social policy, inequality and productivity · of social policy, such as education and ﬁscal...

Social Policy, Inequality and Productivity

Richard Harris text 11/27/02 2:21 PM Page 277

Richard Harris text 11/27/02 2:21 PM 78

279

INTRODUCTION

The equity versus efficiency argumenthas been the bread and butter of eco-nomic policy and social policy discus-

sions since the emergence of the modernwelfare state in the post-Second World Warperiod. In virtually all aspects of policy, thetwin goal of promoting economic progressand promoting social justice stands as a hall-mark of the modern industrial democracy. Bythe late 1960s the general view was that aconflict existed between the efficiency objec-tive and the equity objective, nicely summa-rized in Okun’s famous 1975 book, Equalityand Efficiency: The Big Tradeoff.1 In the 1990sa new debate has emerged covering similar,although conceptually different, ground.Productivity growth is widely regarded asthe major long-run determinant of per capi-ta income growth in industrial countries.Over the last two decades, economists havebeen preoccupied with understanding thesources of productivity growth, and slow pro-ductivity growth in Canada has been a majorpolicy concern for several years. Prior to themid-1980s, traditional economic analysis

focused on the static effects of economic pol-icy — the so-called size-of-the-pie effects.For example, when looking at the impact oftaxes on labour supply, the analysis was con-cerned with the one-time effect an increasein wage taxes could have on the labour sup-ply, rather than its effect on long-run eco-nomic growth. However, it is evident that,in the longer term, how fast the pie grows ismore important. The reason is simple: asmall change in long-term growth rates —on the order of 1.0 percent, or even less —has dramatically larger consequences than asimilar percentage change in GDP. Thisexplains the emphasis put, in both researchand policy, on understanding the factorsleading to higher, or lower, productivitygrowth, as opposed to other factors that donot have permanent consequences on growth.Social policy might well be one factor thathas an impact on growth. The expansion ofthe welfare state was heavily dependent onstrong economic growth in the 1950s and1960s. The fiscal repercussions of slow pro-ductivity growth, which had set in by themid-1970s and were evident in a debt anddeficit build-up by the mid-1980s, raised

Social Policy andProductivity Growth:

What Are the Linkages?

Richard G. Harris


concerns about the sustainability of heavysocial spending. For both of these reasons, thedynamics of social policy became inevitablylinked with the issue of economic growth.

That growth depends on productivity isnot a fact in serious dispute; but the long-runsources, or ultimate determinants, of produc-tivity growth are not completely understood.At the most general level, this is AdamSmith’s question: What are the sources of thewealth of nations? At a more restricted level,there is agreement on the proximate sourcesof productivity growth — new investment,human capital formation, new technology andproduct innovation. What drives these factorsin an economy has been accounted for largelyby economic determinants — that is, thoseimpinging directly on investment, innovation,education and trade, which appear to have adirect and medium-term impact on produc-tivity growth. However, recent research hasput forward the hypothesis that social factorsmay also be a major determinant of produc-tivity growth. Social factors would include thedistribution of income and wealth in an econ-omy, the range of social policy interventionsincluding health, education, labour marketregulation and a variety of income supportprograms. These social policies may be definedto include the tax-transfer system, whichfinances the social budget. The implicationsof this change of perspective are potentiallyquite powerful in making a case for social pol-icy. If it could be established that socialdeterminants are a quantitatively major factorin productivity growth, then the traditionalefficiency-equity tradeoff may not exist. Socialpolicies to promote equity could also bedefended on grounds that they simultaneous-ly increase economic growth. The tradeoff isreplaced by a virtuous circle in which equity-

enhancing policies also promote economicgrowth. This paper provides a critical evalua-tion of these arguments.

The paper present a survey of the evi-dence and debate on the social determinantsof productivity in the context of theCanadian productivity debate. It examinesboth the basic theoretical arguments and theevidence advanced by economists, and theirrelationship to what might be called modernsocial policy. Not all social policy is directlymotivated by equity considerations. In par-ticular, modern social policies in the area ofeducation and health focused on promotingthe growth of human capital represent onecategory where both the evidence and debateon the growth effects are qualitatively differ-ent from those in other areas of social policy.

It is instructive to consider the contextin which this often heated, and at times polit-ically loaded, debate surrounding the impactof social policy on economic growth has takenplace. Three trends have been driving thewider debate in industrial countries — all ofwhich are noticeable in Canada. First, the slowgrowth in Europe, particularly of employ-ment, had led many to put the blame on thewelfare state.2 "Eurosclerosis" became the termemployed to describe the slow growth and pooremployment record of a number of Europeancountries through the 1980s and early 1990s.A parallel debate in the Scandinavian countrieshas led many to the conclusion that theScandinavian welfare state had similar conse-quences. Assar Lindbeck’s critique is one ofthe most well-known (see Lindbeck 1975,1995). Part of the European record was theperception that generous social programs werea major factor responsible for the poor growthrecord. This debate was fuelled in part by thefamous OECD Jobs Study (1994) and an attack

280

Richard G. Harris


by all OECD governments on the growth ofdebt and deficits in the mid-1990s. It maywell be that the factors behind the slowemployment growth in Europe ultimatelyhave little to do with long-term productivitygrowth; but in the popular debate, theimpacts of the European welfare state on pro-ductivity, employment and fiscal policy tendto get lumped together. Canada is typicallyviewed as somewhere between the UnitedStates and Europe on the welfare state spec-trum, so that these arguments have likewiseplayed out here.

A second major element, of more recentorigin, is the debate on the “new economy”in the United States in contrast with the slowgrowth in Europe. The long and extraordi-nary economic expansion in the United Statesthroughout the 1990s was accompanied byhigh employment and strong productivitygrowth. While the sources of this growthremain a matter of discussion, the new econ-omy hypothesis claims that it is driven by theimpact of innovations in the information,communications and telecommunicationsfields, giving rise to an entirely new phase ofeconomic development — the so-called ThirdIndustrial Revolution. Prior to the recentsurge in growth, beginning in the mid- tolate-1970s but continuing into the 1980s,there was a significant rise in market incomeinequality in the United States and theUnited Kingdom. These trends have subse-quently shown up in most OECD countries,including Canada, but in Europe particular-ly it appeared that inequality was not increas-ing to the same degree. The acceleration ofgrowth in the United States during the 1990sled some to infer that inequality contributedto growth. The divergent US and Europeangrowth patterns in the 1990s have brought

the charge that the redistributive and labour-market policies responsible for Eurosclerosishave also prevented Europe from experienc-ing the growth benefits of the new economy.Economic growth and the preservation ofequality as seen through this debate appear tobe conflicting goals, reinforcing the old viewthat equity and growth are in opposition withone another.

Thirdly, an intellectual challenge to theexistence of an equity-efficiency tradeoff emergedat about the time that the Eurosclerosis debatebegan. In the mid-1980s economists began toseriously rethink the sources of economic growth,which led to both the New Growth Theory3 anda large empirical literature on the determinantsof growth and productivity. The developmentof new data sets for a large number of develop-ing and developed countries allowed researchersto pose new and interesting questions about thesources of growth. Much if not all of the intel-lectual impetus to discover links between socialfactors and growth is found in this literature oncross-country growth comparisons. In the early1990s a number of researchers identified arobust negative empirical correlation betweenmeasures of inequality and economic growth— lower inequality would be associated withhigher growth. Other researchers began to lookfor other policy determinants of growth, manyof which bear directly or indirectly on the issueof social policy, such as education and fiscal pol-icy. Lastly, a voluminous literature has emergedon the rising wage inequality in advancedindustrial countries over the last two decades.While not directly about productivity andsocial policy, the wage inequality issue figuresprominently in the productivity-social policydebate, for a simple reason. Much of this litera-ture adopts the opposite perspective — what isdriving inequality is economic growth, which

281

Social Policy and Productivity Growth:What Are the Linkages?


in turn is driven by technological change. Fromthis perspective, understanding the conse-quences of any policy intervention on inequal-ity and growth requires an understanding of thecomplex interaction between technologicalchange and productivity growth, and its impli-cations for wages and employment.

My purpose in this paper is to try tomake sense of these often seemingly contra-dictory pieces of theory and evidence linkingsocial policy to economic growth. Essentiallythe paper looks at two areas of research: thegrowth and inequality debate, and the smallbut growing literature on the policy deter-minants of economic growth. To provide thecontext for this discussion, the paper alsoincludes some background material on eco-nomic growth, productivity and social policyin OECD countries.4

By way of a caveat, the paper is focusedspecifically on issues that are pertinent toCanada, or at least to countries like Canada —those with a democratic, high-income, small,open OECD economy. Nothing in what fol-lows is meant to prescribe what developmentstrategies are, or are not, appropriate for thedeveloping world. The paper does not discussthe other main objectives of social policy thatare not directly related to growth. Lastly, thepaper does not discuss two areas of social pol-icy that do have growth effects but are notdirectly related to the productivity issue.These are: the effects of social security reformon savings — a very active debate driven bythe aging population issue; and the effects oflabourmarket regulation on employment,which have been extensively discussed sincethe release of the OECD Jobs Study.5

My main conclusion is in the form of anon-conclusion. This is one case where strongpolicy conclusions are well ahead of both the-

ory and evidence. Neither provides conclu-sive support for the proposition that either(a) policies directed at reducing inequality willincrease productivity growth, or (b) increasedsocial spending will raise productivity growth.Both advocates and opponents of such policieswill find little comfort in these conclusions:advocates, for the obvious reason that they areleft in the position of dealing with the chargethat equity and efficiency are often conflict-ing goals; opponents, because the evidence isoften sufficiently indecisive to leave ampleroom for a priori reasoned arguments to thecontrary. Lastly, it should be stressed thatmost of the research is relatively recent. It isentirely possible that the balance of evidencewill shift one way or the other as new stud-ies are published.

