise, two strategies, both
unsatisfactory, have been fashionable. One models the levels of
variables with the inclusion of an AR(1) term controlling for
persistence; see Swank 2002; and Garrett
1998. Controlling is noninformative about the
properties of persistence. The other models stationary
first-differences; see Alesina and Perotti 1998. Such strategy simply obliterates
persistence.
The Cointegration Model in VECM Form
The standard VECM model of long-run dynamics is96
Since the analytic focus is on long-run dynamics, the model
omits the matrix Γk
Δzt−k, which
describes short-run fluctuations in the changes of variables that meander
around the long-run relations.
where zt is a vector of at most I(1)
n variables in levels and
zt−1 is its lagged vector;
Δzt is the vector of
differenced, and hence I(0) stationary, variables in changes; and
Dt is a vector of deterministic variables,
such as time trends and step dummies, which capture policy
interventions and regime breaks. This model captures three dimensions
of long-term dynamics: (1) common trends, (2) equilibrium relations,
and (3) adjustments.
Common Trends
Cointegration assumes stationarity of a set n of I(1)
nonstationary variables in zt by linear
combination. The hypothesis of cointegration is formulated as a reduced
rank of the Π-matrix
α and β are p × r
matrices of full rank, where r, the rank of Π, is
equal to the number of independent cointegrating vectors, and
p is the number of common stochastic trends. The hypothesis
implies that Δzt is
stationary, zt is I(1) nonstationary, but
β′zt is stationary.97
See Johansen 1995; and Engle and Granger 1991.
The rank of Π suggests three main scenarios concerning
system (1) stability, (2) instability, and (3) rigidity. A system is
stable if Π has rank 0 < n − r
< n: the nonstationary processes are cointegrated and
coevolve around one or more stochastic trends that drive the stable
long-run evolution of the system. Instability occurs if Π
has rank n − r = n: there are as many
stochastic trends in the cointegration space as I(1) variables. Such
trends evolve independently following unrelated historical logics, and
thus fail to share a common history. Lastly, the higher the number of
trends, but clearly less than n, the greater system rigidity:
a system evolves historically in as many unrelated directions as there
are trends.
Equilibrium Relations
The matrix β′zt
represents up to (n − 1) stationary cointegrated
relations describing the stable coevolution of nonstationary variables
around common trends. The series share a history precisely because they
evolve around such trends. The equilibrium relations represent
long-term stationary associations between variables. They also
represent long-run steady states
(β′zt−1
= 0) toward which variables converge when shocks push them away from
equilibrium. Disequilibrium ensues when
(β′zt−1
≠ 0).
Adjustment Dynamics
The α matrix describes the direction and speed of
adjustments of variables in zt to
disequilibrium in the steady state of
zt−1. Relations are in equilibrium
at
β′zt−1 =
0, and in disequilibrium when
β′zt−1
≠ 0. In the presence of disequilibrium, the variables in
the system adjust to restore equilibrium in the next period such that
β′zt = 0.
Larger αij indicate faster
adjustments. If αij = 0, the
variables are weakly exogenous and do not adjust.
Deterministic Components
The matrix ΨDt describes
the effects of deterministic components, such as trends and
level-shifts, on the equilibrium relations. Especially when level
shifts are observed, prior knowledge suggests that something special
has happened in a given historical period, and that it should be
modeled using dummy variables.
Stability and Rigidity of Regimes: Rank of
Π and Common Trends
Are the regimes of laissez-faire and embedded liberalism stable in
the long run? If they are stable, which of the two regimes exhibits
greater rigidity? The first question—about regime
stability—concerns whether the four nonstationary processes are
cointegrated, that is, stably coevolving around one or more common
trends. The second question—about regime rigidity—concerns
the number of trends driving the nonstationary yet stable evolution of
the regimes. Determining the rank of Π helps answer both
questions.
The reliance of political economy theories on neoclassical models of
stationary dynamics provides little guidance in determining the rank of
Π.98
For a critique of the
weak theoretical guidance, especially in economics, see Juselius 1999; Granger 1997; and
Pesaran 1997.
proposes a strong version of neoliberal and
neointerventionist hypotheses about the stability of regimes for the
four-dimensional (
n = 4) model. Neoliberals claim regime
stability for the laissez-faire regime. Hence,
Π has rank
0 <
n −
r < 4: the random walks are
cointegrated and evolve around one or more stochastic trends. In
contrast, neointerventionists, mindful of Keynes's critiques,
claim system instability. Hence,
Π has rank
n
−
r = 4: with as many trends as variables, the random
walks are not cointegrated and evolve in four unrelated directions. The
opposite set of hypotheses would hold for embedded liberalism.