SOME BACKGROUND: PRODUCTIVITY GROWTH ANDSOCIAL POLICY

Productivity Growth: Conceptsand Framework6

Economic growth is measured as anincrease in real economic output per personat the national level and is generally regard-ed as reflecting four factors:

> capital accumulation> employment growth relative to popu-

lation growth> external market factors> productivity growth

Of these four factors, productivity growthhas generally been found the most importantfor industrial countries. However, all the otherfactors can play an important role at varioustimes. For example, a sudden increase in thefraction of the population that is employed

282

Richard G. Harris


would have substantive effects on growth for afew years. Moreover, a strict additive decompo-sition of these four factors could easily lead toincorrect inferences as to what is drivinggrowth. For example, an increase in productiv-ity growth caused by the availability of newtechnologies can lead to greater investment,which has an additional knock-on effect on thegrowth rate. Causality can also run the otherway — investment can carry spillover effectsthrough improved knowledge flows, leading tohigher productivity.

The productivity of an economic activityis defined by economists as the ratio of anindex of outputs to an index of inputs. It canbe defined at the level of an individual per-forming a certain task, a plant producing aparticular good, a firm carrying out a diverseset of economic activities, an industry, or anentire country. Productivity goes up whenyou can get more output with the sameinputs. The definition of productivity hingescritically upon how one measures the inputsand the outputs. In the economic literature,the starting point is a production functiondepicting a microeconomic relationship at apoint in time and mapping input to outputs.So we write, for example:

Y=AF(K,L)where Y is output, K and L are measures ofcapital and labour, F(K,L) is a time-invariantfunctional relationship between capital andlabour, and A is a time-varying parameter,referred to as an efficiency parameter or totalfactor productivity (TFP) parameter. The pro-ductivity level is defined as the output perunit of labour input — the average labourproductivity — either per worker or per hourworked, defined as Y/L. In this framework,productivity growth is the sum of two effects:the increase in the TFP parameter A, and the

increase in capital per worker K/L. Thisapproach is extremely well-known and is usedat both the individual micro-unit level and thelevel of the entire economy.7 In the latter case,output is measured as real GDP and L is eitherthe working population or the total numberof hours worked. At the macro level, A is alsoreferred to as the stock of knowledge, in line withthe recent emphasis on knowledge as the trulyultimate determinant of technological feasi-bility. In practice, growth in A is invariablydone by attributing to it what other factorscannot explain. In macroeconomics, this isoften referred to as the Solow residual. Formost industrial countries, growth in labourproductivity is accounted for by changes in A,while relatively little growth is accounted forby changes in capital per unit of labour.However, the range of estimates variesconsiderably.8

While this framework is conceptuallysimple and widely used because productivitygrowth can be identified by the residualmethod (i.e., the change in A calculated bysubtracting from the growth in Y a weightedaverage of the growth in K and L), it has longbeen recognized that this approach presentssome serious shortcomings. In particular, thereis no institutional context describing how eco-nomic incentives are determined, where newtechnology comes from, or what factors deter-mine investment. The major accounts of theIndustrial Revolution or of economic devel-opment offered by economic historians placegreat emphasis on these last factors.9

A more general diagram depicting thedeterminants of productivity growth is givenin Figure 1, which distinguishes between threeinterrelated categories — the economic deter-minants of productivity, the social determi-nants of productivity, and the policy and

283



institutional framework in which these factorsinteract. The arrows indicate the possibledirections of causality running between thethree sets of interrelated factors. It is conven-tional to distinguish between the direct effectand the indirect or feedback effect each ofthese variables has on each other. It is gener-ally agreed that investment, particularly inmachinery and equipment, has the most directmeasured impact on business sector produc-tivity. This shows up in both country micro-studies and cross-country studies. Many socialdeterminants could have an impact on pro-ductivity growth through their effect oninvestment. For example, greater political sta-bility contributes to investment growth by

reducing uncertainty; this greater investmentin turn raises productivity growth, whichleads to high economic growth. More gener-ally, government policies — economic andsocial — probably have some medium-termeffect on productivity growth via their impacton the economic determinants of productivi-ty growth, such as investment. However, botheconomic and social policy also impact on thesocial determinants of productivity growth.For example, education policy affects both theaverage level of human capital in the economyand the longer-run wage distribution betweenskilled and unskilled workers, which in turnaffects future investments in human capital.There are also linkages running between the

284

Richard G. Harris

FIGURE 1

A Conceptual Framework for the Analysis of Productivity

Economic Determinants of Productivity • Investment• New Technology and Innovation• Human Capital• Market Structure / Openness • Business Cycle Factors

Policy and Institutional Framework • Macroeconomic Policy• Microeconomic Policy• Social Policy• Financial Market Structure• Education System• Political Structure• Legal System

Social Determinants of Productivity• Wealth Inequality• Income Inequality• Social Cohesion• Trust and Association • Political Stability

A direct or medium- to short-term causal linkage

A foreign or long-term indirect linkage


economic and social determinants to the listof institutional and policy factors. Greaterincome inequality can influence political deci-sions on social policy, for example, whichwould have second-round effects on growthand inequality, and so on. For the purposes ofthis paper, these highly indirect factors will bementioned only occasionally, largely becausethere is not a lot of evidence in the literature.However, they certainly figure prominently inthe larger debate on the sources of differencesin national economic performance.10

One of the major problems affectingresearch on the deeper causal pathways run-ning from policy to growth is the time frameinvolved. Tax policy changes are likely toaffect investment next year; education policyreforms may not change the stock of humancapital in the economy for years to come. Thistime-horizon problem has forced researchersto use empirical data and methods that arecapable of identifying medium-term measur-able linkages between particular inputs andeconomic growth. Much of the cross-countryresearch, for example, tries to identify thelong-term effect of policy on growth by usingaverages of long-term growth rates over longperiods, often two or more decades, and sam-ples of countries with vastly different levelsof economic development. The difficulty withthis approach is that one is forced to assumethat the effect of a given variable on growthis the same for all countries, thus ignoringpotentially significant differences betweencountries in the way a given policy or socialfactor might impinge on growth.

As discussed in a companion paper tothis (Harris 1999), the bulk of the micro evi-dence on productivity is primarily about theso-called economic determinants. This reflectsboth data availability and the fact that eco-

nomic theories linking these factors to pro-ductivity growth have received a lot moreattention from economists than potentialsocial determinants. We now turn to adescription of where this evidence stands, anda review of recent trends in social policy.

Economic Determinants of ProductivityThe bulk of the productivity literature

is concerned with either (a) measuring pro-ductivity, or (b) attempting to assess thequantitative importance of a set of limitedeconomic determinants, largely at the micro-economic level but also at the macroeconom-ic level. The determinants that have receivedthe most attention are investment, human cap-ital, innovation and diffusion of technology,effects of international and domestic competi-tion, various forms of knowledge spillovers,and most recently geographic agglomerationof economic activity. The success of theseexplanations has varied. Beyond the first fourexplanations, the measured effects are high-ly variable and in many cases difficult todetect statistically.

The social policy-inequality-growthdebate has been partially motivated by andconducted almost entirely within a macro-economic framework focused on nationalcomparisons. This is not surprising since dif-ferences in social determinants are generallyregarded as having systemic economy-wideeffects that would tend to impact on all sec-tors of the economy. The search for empiri-cal regularities has therefore largely focusedon differences between economies, averagedover a number of years. Attributing differ-ences in productivity growth across timewithin a national economy to a single policyis fraught with difficulty. In particular, the

285



fact that so many economic variables tend totrend together makes it impossible to provethe importance of one particular factor rela-tive to any number of others. The mostprevalent form of evidence that has beenoffered in the modern debate, therefore, iseither reduced-form or structural growthequations in which the explanatory variableis the average growth of GDP per worker, orper hour, across a number of countries.Researchers in this area are well aware of thepossible complex causal relations linkingthese variables at the aggregate level. Successmay thus be judged by the standard scientif-ic criteria of demonstrating that a few vari-ables explain the data fairly well, or thatparticular variables show up repeatedly asquantitatively significant, despite variationsin the data or statistical methods used. So far,it has been difficult to show that the eco-nomic determinants do a fairly good job inexplaining the growth experience of countriesat all levels of economic development.

Using a full sample of countries at allstages of development and only a limited set ofeconomic variables leaves a lot to be explained.In discussing this issue, Hall and Jones (1999)point out that vast differences in income levelscannot be explained by savings behaviour oreven measured human capital levels:

Output per worker in the five countries withthe highest levels of output per worker in1988 was 31.7 times higher than output perworker in the five lowest countries (based ona geometric average). Relatively little of thisdifference was due to physical and humancapital: differences in capital intensity andhuman capital per worker contributed factorsof 1.8 and 2.2, respectively, to the differencein output per worker. Productivity, however,contributed a factor of 8.3 to this difference:

with no difference in productivity, output perworker in the five richest countries would havebeen only about four times larger than in thefive poorest countries. In this sense, differencesin physical capital and educational attain-ment explain only a modest amount of thedifference in output per worker across coun-tries. (92)International productivity differences (in

levels) are enormous and any coherent expla-nation will have to rely on institutional andsocial infrastructure factors. The relevance ofthis to the OECD countries — many have verysimilar levels of economic development andquite similar institutional structures — isquestionable. For these countries, similaritiesin institutions and developmental stages implythat the sources of growth are more likely tobe found in a common set of factors. Most eco-nomic theories simply assume the problemaway. Contemporary growth theory largelyassumes a well-functioning market systemwith efficient financial markets, and marketsthat clear (most of the time) for labour and cap-ital. Are these theories — now textbook mate-rial for most graduate students — capable ofdescribing the modern economic growth expe-rience of advanced countries? The answer is nota decisive yes or no, but, as we will see below,the support for these models in the case ofindustrial countries is fairly good. In general,however, the task they face is considerably lessdaunting than it is for models attempting toexplain what Hall and Jones describe, giventhat the maximum difference in income levelscan be expressed as factors of 2 to 3.