It is difficult, however, to fathom such an unstable social system in
which historical processes evolve independently of one another.
Therefore, a softer set of hypotheses assumes that both regimes are
stable but differ in rigidity. For neoliberals, the institutions of
embedded liberalism generate greater rigidities; hence, this regime
would exhibit a higher number of trends. For neointerventionists, in
contrast, the stickiness of laissez-faire markets generates greater
rigidity, and hence a higher number of trends.
Table 4 shows eingenvalue tests of the rank
of Π for Britain and the United States. Tests refute the
strong hypotheses of regime instability. They also refute the softer
hypotheses of differences in rigidity.99
Estimation is done in PcFiml 10.0; Doornik and
Hendry 2000. Tests reject the null of
serially correlated, non-normal, heteroskedastic vector residuals. The
λmax and trace rank tests are not appropriate
with dummy variables.
With one root at unity (p = 1), the eigenvalues of the
companion matrix indicate that Π has rank r = 3:
hence, one common trend (n − r = 1) spans the
entire history of each regime—at least for the sample period.
Implicitly, three equilibrium relations coevolve around such trends.
The common trends capture the nonstationary, yet stable, evolution of
the two regimes. The four nonstationary series are cointegrated and
share a history because they evolve around the common trend. Thus,
regardless of institutional differences in the relative dominance of
market and state, both regimes exhibit long-run stability and similar
rigidity. Hence, the neoliberal claims of instability and greater
rigidity of embedded liberalism find no support. Each regime appears to
represent a stable and viable historical experiment: one allegedly
driven by unfettered markets; the other, by the complementarity of
market and state.
Long-Term Equilibrium Relations
(βij): Distinctiveness of
Evolutionary Paths
Given regime stability, how do market and state shape the evolution
of the equilibrium relations? The rank of Π suggests a
four-dimensional model with three equilibrium relations that evolve
around one stochastic trend. The multidimensional model cast in terms
of cointegrated dynamics is100
This
cointegration model has two dimensions: (1) that of the model, which is
four-dimensional since it includes four endogenous variables
(n = 4); and (2) that of the cointegration space, which in
this model happens to have one common trend (p = 1) and three
cointegrating vectors (r = 3). Had there been two common
trends (p = 2), the four-dimensional model would have had two
cointegrating vectors (r = 2).
The β′zt matrix
describes the three relations theoretically normalized to capture the
evolution of GDP (yt−1),
unemployment (ut−1), and government
spending (gt−1).101
Cointegrating vectors are normalized by theoretically
setting variables to 1 to provide meaningful interpretations of
vectors; Boswijk 1996. Normalization
indicates that yt,
ut, and gt
are endogenous, whereas xt is
exogenous.
The time trend
tt in
the cointegration space describes the long-run growth of GDP.
102 The model contains stochastic and
deterministic trends. Stochastic trends represent the contributions to
the variables in the system of “unexpected” events with
permanent effects, whereas deterministic trends represent
“expected” events with permanent influence; see Aghion and
Howitt 1998; and Engle and Granger 1991. With regard to GDP, growth theory interprets
technological progress, and hence productivity shocks, as an expected,
regular phenomenon with permanent contributions. That is, following
“unexpected” innovations, technological change exhibits
some regularity, and the deterministic trend in GDP captures this
regularity. The stochastic trend, instead, captures the irregular, or
unexpected, technological changes, together with other unexpected
permanent changes. Thus, in the cointegrated GDP relations, the common
stochastic trend is shared among the variables in that relationship,
while the deterministic trend is unique to GDP, and therefore is
included in the relation explicitly.
The term
dt identifies the effects of deterministic
components on spending: (1) the Boer War for Britain;
103 See Hobsbawm 1990; and Floud and McCloskey 1994. The deterministic component is described by a
step dummy that takes the value of 1 for the 1900–1903 war period
and of 0 before and after the war years.
and (2) state
retrenchment and weakening of embedded liberalism since the early 1980s
for the United States.
104 For the
retrenchment and weakening in the United States, see Ruggie 1997; and Pierson 1994.
The model tests for such events by including a step dummy that takes
the value of 1 for the post−1982 period and 0 for the period
before.
The
α matrix captures markets adjustments
via GDP (α
2j) and unemployment
(α
3j), and state adjustments via government
spending (α
4j). Trade, assumed to be weakly
exogenous, does not adjust to disequilibrium in domestic variables
(α
1j = 0).
Table 5 summarizes the neoliberal and
neointerventionist hypotheses of long-run dynamics for the two regimes,
and Table 6 shows the empirical findings for
Britain and the United States. The findings challenge neoliberal
trade-off hypothesis of inefficient state interventions in embedded
liberalism while they strongly support neointerventionist claims of the
complementarity between market and state.
Normalized equilibrium relations: Hypotheses
Long-run equilibrium relations: Findings
Spending Relations: β′g
The β′g rows in Table
5 indicate that neoliberals and neointerventionists share a
basic hypothesis: persistent bursts in GDP and unemployment are
associated with stronger long-term spending expansions in embedded
liberalism than in laissez-faire liberalism. Larger wealth
(β32.EL > β32.LF
> 0) and welfare (β33.EL >
β33.LF > 0) effects capture the stronger
associations.105
In the subscripts, LF =
laissez-faire; EL = embedded liberalism.
With regard to
deterministic components, for Britain
dt
describes the temporary increase in spending during the Boer War
(β
36.uk > 0). For the United States,
dt captures state retrenchment, which is
characterized by a permanent new lower equilibrium in spending after
1980 (β
36.us < 0).
Findings in Table 6 challenge claims of
stronger long-term wealth and welfare effects in embedded liberalism.
Unemployment exhibits comparable positive associations
(β32.uk = 0.36 ≈
β32.us = 0.30). Shocks in GDP are more strongly
associated with spending increases in Britain
(β32.uk0.50 > β32.us
= 0.20). Thus, as long claimed, far from being unique to embedded
liberalism, wealth and welfare effects were operative in British
laissez-faire.106
Flora and Heidenheimer
1981.
Finally, for Britain, the Boer
war temporarily increased spending by about 5 percent
(β
36.uk = 0.5). For the United States, the
“Reagan-Volcker legacy” exerted a moderate permanent
downward shift in the equilibrium level of spending after 1980
(β
36.us = −0.12). This shift is
consistent with the hypothesis of neoliberal retrenchment aimed at
disembedding markets and weakening social protection.
Growth Relations: β′y
The β′y rows show a sharp disagreement
over the efficiency of the state: by the trade-off hypothesis,
neoliberals claim that persistent spending expansions are negatively
associated with long-term growth (β14.LF <
0); by the complementarity hypothesis, the association is positive or
growth inducing for neointerventionists (β14.EL
> 0). Disagreement also involves the positive trade-GDP link: to
neoliberals, persistent bursts in trade are associated with stronger
economic expansions in laissez-faire liberalism
(β11.LF > β11.EL
> 0); to neointerventionists, the association is stronger in
embedded liberalism (β11.EL >
β11.LF > 0). With regard to productivity
shocks, neoliberals maintain that trend growth is higher in
laissez-faire liberalism (β15.LF >
β15.EL > 0); to neointerventionists, trend
growth is higher in embedded liberalism (β15.EL
> β15.LF > 0).
Findings challenge all three core neoliberal claims. The
first—economic inefficiency of state interventions—is
unwarranted. Spending in British laissez-faire, unsurprisingly, appears
to be unrelated to growth (β14.uk = 0).
Instead, in postwar United States, persistent spending increases are
associated with long-term economic expansions
(β14.us = 0.40). This positive relation
supports the complementarity hypothesis. The second—trade
globalization in nineteenth-century laissez-faire generated superior
growth—is also unwarranted: both Britain and the United States
exhibit small, and statistically weak, positive associations between
trade and GDP (β11.uk = 0.11 ≈
β11.us = 0.12). The third—higher trend
growth in nineteenth-century laissez-faire—is indefensible:
growth is more than 1 percent higher in the United States
(β15.us = 3.1 >
β15.uk = 1.8).