Growth theory and empirical workhave made some progress in the last decadetoward reducing the uncertainty surround-ing the determinants of growth in industri-al countries. Temple (1999), for example, is

286

Richard G. Harris


cautious but optimistic in his assessment ofthe literature. I would summarize the evi-dence on modern empirical growth modelsas involving three stages — the reduced-formliterature, and then the structural models ofgrowth without and with explicit transition-al dynamics.

First, in the cross-sectional reduced-form literature, there is a consensus that rel-atively few variables are statistically robustin a growth equation (see Levine and Renelt1992; Sala-i-Martin 1997). In a growth equa-tion, average labour-productivity growth isthe dependent variable with a set of poten-tial explanatory variables on the right-handside. The successful variables include:

> the initial income level at the begin-ning of the period

> investment-to-GDP ratios> schooling levels> population growth> indicators of openness in trade and/or

foreign direct investment (FDI)Temple (2000) surveys this literature

and notes that, given the lack of an explicittheoretical structure, a large number of vari-ables have been tried and the whole literaturesuffers heavily from data mining. That said,the growth regression literature has been veryinfluential, although more so with respect todeveloping country issues than advancedcountry issues. The early work also revealeda number of variables that, to some, were notgood explainers of growth. These includedfiscal policy, R&D measures, and variouspolitical and legal variables.

Second, an important structural modelof growth is the Mankiw-Romer-Weil (1992)augmented Solow model. This is the basicneoclassical growth model of Robert Solowwith exogenous savings in physical capital, to

which is added a third factor input — humancapital. This is all done within a constantreturns to scale aggregate production frame-work. The model is empirically implementedby imposing a steady-state restriction whichimplies that countries are on a steady-statelong-run growth path for the period exam-ined. Under this assumption, growth rates(the dependent variable) can be expressedwithout reference to the stocks of physical orhuman capital, but as functions of the savingsrate, a schooling variable, and an initial pro-ductivity level assumed to be randomly dis-tributed across countries. Attempts to makethis model fit OECD cross-sectional data havenot met with much success. This can beregarded as either a failure of the theory or areflection of the fact that the steady-staterestriction is too constraining.11

Third, the 1990s have brought a varietyof structural growth models that incorporatehuman capital and drop the assumption thatobserved growth is of the steady-state kind. Byincorporating dynamic transition effects toallow theoretical growth rates to vary overtime, the models have met with somewhatmore success. Barro (1991) was an early pioneerin this area, but numerous methodological,measurement and econometric improvementshave been made over the last decade. A goodtechnical survey of this literature is provided byDurlauf and Quah (1999), and it is covered inpart in the Barro and Sala-i-Martin (1995) text-book. More significantly, the most recent ver-sions of these models use panel data thatexploit both cross-sectional and time seriesvariation and are estimated using a variety ofwhat are referred to as dynamic panel methods.Initially, there was some debate about the wayin which the human capital variables shouldenter the model and some of the early results

287



on human capital were quite odd. However,this human capital paradox has recently beenlargely resolved. Many of these estimates sup-port the view of close to non-diminishingreturns to a broad measure of human and non-human capital. Non-diminishing returnsimply that increases in broad capital per work-er yield incremental output increases that donot diminish as more capital is added. Thiscomes very close to supporting what is knownas endogenous long-run growth. Endogenousgrowth, as developed by Romer (1990) andLucas (1988), occurs when a policy variable,such as the savings rate, can have a permanenteffect on the growth rate as opposed to the long-term level of income. Non-diminishingreturns to capital are a sufficient condition fora growth model to generate endogenousgrowth. A model exhibits exogenous growthwhen policy variables have only transitionaleffects on growth rates, although they canimpact on steady-state levels of income. TheMankiw-Romer-Weil model is an example ofan exogenous growth model.

Measurement and data issues have turnedout to be quite important in this literature.Changes in data on capital stocks, human cap-ital and specific economic policy variables havetended to have a substantial effect on estimat-ed parameter values (see Temple 1999).

Policy enters these models either as anadditional explanatory variable or as a struc-tural characteristic of the model. While inprinciple one can distinguish between endoge-nous and exogenous growth models, empiri-cally identifying the effect of a policy variableon the steady-state income level versus themedium-term growth rate has proven to bevery difficult with data sets covering 20 to 30years. This is simply because convergence inthese models is relatively slow, and when the

share of profit and returns to human capitalbecomes large (on the order of 2/3 or greaterfor most high-income countries), endogenousand exogenous growth models begin to behavequalitatively in a very similar fashion. A lot ofthe most recent literature works largely with-in an augmented Solow framework, in whichpolicy impacts on the transitional growth rate,although the effects can last for a couple ofdecades. Policy is often discussed in terms ofits impact on the rate of convergence. This refersto the fact that holding policy constant, thesetheories predict income levels that tend to con-verge to the steady-state income level. The rateof convergence is defined by reference to howlong the process takes. Typical estimates are inthe range of 15 to 30 years. When an economyis out of steady-state growth, which is usuallyassumed to be the case of interest, changes inpolicy impact on the rate of convergence aswell as on the long-run level of income. Otherthings being equal, a policy that raises long-run income and has a shorter period of conver-gence is preferable to one that has a longerperiod of convergence.12

A recent paper by Bassanini et al.(2001) provides a good example of the use ofthis type of econometric model for a cross-country analysis of growth in OECD countriesover the 1971-98 period with a specificemphasis on economic determinants. Thebasic growth model is a dynamic version ofthe augmented Solow model discussed inchapter 5 of Barro and Sala-i-Martin (1995)with human capital and R&D. Policy variablesinteract with accumulation variables and alsohave a potential impact on long-run steady-state levels of productivity. The model doesnot impose similar dynamics on all countries— rates of convergence are allowed to varyamong countries — but it does assume that in

288

Richard G. Harris


the long run all countries are governed by sim-ilar parameter values up to a constant level ofdifference between countries. The model doesquite well at tracking the data, and theauthors provide an illustrative decompositionof the factors that determine aggregate pro-ductivity growth. The set of variables thatexplain growth includes a group of baselinevariables (those derived from the basic theory)and a group of economic policy variables thatshift the growth path:

Baseline variables:>the initial productivity level>the share of investment in GDP>population growth>human capital

Policy variables:>trade intensity>R&D expenditures> inflation variability>government investment>government consumption In the estimation of the model, govern-

ment investment turned out to be insignifi-cant, while the R&D variable had to bedropped due to limited country coverage,although both were significant on a more lim-ited data set. Table 1 reports the decomposi-tion of the growth rate for each countryexpressed as a deviation from the OECD aver-age. Looking at the row for Canada, we seethat the country’s annual growth rate of

289


TABLE 1Economic Determinants of Economic Growth in OECD Countries, 1971–98

Country

Australia 1.68 0.13 -0.37 0.20 0.52 -0.25 0.03 0.01 -0.41 0.40Austria 1.57 0.02 -0.41 0.07 0.26 0.01 0.05 0.00 0.03 0.01Belgium 1.66 0.11 -0.53 0.02 -0.15 0.20 0.03 -0.05 0.53 0.06Canada 1.32 0.23 -0.90 -0.21 0.62 -0.18 0.04 -0.07 0.14 0.32Denmark 1.69 0.14 -0.57 0.28 0.21 0.12 0.02 -0.14 -0.05 0.27Finland 1.82 0.27 0.51 0.05 0.02 0.15 0.00 -0.06 -0.26 -0.14France 1.35 0.20 -0.59 -0.09 -0.10 0.07 0.07 -0.08 0.05 0.48Greece 1.15 -0.40 2.00 0.19 -0.56 -0.07 -0.16 0.17 -0.51 -1.48Ireland 3.02 1.47 1.54 -0.18 -0.32 -0.18 0.01 0.09 0.17 0.34Italy 1.73 0.18 0.22 -0.13 -0.69 0.13 0.02 0.01 0.14 0.48Netherlands 1.26 -0.29 -0.47 -0.03 0.25 0.01 0.06 -0.13 0.52 -0.50New Zealand 0.53 -1.02 0.34 -0.17 0.31 -0.29 -0.07 0.10 -0.36 -0.87

Norway 1.72 0.17 -0.12 -0.05 0.35 0.07 0.03 -0.06 -0.04 -0.01Portugal 2.15 0.60 2.56 0.58 -1.20 0.07 -0.10 0.10 0.11 -1.52Spain 1.28 -0.27 0.73 0.04 -1.12 0.00 0.03 0.07 -0.14 0.11Sweden 1.20 -0.35 -0.60 -0.10 0.21 0.11 -0.10 -0.17 0.01 0.30Switzerland 0.81 -0.74 -1.75 0.08 0.59 -0.04 0.00 0.15 0.02 0.21United Kingdom 1.63 0.08 0.05 -0.21 0.17 0.15 -0.03 -0.02 0.31 -0.34

United States 1.93 0.38 -1.62 -0.34 0.63 -0.09 0.07 0.09 -0.25 1.89

Source: Bassanini et al. (2001), Table 9.

Ann

ual A

vera

ge

Gro

wth

Rat

e

Gro

wth

Diff

eren

tial

Init

ial C

ond

itio

nG

DP/

Pop

.

Inve

stm

ent

Shar

e

Hum

an

Cap

ital

Pop

ulat

ion

Gro

wth

Var

iab

ility

of

Infla

tion

Gov

ernm

ent

Con

sum

p.

Trad

e E

xpos

ure

Res

idua

l Cou

ntry

-sp

ecifi

c E

ffec

t


labour productivity was 0.23 percentagepoints above the OECD average for the peri-od. The last column reports the country-specific residual effect, which is that part ofthe growth differential unexplained by themodel. For Canada, it turns out that 0.32 per-centage points of growth are unexplained.Factors that impact on Canada’s growth rela-tive to the OECD average include:

> a high initial income, which tended toreduce Canada’s growth relative toother OECD countries, which startedthe period at much lower productivitylevels;

> a share of investment in GDP that waslower than in other countries;

> human capital levels that account for alarge positive effect on the Canadiangrowth differential (0.62 percentagepoints per year);

> openness to trade, which accounts for apositive 0.14-percentage-point growthdifferential; and

> population growth, government con-sumption levels and inflation variabil-ity, which account for very little of thegrowth differential.The model performs well except for two

countries, Greece and the United States. Theauthors note that Greece is an unusual case,which also raises some data issues. However,the US results are quite interesting. The largepositive unexplained residual for the UnitedStates reflects the inability of the model toexplain the acceleration of labour-productivi-ty growth in the 1990s. To that extent, it isclear that the explanation of growth beingoffered by this model is less than complete.Nevertheless, the model provides an impres-sive example of how far modern theory andeconometric methods can go in terms of

explaining the growth performance of indus-trial countries. Providing explanations for thecountry-specific effects remains an importantissue. There could be either social determinantsor other unaccounted-for economic determi-nants at work. It is important to emphasizethat it would appear that a large portion of eco-nomic growth can be accounted for by a rela-tively small set of determinants.

Social PolicyThe basic policy question to be

addressed is the extent to which social poli-cy might have consequences for productivi-ty. As most of the empirical work in the areahinges on differences among countries insocial policies, this subsection provides a briefreview of some indicators of social policy. InCanada, social government expenditure cov-ers a range of public-sector activities. A typ-ical classification scheme based on publicfinance theory would be as follows:

> Public goods and services — pure pub-lic goods such as national defence andgeneral public services such as admin-istration, legislation and regulation.

> Merit goods and services — quasi-pub-lic goods provided on grounds of mar-ket failure, externalities or economicjustice principles. For example, govern-ment provision of education is commonbecause citizens may ignore the socialreturns of human capital investment ormay have limited access to capital mar-kets. Health care is another example.

> Economic services — private goods orservices prone to natural monopoly orstrong externalities. Examples includepublic utilities and financial supportfor specific activities such as researchand development.

290

Richard G. Harris


> Social transfers — transfers providingsupport for income and living stan-dards that have declined sharply or toindividuals who face exceptionalexpenses due to old age, disability,sickness, unemployment or family cir-cumstances.Under this classification, social policy

would tend to be defined in terms of spend-ing under the merit goods and services and socialtransfers categories. An alternative perspec-tive would be focused not on the classifica-tion of spending, but more directly on thegoals of social policy. Social policy pursues anumber of goals, including:

> increasing self-reliance> readjusting intergenerational burdens> improving flexibility and economic

growth> reducing the incidence of low income

and child poverty> improving the efficiency and quality of

service delivery> improving public finances> improving social cohesion> ensuring that basic social needs are met

Clearly, economic growth is one goal,but only one of many, and almost certainlynot the most important. The recent socialpolicy debate in many OECD countries tend-ed to emphasize the cost side of the ledger.The incentive cost argument emphasizes thatsocial protection can generate long-term wel-fare dependency and the capacity for flexibleadjustment to shocks. The funding of socialsecurity contributions in the form of payrolltaxes or general tax revenues increases the dis-tortionary welfare cost of taxation. High socialsecurity and health-care contribution liabil-ities for employers and other non-wagelabour costs can lead to lower employment,

especially for low-wage unskilled workers.All of these factors might contribute to lowerproductivity growth.

However, in principle, social programs canfacilitate economic adjustment and thus eco-nomic growth. For example, unemploymentbenefits can provide replacement income whilepeople search for a job. Social protection providescollective insurance to cover risks that may occurduring a person’s life (such as unemployment,sickness, disability, maternity), usually at a muchlower cost than if such risks were insured pri-vately, leading to increased investments inhuman capital and greater mobility. Activemeasures to encourage and facilitate labour forceparticipation contribute to economic growth byenhancing the flexibility of the labour force.Policies to improve the health and safety of theworkforce can increase labour productivity.13

Assessing the productivity effects ofsocial policy is inherently difficult. Aside fromthe direct human capital effects, a lot of theimpact is likely to be indirect, workingthrough changes in incentives to invest, saveor work or through the induced fiscal effects onsimilar variables. The search for empirical reg-ularities linking growth to social policy isalmost non-existent. OECD comparisons areinevitably going to be the data most discussedin this respect. To make matters worse, thesecomparative data are almost all related toexpenditures — that is, they measure inputsto social programs but not their outputs, whichwould be preferable in a productivity study.The growth literature has investigated quiteextensively two categories of public spending— public investment and government con-sumption. Generally, the results are mildlyfavourable toward the productivity or growtheffects of public-sector investment, and dis-tinctly negative with respect to public-sector

291



consumption, as is illustrated by the resultsreported in the last subsection. However, nei-ther of these captures what would be calledvarious forms of social expenditure. Differencesbetween countries in social spending is theonly form of evidence available thus far toestimate the growth effects of social policy.

Under the public finance classificationof spending, Canada tends to spend relativelylittle on what might be called public goods oreconomic services. Of total public spending, agreat deal is accounted for by social spending.In 1995, public goods accounted for 2.6 per-cent of GDP, merit goods (health, educationand other social services) 12.3 percent, incometransfers 11.5 percent, economic services 2.4percent and interest on the public debt 9.6percent. However, comparative numbers aremore interesting. Table 2 compares Canada totwo other countries perceived to be at oppo-site ends of the social policy spectrum withrespect to spending on education, health andtransfers — Sweden and the United States.

While there were substantial differencesamong the three countries in 1980, someconvergence has occurred between Canada andthe United States while Sweden continues tostand out in its spending on social transfers.

Here are some other characteristics ofOECD social spending patterns worth noting:

> A well-established empirical regularityin public finance is what is known asWagner’s Law. The demand for certaintypes of social protection rises more thanproportionately with the level of percapita income. While this relationship isnot observed in a cross-section of coun-tries, it holds very strongly in almostevery national time series on publicexpenditure. This fact, usually explainedby using simple arguments about voterpreferences, implies that economicgrowth is likely to have a positiveimpact on social spending, confoundingthe detection of causal channels runningin the other direction — from socialspending to economic growth.

> Much of what government does is redis-tributive (Boadway, 1998), but theinteresting fact is that the bulk of theredistribution is not from the rich to thepoor. During the 1980s and 1990s thereforms to the personal tax system innearly all OECD countries and the pres-sure on public budgets meant that thegenerosity of benefit schemes wasreduced. While benefit systems redis-tribute income, they redistribute prima-rily not from the rich to the poor but,rather, from the young to the old, fromthose who work to those who do not,and from childless families to familieswith children. Social policy, therefore, is

292

Richard G. Harris

TABLE 2

Selected Social Expenditures as aPercentage of GDP, Canada, Sweden and the United States

1980 1990 1995

HealthCanada 5.0 5.4 5.8Sweden 8.4 7.6 5.7United States 4.0 5.2 6.5

EducationCanada 5.4 6.7 6.5Sweden 7.6 6.8 6.6United States 5.3 5.3 5.0

TransfersCanada 8.1 10.8 11.5Sweden 16.5 19.2 21.2United States 9.3 8.5 9.4

Source: OECD (2000).


not primarily directed at equity per se,and its growth effects are dependent onthe details of specific programs.

> There has been a general and persist-ent upward trend in total governmentspending within OECD countries.From 1970 to 2000, the OECD aver-age went from 29.2 percent to 36.5percent of GDP. Canada went from33.8 percent in 1970 to 46 percent in1990, and then down to 37.8 percentin 2000 with the successive Martinbudgets. The major factor to whichmost analysts attribute this growth isthe creation and expansion of pro-grams and the provision of services inthe social policy area. The incomesupport element of these entitlementsis reflected in a persistent rise inincome transfer payments until themid-1990s.The common social policy experience

of so many countries points to the difficultyinherent in attempting to use these variablesto explain differences in the growth experi-ences within OECD countries. However,there are some notable differences, as noted

above, and these will prove important in theidentification of the effects of social expendi-tures on productivity.