Unemployment Relations: β′
The β′u rows show that neoliberals and
neointerventionists share a basic agreement on the trade-unemployment
relations. In laissez-faire liberalism, persistent trade expansions are
associated with long-term declines in unemployment
(β21.uk < 0). In contrast, given hypotheses
of socioeconomic structural changes—whether due to processes of
modernization, or postindustrialism, or deindustrialization—in
embedded liberalism, persistent trade expansions are associated with
long-term increases in unemployment (β21.us
> 0). The negative GDP-unemployment relations suggest another
agreement concerning long-run business cycles: persistent economic
expansions are associated with long-term declines in unemployment
(β22 < 0). However, the greater flexibility of labor
markets in the laissez-faire regime yields stronger associations
(β22.uk < β22.us
< 0).
Findings lend partial support to these claims. The negative
trade-unemployment relation finds significant support in laissez-faire
Britain: labor markets appear to be synchronized with long-term trade
fluctuations, whereby persistent trade expansions are associated
strongly with long-term declines in unemployment
(β21.uk − 4.0). The positive
trade-unemployment relation also finds significant support in
U.S.-embedded liberalism: trade expansions are associated, also
strongly, with long-term unemployment increases
(β21.us = +5.4). In contrast, the hypotheses of
long-term domestic business cycles find support only in the United
States: persistent economic expansions are strongly associated with
long-term declines in unemployment (β22.us =
−5.1). Laissez-faire Britain seems not to exhibit such a link
(β22.uk = 0).107
Tests could not reject the null of no long-term
association.
This finding about Britain is consistent with the
form of imperial industry based on imports of raw materials and export
of finished goods, with growth depending more on foreign markets than
on domestic consumption.
108 See Floud
and McCloskey 1994; and Hobsbawm 1990.
The two
relations—trade-unemployment and trade-GDP—indicate the
high dependence of the British economy on foreign markets. Finally, the
“Reagan-Volcker legacy” since the early 1980s permanently
shifted the equilibrium level of unemployment upward by a sizeable 1.7
percent (β
26.us = +1.7).
109 Tests rejected the null of no shift in the unemployment
relation since the early 1980s.
This shift, together with the
downward shift in the spending relation, provides further evidence of
neoliberal retrenchment: the significant increase in unemployment is
accompanied by a contraction of government spending.
Distinctiveness of Adjustment Dynamics to
Disequilibrium: αij
Lastly, how do market and state adjust to maintain the three
cointegrated relations in long-run equilibrium? Neoliberals and
neointerventionists largely share the belief that market adjustments
dominate in laissez-faire liberalism whereas state adjustments dominate
in embedded liberalism. Table 7 summarizes
the hypotheses and Table 8 shows the
empirical findings. The findings lend no support to beliefs that market
adjustments dominate in laissez-faire liberalism and state adjustments
dominate in embedded liberalism. Crucially, the state plays a crucial
role in both regimes.110
Three main
types of dynamics are of interest: (1) αij <
|1| indicates either gradual or oscillatory convergence;
(2) αij = |1| suggests that 100
percent convergence is achieved very fast within one year; (3)
αij > |1| implies
overreaction, or overshooting, which is followed by stable convergence
since none of the roots lies outside the unit circle.
Adjustment dynamics: Hypotheses
Adjustment dynamics: Findings
State Adjustments (gt)
The gt row in Table
7 captures the absence of the state in maintaining the long-run
equilibria in laissez-faire liberalism
(α4j.LF = 0). Government spending is
weakly exogenous to the GDP, unemployment, and spending relations. Only
markets, via yt and
ut, adjust to disequilibrium in the
relations. In contrast, spending plays a crucial role in maintaining
equilibria in embedded liberalism. Congruent with Wagner's wealth
effects, the state absorbs disequilibrium growth in GDP
(β′yt−1
> 0) via spending expansions (α41.EL >
0). Congruent with welfare effects, disequilibrium growth in
unemployment
(β′ut−1
> 0) also triggers spending expansions
(α42.EL > 0). Lastly, disequilibrium growth
in spending
(β′gt−1
> 0) is offset by endogenous mechanisms that curb spending
(α43.EL < 0).
Findings in the last row of Table 8 refute
beliefs in the absence of the state in nineteenth-century
laissez-faire: government spending plays a crucial role in maintaining
the long-run equilibria of both nineteenth-century Britain and postwar
United States. First, supporting Wagner's wealth effects,
disequilibrium growth in GDP triggers spending increases in both
regimes. Indeed, state adjustments are significantly faster in Britain
(α41.uk = 1.0 >
α41.us > 0.4): government spending absorbs
100 percent of disequilibrium within one year; postwar United States
absorb about 40 percent.111
With
α41.us = 0.4, 40 percent of disequilibrium is
absorbed in the first year, 20 percent in the second year, and so on in
geometric progression. With α41.uk = 1.0, all
(100 percent) disequilibrium would be eliminated within one year. In
reality, the dynamics of other variables do not allow the stochastic
equilibrium to be actually restored.
Second, contradicting the
welfare effects hypothesis, excess unemployment triggers slow gradual
declines in spending: the state confronts increases in unemployment
with fiscal contractions, not expansions. Adjustments, while slow, are
faster in Britain (α
42.uk = −0.10 >
α
42.us = −0.02). Third, disequilibrium
growth in spending is followed by comparable moderate gradual declines
in spending (α
43.uk =
α
43.us = −0.3): both Britain and the
United States absorb 30 percent of disequilibrium after one year.
Clearly, the state was not absent from the evolution of
nineteenth-century laissez-faire. It contributed to maintain
market-driven equilibria in Britain and state-steered equilibria in the
United States.
Market Adjustments (yt, ut)112 As with equilibrium relations, trade is
exogenous to domestic variables—GDP, unemployment, spending
(α1j = 0).
β′yt−1
> 0 Disequilibrium growth in GDP is followed by
economic contractions (α21 < 0) and by concurrent
increasing unemployment (α31 > 0) to restore
equilibrium. Given the high flexibility of laissez-faire markets, GDP
contracts faster (α21.LF <
α21.EL < 0) and unemployment increases
faster (α31.LF >
α31.EL > 0). Findings support the hypotheses
for Britain. Excess GDP triggers very quick GDP contractions
(α21 = −1.0) and unemployment increases
(α31 = 8.4).113
For
α31.uk = 8.4, it is possible to interpret only
the sign. Interpretation of the size/speed would require the
renormalization of the β′y vector to
u, which is not possible because u has a zero
coefficient in β′y.
The United
States exhibits no significant adjustments in
yt and
ut.
Only government spending appears to restore equilibrium GDP in the
United States.
β′ut−1
> 0 Excess unemployment is offset by higher
economic growth (α22 > 0) and by declining
unemployment (α32 < 0). Given the flexibility of
laissez-faire markets, GDP expands faster
(α22.LF > α22.EL
> 0) and unemployment declines faster
(α32.LF < α32.EL
< 0). Findings support this stabilization pattern only for the
United States: GDP increases, albeit very slowly (α22 =
0.03), and unemployment declines slowly (α32 =
−0.20). In Britain, excess unemployment seemingly foreshadows
recessions that result in very slow GDP contractions
(α22 = −0.03).114
Tests fail to reject the hypotheses of exogeneity of
trade for the United States. Trade adjusts slowly
(α12.us = 0.05), which indicates that U.S.
trade is not exogenous to the large domestic economy; see Ruggie 1982; and Keohane 1984a.
The small adjustment coefficients
provide further evidence of the stickiness of unemployment in both
regimes: any deviations of unemployment from equilibrium persist for a
relatively long time.
β′gt−1
> 0 Neoliberals and neointerventionists largely
agree on the adjustment dynamics to the disequilibrium growth in
government spending in the laissez-faire regime: by the trade-off
hypothesis, neither GDP nor unemployment adjusts
(α23.LF = α33.us = 0).
The “no-effect” outcome would be congruent with the
equilibrium relations where spending does not enter the GDP and
unemployment relations. Thus, spending itself would bear the burden of
reestablishing its own equilibrium (α43.LF <
0). Indeed, findings show that both GDP and unemployment in Britain are
unresponsive to deviations in government spending, and that
contractions in spending itself reestablish equilibrium
(α43.uk = 0.3).