INEQUALITY, SOCIAL POLICY AND PRODUCTIVITY

In this section we review the theoreticaland empirical literature that points to a causallinkage running from inequality and socialpolicy to productivity growth. It is instructivefirst to assess what has been a key driving forcebehind the policy dimension of this debate —the recent changes in income inequality.Looking at the total income of the workingpopulation, the changes have not been as dra-matic as one might imagine from the populardebate on this topic. In Table 3, the levels andchanges of two standard inequality indexes,the Gini coefficient and the ratio of income ofthe 90th decile to the 10th decile are record-ed.14 It is well-known that total incomeinequality rose in the United States and theUnited Kingdom from the mid-1970sthrough the mid-1980s. These trends werenever as evident in other countries. However,

293


TABLE 3Inequality Levels and Changes, Working-age Population, Mid-1970s to Mid-1990s

Levels Absolute Changes Between Periods

P90/P10 P90/P10 P90/P10Gini Decile Gini Gini Decile Decile

Coefficient Ratio Coefficient Coefficient Ratio Ratio

mid-70s / mid-80s / mid-70s / mid-80s /mid-90s mid-90s mid-80s mid-90s mid-80s mid-90s

Canada 28.7 3.9 0.1 0.1 -0.1 0.0Sweden 24.7 3.1 -0.6 2.3 0.0 0.2United Kingdom 30.4 4.1 3.7 2.7 0.7 0.4United States 33.3 5.3 2.9 0.6 1.0 -0.1

Source: Förster and Pellizzari (2000).


from 1985 to 1995 the trends slowed some-what. The effects on the distribution ofincome for the working-age population areshown for four countries: Canada, the UnitedKingdom, the United States and Sweden.While the level of inequality of income with-in the working-age population would be con-sidered higher in Canada than in Sweden,there has been virtually no change from themid-1970s to the mid-1990s. However, withrespect to market income, the underlyingtrend has been similar in most countries. Arecent OECD summary of the trends withrespect to Canada is provided in Box 1.

What has happened in Canada is typi-cal of a number of OECD countries: from the1980s to the mid-1990s there was a fairlysignificant change in the distribution of mar-ket income towards the upper end of the dis-tribution despite the relatively mild changesin total inequality, which measures incomeafter taxes and transfers.15 Specifically forCanada, from 1988 to 1995, changes record-ed in the market income share of differentdeciles are presented in Table 4.

There is little doubt that these data havebeen a major factor behind the renewed inter-est in growth and inequality. Specifically, it isbeing argued that there is a causal chain run-ning in the following sequence:

Social policy ➾ Incomeinequality ➾ Economic growth

with the presumption that increased incomeinequality lowers growth. The debate wasgiven a great deal of impetus by two relateddevelopments in the field of economicgrowth: first, an empirical finding thatclaimed to show a positive link between lowerinequality and higher growth, based on cross-sectional growth regressions; second, sometheoretical work in the new growth theory tra-dition that provided a rationale for this link.In this section, we look at both developments.Finally, it should be pointed out that it haslong been recognized that causal links couldalso run the other way — from growth toinequality — although the sign of the effectis largely viewed as ambiguous. In the broadsweep of evidence on the IndustrialRevolution and economic development, thereceived wisdom was summarized by a con-cept known as the Kuznets (1955) curve,which showed that as income levels riseinequality first increases and then decreases.However, the existence of an inverted U-shaped Kuznets curve says nothing directlyabout growth and inequality, other than toargue that as income levels get sufficientlylarge, inequality will fall.

Growth-Inequality RegressionsEvidence on the positive link running

from inequality to growth was first providedby Persson and Tabellini (1994), who lookedat cross-sectional and time-series data for bothdeveloping and OECD countries. They founda significant order of magnitude effect of inequal-ity on growth. The equations were a reduced-form growth regression with per capita GDPgrowth as the dependent variable and con-trolled for the initial GDP level (per capita)

294

Richard G. Harris

TABLE 4

Market Income Share Levels and Changes, Canada

Share in Change over1995 1988-95

(percent) (percentage points)

Three Bottom Deciles 9.6 -0.9Four Middle Deciles 35.5 -0.5Top Three Deciles 54.9 1.4


and schooling. They estimated that a 0.07-increase in the income share held by the top20 percent of the population lowered thegrowth rate of per capita income by just under0.5 percent — a very large effect. They arguedthat this result also holds for OECD historicaldata. Using a 70-country post-war data set,Alesina and Rodrik (1994) found that a one-standard-deviation increase in the Gini coef-ficient of land distribution affects growth ratesby 0.8 percentage points per year. A numberof studies came to similar conclusions,although it is important to note that themajority of these studies were done with sam-ples dominated by developing countries.16

Very few empirical variables that havebeen asserted to explain growth have not goneunchallenged. The same can be said forinequality within samples of both OECD anddeveloping countries. Here are some of theissues that have been raised in the growth-inequality context:

> Empirical growth regressions are very sen-sitive to the set of explanatory variablesused. The significance and magnitude ofcoefficients often change when the set ofexplanatory variables changes. For exam-ple, most theory suggests that bothinvestment levels and human capitalshould be important conditioning vari-

295


BOX 1

Inequality Trends in Canada: An OECD Summary

In Canada, the distribution of disposable incomes remained broadly stable over the last two decades,

and some summary measures point to a slight decrease in inequality. This holds for both the working-

age and the elderly population. During the first period, mid-1970s to mid-1980s, there was some "hol-

lowing out" of the middle incomes, as both the bottom and the top incomes gained income shares at

the expense of the middle incomes. This trend did not continue into the second period, from the mid-

1980s to the mid-1990s. Real incomes, on average, did not improve in Canada over the last 10 years;

they fell for the upper incomes while the real value was maintained for those at the bottom. There was

redistribution across age groups in the last ten years: relative incomes of the elderly, in particular older

senior citizens, increased more than in all other OECD countries (Austria excepted), namely by 3 per-

cent for those aged 55 to 64, by 8 percent for those aged 65 to 74 and by 10 percent for those aged 75

and over. All other age groups lost ground.

As in most other countries, the share of market income, in particular capital and self-employment

income, going to the bottom deciles among those of working-age decreased, and related to that, tax

shares fell, too. At the same time, Canada is one of the few countries in which the transfer share of bot-

tom incomes did not increase during the past ten years. Nevertheless, a decomposition of levels and

trends in inequality among the working-age population shows that both taxes and transfers contributed

to equalize the distribution of disposable incomes over time. As in a majority of countries, a process of

"employment polarisation" took place in Canada in the last ten years. However, both fully employed

and workless households increased their relative incomes while those of multi-adult households with

only one worker fell. The contributions of these three groups to the slight decrease in overall inequali-

ty were different: while inequality within and between those groups contributed largely to the decrease,

structural changes drove overall inequality up but did not outweigh the other decreasing effects.

Source: Förster and Pellizzari (2000, pp. 36-37).


ables. Barro (1999) noted this sensitivityand specifically found that when fertilityrates are included in the full sample(developed and developing countries), theinequality variable becomes insignificant.

> One of the major problems in thisdebate relates to the inclusion of bothdeveloping and high-income countriesin the data sets. These countries differnot only in income per capita but alsofor a wide range of political and insti-tutional factors. The convergence liter-ature on developing countries has cometo the conclusion that there appears tobe evidence of non-convergence, sug-gesting that these differences are verypersistent. How this should be dealtwith statistically is a major issue.Purely cross-sectional methods have thedisadvantage of imposing commonparameters on a number of effects thatmight be expected to differ betweencountries at different levels of develop-ment. One way around this issue is touse dynamic panel methods of estima-tion that attempt to use both time-series and cross-sectional variation as away of identifying the determinants ofgrowth while controlling for country-specific effects.17 One of the first to usethis methodology with respect to theinequality issue was Forbes (2000), whofound that once country-specific fixedeffects were included, changes ininequality either had the opposite effecton growth rates or were insignificant.

> Arjona et al. (2001) adopt a panelapproach to look specifically at this issueand at the level-of-development issue ina sample of OECD countries. They usethe transitional version of the Mankiw-

Romer-Weil model, in which growthdepends on population growth, invest-ment, initial income and human capi-tal. They find virtually no evidence thatinequality affects growth.

> Another major issue is causality. Astandard criticism of much of the cross-sectional growth literature is that onecan never be certain that correlation iscausation. Usually, there is an attemptto control for this by using data cover-ing long periods of growth as well asconditioning variables measured at thebeginning of the period. More sophis-ticated studies will often try to esti-mate a structural model in which thecausal linkages are more precise. Thereare a number of different theories link-ing inequality to growth, and thetransmission channel is quite differentin each case. It is unfortunate that therehave been few attempts to identify theunderlying structural link. For exam-ple, if increased inequality is assumedto lower human capital investment itwould be useful to check whether thisstructural relationship exists. Perhapsfuture work will take this into account,but at the moment it is a major weak-ness of the underlying methodology.18

Should any of this be very surprising?Hardly, for two reasons. First, it has longbeen known that relatively few variables arerobust in growth regressions (see Levine andRenelt 1992; Sala-i-Martin 1997). Second,there are the basic data one has to work with.With a few exceptions, there is not muchvariation in inequality across OECD countriesrelative to developing countries. The UnitedStates and the United Kingdom tend to havehigher levels of inequality, but their long-

296

Richard G. Harris


term growth performance was not very dif-ferent from that of most other industrializedcountries until very recently. The recent surgein US growth has, if anything, added to theperception that the causality runs in the otherdirection. Chart 1 presents a simple plot ofgrowth versus average income inequality.

The chart is plotted for the subset ofolder OECD countries (it excludes the recentjoiners: Mexico, South Korea, Greece, Spain,Portugal and Turkey). Not surprisingly,there is not much to be detected here usingocular statistical methods. The search for amore complicated correlation in these data islargely what the empirical debate has beenabout.

On balance, the empirical case for alink running from growth to inequality forthe high-income countries is at best statisti-cally fragile and at worst insignificant. Notethat none of this points to the opposite con-clusion — that increases in inequality causehigher economic growth.

The Theoretical LinkagesOften in economics, in the absence of

decisive evidence for or against a hypothesis,economic theory plays an important role indetermining the priors of economists both associal scientists and as policy advisers. Part ofthe renewed interest in this debate is the newtheoretical literature that shows that increasesin inequality can hurt growth. Most of this the-ory is rooted in endogenous growth theory,19 inwhich productivity growth is an endogenouscharacteristic of the economic system. Recentsurveys that focus on inequality include Aghionet al. (1999) and Lloyd-Ellis (2000). As it turnsout, however, these theoretical developments,while insightful, do not establish a strong case.They provide interesting examples of models

where changes in inequality can lead to lowergrowth under highly specialized assumptions.To obtain these results, one must dramaticallysimplify the models themselves. Now, this isnot a criticism. It merely serves to point outthat often, in economics, theory does not sug-gest a one-sided causal pathway between twovariables. In this particular case, there is also anolder literature that suggests the opposite effect— higher inequality can raise growth. There isalso a political economy literature that empha-sizes the endogenous nature of policy andgrowth consequences.

A brief summary of the theoreticalarguments is provided below.

Traditional Theory> Kaldor (1957): With savings-driven

accumulation, and assuming the richhave a greater propensity to save thanthe poor, more inequality leads togreater savings, which can lead tohigher transitional growth rates.

297


45

40

35

30

25

20

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Gin

i ind

ex

Percentage rate

CHART 1

Growth and Inequality Scatter Plot,Average Annual Productivity Growth,OECD, 1971–98


> Large investment indivisibilities: Assumingthat capital markets are very imperfect, sig-nificant individual wealth accumulationmay be necessary to make an invest-ment. More inequality could help growthin these circumstances by facilitating theconcentration of large pools of investmentfunds.

> Incentive or Mirrlees-type theories(Mirrlees 1971): With imperfect mon-itoring of contracts due to transactioncosts, moral hazard is to be expected.Borrowers using traditional debt con-tracts are quite likely to behave oppor-tunistically and not always in thelenders’ interest. In such cases, optimalcontracts should reward output, andwith heterogeneity among borrowersthe successful would be rewarded, notthe unsuccessful. This implies a needfor ex post inequality in rewards tomaintain incentives. Similar argumentscarry through to the taxation of savingsin endogenous growth models drivenby capital accumulation. Taxing sav-ings results in lowered growth (Rebelo1991). Both classes of arguments sug-gest that increased income inequality,as opposed to more equality supportedby a highly progressive tax system,leads to higher growth.

Political Economy Models(Persson and Tabellini 1994)

> Inequality affects taxation through thepolitical process: In unequal societies,more voters prefer redistribution assum-ing the median voter determines policyoutcomes. They consequently vote forredistribution, which reduces the incen-tives to invest and hence lowers the

growth rate. Note that this argumentassumes that: more inequality ➾ moreredistribution ➾ less growth.20

> Social protection reduces growththrough rent-seeking: This argumentwas made by Lindbeck (1975, 1995),who looked at the link between growthand social protection. He suggested thatthe universality of Scandinavian welfarestates politicized the returns to econom-ic activity and thus encouraged peopleto seek material gain through the polit-ical process, by passing redistributivelegislation, rather than through entre-preneurial and innovative activity.

> A variant on the first set of theories, butwith reverse implications, assumes thatinterest groups determine policies andthat a strong social safety net exists: Inthe presence of a free-rider problem,interest groups work hard at preventingpolicies that hurt them but that other-wise may have positive, widely diffusedgrowth effects (e.g., trade liberalization,labour market reforms). With socialprotection, these losses are partiallyinsured against, thus reducing the oppo-sition of interest groups to growth-pro-moting policies and increasing thelikelihood that they will vote in favourof such measures.

New Growth Theory> Imperfect markets and diminishing

returns to investment: Aghion et al.(1999) refer to this as the opportunity-enhancing effect of redistribution withimperfect capital markets. Given thediminishing returns on individual invest-ments and restrictions on the ability ofindividuals to pool funds, people with

298

Richard G. Harris


large endowments have low marginalreturns on investment, and converselyfor the poor. Redistribution from therich to the poor raises the average returnand thus enhances growth.

> Reversal of the traditional incentive argu-ment: This argument stresses the Mirrleescase, but with the added assumptions thatthe effort of borrowers is related to initialincome and that limited liability effectsare important. Let us assume that theprobability of success of an investmentproject depends on the effort of the bor-rower, but that moral hazard exists for theusual reasons. With limited liability, indi-vidual borrowers do not bear the risk offailure (the lenders lose), and this affectstheir effort. If the effort increases the bor-rowers’ own wealth, then redistributiontowards poor borrowers will have a posi-tive effect on their effort, thus promotinggrowth. Aghion et al. argue that redistri-bution will increase the effort because itreduces borrowing by the poor, who nowget a larger share of residual output; witha larger share, they have an incentive towork harder.As is evident, there are a variety of the-

ories suggesting alternative linkages betweeninequality and growth. Note that most eco-nomic theories hinge on one market failureargument or another, and particularly onimperfect capital markets. In the case of adeveloped country, this would seem to makesense only in the context of human capital,given well-developed capital markets forother forms of investment in physical capi-tal. If redistribution is to occur, it wouldhave to be financed by distortionary taxes onwages and savings. This would have the tra-ditional negative-incentive effects on growth,

which are offset or perhaps overcome by theopportunity-enhancement effect. However,it is far from evident that the appropriatepolicy to stimulate growth is passive redis-tribution of income. With inequality ofaccess to investment across individuals, amore suitable policy response would be toeither (a) reform financial institutions and mar-kets such that able individuals could invest ineducation, or (b) provide more direct supportfor public education.

The political economy theories point tothe fact that one must distinguish carefullybetween three related factors: inequality, whichcan be measured before the tax and transfer sys-tem apply; redistribution, which is income-based;and social insurance, which is situation-specific.Depending upon the assumptions made, moremarket income inequality before taxes and trans-fers may lead to greater or less redistribution expost. Lindbeck views social protection as induc-ing greater political rent-seeking, whose oppor-tunity cost is growth; the other view of socialpolicy is that it provides insurance in a worldwith insufficient private markets for insuring riskagainst sickness, unemployment and so forth.

Thus, social safety nets (a) promoteindividual investments in human capital, and(b) reduce political opposition to growth-pro-moting adjustments and policies. Which ofthese effects is more important?

In this instance, economic theory pointsto interesting hypotheses and provides theempirical economist, or policy-maker, withsome insight on what road marks to look forin determining the set of interactions amongstvariables. Beyond that, however, the theoriesthemselves are too diverse and too malleableto changes in assumptions or parameter choiceto form a basis for reliable policy formulationwithout empirical validation.

299



Social Policy and GrowthEvidenceIt is entirely possible, and theoretically

reasonable, that social policy might affectgrowth without greatly affecting income dis-tribution. For example, many of the theoreticalarguments about the consequences of active-labour market policies suggest that these could,in principle, be growth-enhancing. These samepolicies might also reduce the degree of mar-ket-income inequality, but one cannot be cer-tain of this without carefully specifying thedynamic feedback effects from growth toincome distribution. It is, however, reasonableto ask whether one can empirically identify thelinkage between social policy and growth with-out reference to an intervening effect oninequality. Unfortunately, very few studies havebeen published on this issue, and it is one thatrequires further research. There is a fairly well-developed body of evidence on the effects ofgovernment spending on growth, but it gener-ally does not distinguish government spendingdirected at a social policy objective from spend-ing directed towards other objectives.21 A largenumber of studies on the growth consequencesof fiscal policy have documented a significantand negative effect of government consumptionon growth.22

One innovative study that attempts tolook specifically at social policy for OECDcountries is Arjona et al. (2001, 2002). Theauthors use a framework similar to that dis-cussed in the second section of the presentpaper to infer the impact of social expenditureson growth in OECD countries. The growthequation is a Mankiw-Romer-Weil transition-al one that controls for investment and humancapital intensity across countries, and is esti-mated using an annual sample of 21 OECDcountries over the period 1970-88. The authors

find virtually no evidence that post-tax-transferinequality affects growth rates in OECD coun-tries. There is some evidence that total gov-ernment spending on social programs reducesgrowth. The magnitude of the effects is conse-quential. In the basic model, with aggregatesocial expenditure as a fraction of GDP, thecoefficient is -0.134. This compares with acoefficient on the investment share of 0.345.Both are significant at the 95-percent level.23

Quantitatively, the implication is that if onewere to decrease social spending by 1.0 percentof GDP and increase investment by 1.0 per-cent of GDP, the impact on aggregate labour-productivity growth would be on the order of0.5 percent per year. Not a large impact, butover a number of years it would begin to havea significant effect on income levels. Recallthat until recently annual labour-productivi-ty growth was in the area of 1.5 percent.

The authors do find, however, that whensocial spending is disaggregated by functionthe results are cleaner in terms of both signifi-cance and magnitude. Passive social spendingis prejudicial to growth, while active socialspending promotes growth. Interestingly, theyalso find that when the definition of activesocial spending is expanded to include healthexpenditures, the coefficient estimates on socialspending become insignificant. When theyinclude both passive and active social spend-ing as explanatory variables the coefficient onpassive social spending is significant and neg-ative, while the coefficient on active socialspending is significant and positive. The ordersof magnitude are interesting. The coefficientestimates imply that a shift of 1.0 percent ofGDP from passive to active spending producesa positive effect on growth of about 0.5 per-cent. Overall, the results suggest that socialexpenditures that promote adjustment and

300

Richard G. Harris


labour market participation tend to increaselabour-productivity growth, while other formsof social expenditures do not contribute togrowth and in fact may reduce it.