In contrast, disagreement is sharp with regard to public spending in
embedded liberalism. For neoliberals, the growth of inherently
inefficient spending results in economic slowdown
(α23.EL < 0) and increasing unemployment
(α33.EL > 0). For neointerventionists, the
growth of spending results in stabilization patterns of higher growth
(α23.EL > 0) and declining unemployment
(α33.EL < 0). Findings for the United States
provide partial support to claims of complementarity: growth in public
spending eliminates half of disequilibrium unemployment within one year
(α33.us = −0.5).
Conclusions
These results yield seven major conclusions.
- Dynamics of nineteenth-century laissez-faire. Reliance on
neoclassical models of short-run equilibrium leads neoliberals to claim
stationary dynamics: shocks have transitory effects and fluctuations
quickly converge toward stable intertemporal equilibria. The
nonstationary persistence of historical processes in Britain shatters
claims of stationary dynamics while lending implicit support to
Keynesian claims about the “stickiness” of laissez-faire
markets. In the univariate domain, no significant differences
distinguish the two regimes. If nonstationarity of single
variables—hysteretic unemployment, “unsustainable”
growth of government spending, acceleration of
globalization—justifies claims of regime instability, then both
regimes would be unstable.
- Regime stability. By relying on persistent, and allegedly
unstable, dynamics of single historical processes, neoliberals and
neointerventionists alike fear the disintegration of embedded
liberalism—as happened for nineteenth-century laissez-faire.
Cointegrated dynamics refute the simplistic derivation of regime
instability and disintegration from the nonstationarity of single
variables. The common stochastic trend driving the evolution of both
regimes captures their long-run stability—at least for the sample
periods.
- Regime rigidity. Neoliberals associate the institutions of
embedded liberalism with market rigidities and “stickiness”
of government spending, whereas neointerventionists point to the
“stickiness” of laissez-faire markets. The one trend
spanning the history of each regime challenges neoliberal claims of
greater rigidity of embedded liberalism as well as neointerventionist
claims of greater rigidity of laissez-faire liberalism. Despite
differences in the normative structure defining the relative dominance
of market and state, the two regimes exhibit comparable
rigidities.
- Efficiency of market and state. The equilibrium relations
challenge neoliberal beliefs in the ineluctable trade-off between
market and state, lending support instead to neointerventionist claims
of complementarity. Market dominance in laissez-faire Britain is
associated with lower growth and weaker social protection—as
captured by weaker public spending and persistent unemployment.
Complementarity of market and state in U.S.-embedded liberalism is
associated with superior growth and stronger social
protection—the persistence of unemployment notwithstanding.
Neoliberal beliefs in efficient markets and inefficient state appear to
be a myth.
- Trade openness and economic performance. For
nineteenth-century Britain, the association of trade with weak economic
expansions and significant declines in unemployment lends partial
support to neoliberal claims about the benefits of trade openness in
laissez-faire regimes. However, it also feeds neointerventionist
concerns that unemployment bears the burden of stabilization during
economic downturns. In postwar United States, the association of trade
with weak economic expansions and significant increases in long-term
unemployment provide fertile ground for contention. Neoliberals blame
persistent unemployment on labor market rigidities, and hence they call
for deregulation. Neointerventionists blame deregulation, lack of
global labor standards, and the asymmetric North-South trade relations,
and hence they call for a stronger embedding of markets and for global
labor standards.115 These rival
interpretations require additional investigations of the supply side of
the economy—specifically, of how the evolution of labor market
institutions affects the persistence of unemployment.116
Established political economy research
addressing these issues remains dubious because of its reliance on
stationary theories and methods that are inappropriate for
nonstationary processes.
- Adjustment dynamics. Such dynamics challenge the core
neoliberal beliefs that markets provide the key mechanisms to
reestablish equilibria in nineteenth-century laissez-faire whereas
state interventions do so in embedded liberalism. State adjustments are
ubiquitous, yet play distinctive roles. For nineteenth-century Britain,
they contribute to maintain market-driven equilibria, whereas for
postwar United States, they maintain the equilibria stemming from the
complementarity of market and state. The belief in the absence of the
state from the long-run evolution of nineteenth-century laissez-faire
appears to be a myth. The belief that the state in embedded liberalism
undermines convergence toward long-run stable equilibria also appears
to be false.
- Policy lesson. Equilibrium relations suggest that the
political project “to go back to the future”—by
politically reengineering novel forms of laissez-faire—would mean
a reversion to a highly globalized world, albeit one with lower
long-run growth, persistent unemployment, and weak social protection.