Obviously, one should interpret theseresults with caution, given the limited time-series variation in the data and other poten-tially omitted variables in the growth equationsuch as R&D and openness. Nevertheless, thisis a good start on an important research andpolicy issue.

An alternative and in many ways unre-lated body of evidence links social capital toeconomic growth. Social capital as defined byPutnam (1993) and Woolcock (1998) refersto the nature of trust in societies engenderedby various forms of community association.One of the best known and most representa-tive definitions can be found in the highlyinfluential work of Putnam (1993):

Social capital…refers to features of socialorganisation, such as trust, norms, and net-works, that can improve the efficiency of socie-ty by facilitating co-ordinated actions. (167)

To an economist, as Arrow pointed out longago, trust is an important substitute for mar-kets and contracts. A priori, one would imag-ine that more trust would imply highergrowth. The issue is pertinent to the debateon social policy because there is a strong pre-sumption that social cohesion and social cap-ital are closely related, as argued by Ritzenet al. (2000). A major objective of social pol-icy is the creation of social cohesion. Theseauthors argue that social cohesion creates anenvironment in which good policy is possi-ble if policy-makers are given room tomanoeuvre. The latter is created by reducingsocietal conflict over distributional objectives,in part through common institutions such associal policy.

However, the empirical evidence onsocial trust and growth is simply non-existent,so there seems to be little point in continuingin this vein. What evidence exists from cross-country comparisons based on the World ValuesSurvey seems to show that a higher index of trustactually leads to lower growth rates (see, e.g.,Knack and Keefer 1997). When Knack andKeefer exclude socialist countries and focus on amore recent period (1980-92), they get strongerresults. Controlling for initial income per head,human capital and the relative price of invest-ment goods, an increase of 10 percentage pointsin the level of their trust index (slightly less thanone standard deviation) is associated with anannual growth rate higher by 0.8 percentagepoints. Typically, the results are weaker whenattention is restricted to a sample of OECDcountries. Also using World Values Surveydata, Helliwell (1996) found that trust has anegative effect on growth in a sample of 17OECD countries. Knack (2000) reports that ina sample of 25 OECD countries the impact oftrust is imprecisely measured, and the hypoth-esis that it has no effect cannot be rejected atconventional significance levels. This literaturemay prove to be influential at a future date, butthus far there is little in it that could be used asa major justification for policy.

CONCLUSION

The linkages between economic growthand productivity are both complex and sub-ject to a variety of potential causal mecha-nisms. This paper has reviewed the evidenceand theory linking the social determinants ofproductivity growth, which include such fac-tors as the distribution of income and wealthin society; the set of social policies existing in

301



a country, including social insurance andredistributive programs and the education andhealth systems; and the degree of social cohe-sion. The complexity in uncovering a linkrunning from social factors to productivitygrowth is compounded by the fact that thesebroad institutional arrangements, includingthe social determinants but also the politicaland legal systems, may have indirect effects inthe long run that are difficult if not impossibleto detect in conventional economic data. Inspite of these problems, there is a new body ofresearch, both theoretical and empirical, thatattempts to identify the relationship amongsocial policies, economic inequalities and pro-ductivity growth.

The traditional economic debate on thesematters was usually framed in terms of theequity-efficiency tradeoff, in which more eco-nomic growth could be achieved only at theexpense of increased economic inequality. Thenewer literature suggests that, in fact, growthand social objectives may be complementsrather than substitutes. This certainly providesa more optimistic view of the choices facinggovernments than has been the case based onthe existence of a growth-equity tradeoff.

While these recent empirical and theo-retical contributions are interesting and sug-gest some important new areas for research, itis premature to assume that this literatureproves the existence of a robust linkage run-ning from social policy and inequality to pro-ductivity growth. One cannot conclude thatreduced income inequality leads to increasedproductivity growth or that more socialspending leads to increased productivitygrowth. The empirical evidence establishingsuch a linkage, which at this point is largelybased on macroeconomic cross-country com-parisons, either is simply not in the data or is

statistically fragile. Moreover, much of whathas been offered as evidence in favour of thishypothesis rests on developing-country data,which are of questionable relevance to anadvanced industrialized country like Canada.It is important to emphasize the recent ori-gins of this research. Virtually all of it hasbeen done in the last 10 years, and the totalnumber of studies is still quite limited. It ispossible, therefore, that our views based onthe weight of evidence will change in the nextfew years. The one major exception to theseobservations concerns education. There is avery large body of evidence showing thatincreasing education has a substantial effecton productivity. The role of human capital inCanada’s economic growth has been an endur-ing theme of both social policy and econom-ic policy. The evidence surveyed in Harris(2002) provides a strong endorsement of thisview. The evidence on health expenditures isless convincing, but in general the produc-tivity case for improving human capital iscompelling and warrants further research.

In summary, the major conclusion ofthis paper is as follows:

> The general case linking social policiesor inequality to productivity growthremains unproven. Justification for anyparticular social policy innovation mustrest on its cost-effectiveness in reachingits stated social goals. What little evidencewe have suggests that social policies pro-moting labour market participation, ratherthan passive cash-transfer programs, aremost likely to generate productivity ben-efits, although the magnitude of theeffects remains uncertain. A great dealmore research is necessary to link socialpolicies to productivity, particularly atthe micro level, before a productivity

302

Richard G. Harris


argument can to be used to promote aparticular social policy.To this can be added the further con-

clusions of Harris (2002):> Policies that have been proven to most

likely increase productivity are thosefocused on the proximate economiclevers to productivity growth — thosethat stimulate investment, innovationand competition, and facilitate theinternational diffusion of knowledge.

> The one social policy for which there isample evidence of positive productivityeffects is education. A substantial portionof Canada’s economic growth appears tobe attributable to the country’s high lev-els of educational attainment.

> The “new economy” perspective providesa coherent explanation of both recentgrowth and inequality trends as endoge-nous reactions to a common cause — theacceleration of technological change. Thegrowing evidence linking both recent andpast productivity data, together with evi-dence on wage inequality trends in indus-trial countries, provides a more coherentperspective from which to assess policieslinking productivity and inequality. Agrowth-oriented policy must both pro-mote technological adaptation throughinvestment and skills acquisition, andfacilitate the required structural changeacross regions, industries, firms and work-ers. Social policy can facilitate theseadjustments by providing the least well-off with the resources to make therequired investments in human capitalboth for themselves and for their children.The major rationale underlying social

policies in the modern mixed economy hasnever been higher productivity growth. The

general concerns for social justice and thepolitical demands of an increasingly wealthysociety for improved education, health andsocial insurance have long been the majorreasons why voters have requested these poli-cies in Canada. This will undoubtedly con-tinue to be true provided economic growthis sustained. Failure to increase or keep pacewith living standards in other advancedcountries is ultimately the greatest threat toCanada’s social programs. In that sense, pro-ductivity issues and social policy will alwaysbe linked.

NOTES

This article is an abridged version of Harris (2002),published with the permission of Industry Canada.

1 For a recent review of these arguments from aCanadian perspective, see Osberg (1995).

2 Krugman (1994) provides a very readable statementof this argument.

3 Also referred to as endogenous growth theory.Surveys of this field are presented in Aghion andHowitt (1998) and Jones (1999).

4 Harris (2002) contrasts the social determinants ofproductivity with more conventional economicdeterminants, such as investment and innovation, byexamining two specific social policies — educationand health — and the literature on majortechnological change, wage inequality and the neweconomy.

5 On aging and social security reform, see OECD (1998).The literature subsequent to the OECD Jobs Study isvoluminous. A review is provided by Disney (2000).

6 This section draws on material in Harris (1999).

7 For a brief and non-technical review of productivitymeasurement, see Harris (1999). For an extensivereview of the literature and a history of the subject,see Hulten (2000).

8 In the Canadian data, the majority of productivitygrowth is accounted for by TFP growth ormultifactor productivity (MFP) growth. MFP growthdata are published regularly by Statistics Canada.

9 A good example is Mokyr (1990).

10 For a recent survey, see Ritzen et al. (2000).

303



11 For the non-OECD sample, the model was actuallysomewhat more successful, although this result hasbeen criticized on a number of fronts.

12 These models almost always ignore adjustment costs,which is a serious problem in using them for welfareevaluations. With high adjustment costs, fastconvergence is not always a good thing.

13 These are covered in greater detail in the section “TheHuman Capital Dimension of Growth” in Harris (2002).

14 An increase in the Gini coefficient corresponds to anincrease in inequality.

15 Beach and Slotsve (1996) document these trends forCanada.

16 A survey of this literature is provided in Benabou (1996).

17 Contributions to the analysis of growth using paneldata sets and fixed-effects estimation include Barroand Lee (1994) and Barro and Sala-i-Martin (1995).

18 An exception is Perotti (1996), who looks at theeffect of inequality on female education and fertilityfor developing countries and finds a significant effect.This suggests that it may be the important causalchannel in developing country data.

19 For a comprehensive survey, see Aghion and Howitt (1998).

20 Aghion et al. (1999) claim that this is inconsistentwith evidence showing that redistribution has apositive effect on growth and that measures ofredistribution are uncorrelated with inequality —they cite Perotti (1994), whose Tables 4 and 8 reportregression results. The measure of redistribution isthe marginal tax rate.

21 There are a few older studies that claim to focus on the linksbetween social expenditure and growth. Unfortunately,they rely on the cross-sectional approach and most sufferfrom data deficiencies. Results have generally been mixed,but most come to the conclusion that social expenditure isbad for growth. See, for example, Landau (1985), Gwartneyet al. (1998), Hansson and Henrekson (1994), Lindert(1996) and Weede (1986, 1991).

22 This is the literature on fiscal policy and growth. Amodern example is Easterly and Rebelo (1993).Temple (1999) covers the evidence in his survey.

23 Results reported in Table 6.4, column 2, of Arjona etal. (2001).

REFERENCES

Aghion, P., E. Caroli, and C. Garcia-Penalosa. 1999.“Inequality and Growth: The Perspective of the NewGrowth Theories.” Journal of Economic Literature37(4): 1615-1660.

Aghion, P., and P. Howitt. 1998. Endogenous GrowthTheory. Cambridge, MA: MIT Press.

Alesina, A., and D. Rodrik. 1994. “Distributive Politicsand Economic Growth.” Quarterly Journal of Economics109(2): 465-490.

Arjona, R., M. Ladaique, and M. Pearson. 2001. “Growth,Inequality and Social Protection.” Labour Marketand Social Policy Occasional Paper #51. Paris:OECD Directorate for Education, Employment,Labour, and Social Affairs.

Arjona, R., M. Ladaique, and M. Pearson, 2002. “Growth,Inequality and Social Protection.’” Canadian PublicPolicy 28(2).

Barro, R.J. 1991. “Economic Growth in a Cross-Section ofCountries.” Quarterly Journal of Economics 106(2):407-443.

——. 1999. “Inequality, Growth and Investment.”NBER Working Paper #7038. Cambridge, MA:National Bureau of Economic Research.

Barro, R.J., and J.W. Lee. 1994. Sources of EconomicGrowth. Carnegie-Rochester Conference Series onPublic Policy #40, pp. 1-46.

Barro, R.J., and X. Sala-i-Martin. 1995. Economic Growth.New York: McGraw-Hill.

Bassanini, A., S. Scarpetta, and P. Hemmings. 2001.“Economic Growth: The Role of Policies andInstitutions. Panel Data Evidence from OECDCountries.” Working Paper #283. Paris: OECDEconomics Department.

Beach, C.M., and G.A. Slotsve. 1996. “Are We BecomingTwo Societies? Income Polarization and the Myth ofthe Declining Middle Class in Canada.” In The SocialPolicy Challenge, 12. Toronto: C.D. Howe Institute.

Benabou, R. 1996. Inequality and Growth. NBERMacroeconomics Annual #11. Cambridge, MA:National Bureau of Economic Research.

Boadway, R. 1998. “Redistributing Smarter: Self-Selection, Targeting and Non-Conventional PolicyInstruments.” Canadian Public Policy 24(3): 363-369.

Disney, R. 2000. “The Impact of Tax and Welfare Policieson Employment and Unemployment in OECDCountries.” Working Paper #WP/00/164.Washington, D.C.: Fiscal Affairs Department,International Monetary Fund.

Durlauf, S.N., and D.T. Quah. 1999. “The New Empiricsof Economic Growth.” In Handbook of Macroeconomics,ed. J. Taylor and M. Woodford. Amsterdam: North-Holland.

Easterly, W., and S. Rebelo. 1993. “Fiscal Policy andEconomic Growth: An Empirical Investigation.”Journal of Monetary Economics 32(3): 417-458.

304

Richard G. Harris


Forbes, K. 2000. “A Re-assessment of the RelationshipBetween Inequality and Growth.” American EconomicReview 90(4): 869-887.

Förster, M.F., and M. Pellizzari. 2000. “Trends andDriving Factors in Income Distribution and Povertyin the OECD Area.” Labour Market and Social PolicyOccasional Paper #42. Paris: OECD.

Gwartney, J., R. Lawson, and R. Holcombe. 1998. “TheSize and Functions of Government and EconomicGrowth.” Joint Economic Committee Paper.Washington, D.C.

Hall, R.E., and C.I. Jones. 1999. “Why Do SomeCountries Produce So Much More Output perWorker Than Others?” Quarterly Journal of Economics114: 83-116.

Hansson, P., and M. Henrekson. 1994. “A NewFramework for Testing the Effect of Growth.” PublicChoice 81(3-4): 381-402.

Harris, R.G. 1999. “The Determinants of ProductivityGrowth: Issues and Prospects.” Discussion Paper #8.Ottawa: Industry Canada. Also published inProductivity Issues in Canada, Industry Canada ResearchVolume, ed. S. Rao and A. Sharpe. Calgary: Universityof Calgary Press, 2002.

——. 2002. “Social Policy and Productivity Growth:What Are the Linkages?” In Productivity Issues inCanada, Industry Canada Research Volume, ed. S. Raoand A. Sharpe. Calgary: University of Calgary Press.

Helliwell, J. 1996. “Economic Growth and Social Capitalin Asia.” NBER Working Paper #5470. Cambridge,MA: National Bureau of Economic Research.

Hulten, C. 2000. “Total Factor Productivity: A ShortBiography.” NBER Working Paper #7471.Cambridge, MA: National Bureau of EconomicResearch.

Jones, C.I. 1999. Introduction to Economic Growth. NewYork: W.W. Norton.

Kaldor, N. 1957. “A Model of Economic Growth.”Economic Journal 67, 91-624.

Knack, S. 2000. “Associational Life and EconomicPerformance in the OECD.” Manuscript.Washington, D.C.: World Bank.

Knack, S., and P. Keefer. 1997. “Does Social Capital Havean Economic Payoff? A Cross-CountryInvestigation.” Quarterly Journal of Economics 112(4):1251-1288.

Krugman, P. 1994. “Part and Prospective Causes of HighUnemployment.” In Reducing Unemployment: CurrentIssues and Policy Options. Kansas City, MO: FederalReserve Bank of Kansas City.

Kuznets, S. 1955. “Economic Growth and IncomeInequality.” American Economic Review 45(1): 1-28.

Landau, D.L. 1985. “Government Expenditure andEconomic Growth in the Developed Countries:1952–1976.” Public Choice 47(1): 459-477.

Levine, R., and D. Renelt. 1992. “A Sensitivity Analysisof Cross-Country Growth Regressions.” AmericanEconomic Review 82(4): 942-963.

Lindbeck, A. 1975. “Inequality and Redistribution PolicyIssues (Principles and Swedish Experience).”In Education, Inequality and Life Chances 2. Paris:OECD, 229-385.

——. 1995. “Hazardous Welfare State Dynamics.”American Economic Review 85(2): 9-15.

Lindert, P. 1996. “What Limits Social Spending?”Explorations in Economic History, 33:1-34.

Lloyd-Ellis, H. 2000, May. “The Impacts of Inequality onProductivity Growth: A Primer.” Strategic PolicyResearch Paper #R-00-3E. Ottawa: HumanResources Development Canada, Applied ResearchBranch.

Lucas, R.E. 1988. “On the Mechanics of EconomicDevelopment.” Journal of Monetary Economics 22: 3-42.

Mankiw, N.G., D. Romer and D. Weil. 1992. “AContribution to the Empirics of Economic Growth.”Quarterly Journal of Economics 107(2):407-437.

Mirrlees, J.A. 1971. “An Exploration into the Theory ofOptimum Income Taxation.” Review of EconomicStudies 38: 175-208.

Mokyr, J. 1990. The Levers of Riches: Technological Creativityand Economic Progress. New York: Oxford UniversityPress.

OECD. 1994. OECD Jobs Study: Evidence and Explanations.Paris: OECD.

——. 1998. Maintaining Prosperity in an Ageing Society.Paris: OECD.

——. 2000. OECD Social Expenditure Database, 1980-1997. Paris: OECD.

Okun, A. 1975. Equality and Efficiency: The Big Tradeoff.Washington: Brookings Institution.

Osberg, L. 1995. “The Equity-Efficiency Tradeoff inRetrospect.” Canadian Business Economics 3(3): 5-19.

Perotti, R. 1994. “Income Distribution and Investment.”European Economic Review 38: 827-835.

——. 1996. “Growth, Income Distribution andDemocracy: What the Data Say.” Journal of EconomicGrowth 1(2): 149-187.

Persson, T., and G. Tabellini. 1994. “Is InequalityHarmful for Growth?” American Economic Review84(3): 600-621.

Putnam, R. 1993. Making Democracy Work. Princeton, NJ:Princeton University Press.

305



Rebelo, S. 1991. “Long-Run Policy Analysis and Long-RunGrowth.” Journal of Political Economy 99(3): 500-521.

Ritzen, J., W. Easterly, and M. Woolcock. 2000. “On“Good” Politicians and “Bad” Policies: SocialCohesion, Institutions, and Growth.” WorkingPaper. Washington: World Bank.

Romer, P.M. 1990. “Endogenous Technological Change.”Journal of Political Economy 98(5):S71-S102.

Sala-i-Martin, X. 1997. “I Just Ran Two MillionRegressions.” American Economic Review 87(2): 178-183.

Temple, J. 1999. “The New Growth Evidence.” Journal ofEconomic Literature 37: 112-156.

—— 2000. “Growth Effects of Education and SocialCapital.” Working Paper #263. Paris: EconomicsDepartment, OECD.

Weede, E. 1986. “Sectoral Reallocation, DistributionalCoalitions and the Welfare State as Determinants ofEconomic Growth Rates in OECD Countries.”European Journal of Political Research 14: 501-519.

——. 1991. “The Impact of State Power on EconomicGrowth Rates in OECD Countries.” Quality andQuantity 25: 421-438.

Woolcock, M. 1998. “Social Capital and EconomicDevelopment: Toward a Theoretical Synthesis andPolicy Framework.” Theory and Society 27: 151-208.

306

Richard G. Harris

Richard Harris text 11/27/02 2:21 PM 6

social policy, inequality and productivity · of social policy, such as education and ﬁscal...

Documents