Moreover, the adjustment dynamics reveal that such a laissez-faire
world would still require state interventions to maintain market-driven
equilibria. Instead, the neointerventionist project “to go
forward to the future”—by enhancing political control of
domestic and global markets—seems to promise superior growth and
social protection. Persistent unemployment remains as puzzling for
embedded liberalism as for laissez-faire liberalism.
Beyond these conclusions, cointegration analyses of the two regimes
make significant contributions to several debates in political
economy.
Theoretical contributions principally regard historical dynamics.
Cointegrated dynamics challenge the reliance of political economy on
neoclassical models of short-run dynamics. The fundamental dynamics are
long term. The persistent dynamics of the trade-unemployment relation
in nineteenth-century Britain contradict neoliberal claims of temporary
unemployment in laissez-faire regimes. Keynes may have been correct
about the “stickiness” of markets in nineteenth-century
laissez-faire. In addition, persistent dynamics in the spending-GDP
relation for embedded liberalism challenge neoliberal claims of policy
ineffectiveness and inefficiency. Government spending has long-term
positive effects on the real economy—growth and employment.
Rather, cointegrated dynamics are congruent with endogenous growth
theory, which allows for the stable coevolution of persistent
historical processes.117
Moreover, such dynamics
lend validity to the emerging research in historical institutionalism
and historical evolution, which acknowledges historical persistence.
118 This article simply
demonstrates that the fundamental macro-historical dynamics driving the
evolution of global capitalism belong to the family of cointegrated
processes.
Methodological contributions concern the significance of the
cointegration framework for the investigation of long-term historical
dynamics. Analyses of the long-run evolution of the two regimes lend
strong support to recent critiques of the disjunction between
theoretical claims about persistent, nonstationary dynamics and
theoretical models and statistical methodologies that at best capture
temporary fluctuations.119
Comparisons of the
long-run evolution of the two regimes are best accomplished within the
cointegration framework. Reliance on means and variances in the
univariate domain and on correlations and single-equation OLS-based
regressions in the multivariate domain simply fails to capture
historical persistence and its complexity. Worse, it results in model
misspecification and spurious findings that ultimately inform mistaken
inference and unwarranted policy solutions. The multidimensional VECM
best captures the complexity of persistent historical dynamics stemming
from feedbacks—multiple trends, equilibrium relations, and
adjustment mechanisms.
Substantive contributions concern debates specific to embedded
liberalism. Cointegration analyses demonstrate the analytic usefulness
of constructivist social theory for investigating how the normative
structures of social purpose shape the relative dominance of market and
state in the two regimes as well as their relevance for the coevolution
of economic performance and social protection. Specifically,
equilibrium relations unfolding around the common trend reveal that
market dominance in nineteenth-century Britain and complementarity of
market and state in postwar United States differentiate sharply between
the laissez-faire and embedded liberal regimes. Similarly, adjustment
dynamics show the uniqueness of equilibrium mechanisms in each regime.
Cointegration analyses also provide novel insights into the long-term
stability and viability of embedded liberalism. The unfolding of the
equilibrium relations around a common trend that spans the entire
post–World War II era does not suggest a rupture in the normative
structure.120
The breaks at the onset of the 1980s
indicate that shifts in the unemployment and spending relations have
contributed to disembed labor markets and to decommodify social life
but have not yet changed fundamentally the normative structure.
Policy contributions mainly concern fears of backlashes against the
intensification of unfettered globalization in embedded liberalism, and
even the possibility of its disintegration—as allegedly happened
in the first wave of globalization.121
See Williamson 1998; and Aghion
and Williamson 1998.
Such
fears—best captured by the query: “Has globalization gone
too far?”
122—seem to stem from the
interaction of two phenomena associated with the acceleration of trade
openness: weak growth and persistent unemployment; and reduction of
public spending for social protection in the name of international
competitiveness. The complementarity of market and state, which is
associated with superior economic growth and social protection in
post–World War II embedded liberalism, may best preempt feared
backlashes against globalization. Nevertheless, reconstituting
politically novel laissez-faire worlds, akin to nineteenth-century
liberalism, may represent a viable and stable, but seemingly inferior,
historical alternative.
