The Welfare State and Economic Performance A.B

THE WELFARE STATE AND ECONOMIC PERFORMANCE A.B. ATKINSON*

Abstract - The Welfare State has come unern world. In the United States, the der attack from economists, particularly in movement to halt and, if possible, re- Western Europe, who argue that it is reverse the upward trend of government sponsible for poor economic performance, spending has gathered strength in re- and that public spending should be recent years. . In Great Britain [the duced. The present paper seeks to clarify Thatcher government] has committed the nature of the charges leveled against itself to a gradual reduction in the share the We/fare State and the mechanisms by of national income taken by public ex- which it may adversely affect economic penditure. . . Pressures to restrain the performance. The first section considers growth of government spending . . the aggregate empirical evidence. Not seem to be intensifying in other Euro- only is the evidence mixed, but also such pean countries too” (Morris Beck, Pref- an argument is difficult to establish. The ace to Government Spending, 1981, p. second section of the paper describes a ix). number of the problems with aggregate cross-country evidence. In particular, the interpretation depends on the underlying theoretical framework. The third and INTRODUCTION fourth sections of the paper examine a selection of the theoretical mechanisms, What Morris Beck wrote remains as valid distinguishing between those that affect today as it was 15 years ago, and in no the level of output, taking a model of the field of public spending is the pressure labor market, and those that influence for reductions greater than that of the the rate of growth, drawing on recent Welfare State. In many OECD countries developments in growth theory. An im- there are calls for the Welfare State to portant role is played by the institutional be scaled down. It is argued that the structure of benefits, which can signifi- size of transfer programs is responsible cant/y change their impact on economic for a decline in economic performance, behavior. and that cuts in spending are a prereq- uisite for a return to the golden age of “The size of government has become a full employment and economic growth. major issue in many parts of the West- The critique of government spending has *Nuffleld College, Oxford, United Kingdom, OX1 1NF been especially forceful in Europe, where

171 the Welfare State has traditionally played They conclude that a major social role. In Sweden, the Eco- nomics Commission, chaired by Assar “the agenda should be to make the Lindbeck and including distinguished Welfare State leaner and more effi- economists from other Nordic countries, cient” (Dreze and Malinvaud, 1994, p. has referred to “the crisis of the Swed- 82). ish model,” arguing that it has While recognizing the diversity of na- “resulted in institutions and structures tional circumstances within Europe, and that today constitute an obstacle to that in sorne countries spending may be economic efficiency and economic too low, their overall recbmmendation is growth because of their lack of flexibil- to ity and the/r one-sided concerns for income safety and distribution, with lim- “reduce expenditure in some countries, ited concern for economic incentives” perhaps by 2 percent of GDP or so” (Lindbeck ef a/., 1994, p 17). (Dreze and Malinvaud, 1’994, p 98). They seek cuts in social security benefit The present paper does not attempt ‘to levels in order that determine whether or ndt spending “the social-security (or social insurance) should be cut. The aim is rather to clar- system should not overburden the ify the nature of the charges leveled economy through distorted incentives against the Welfare State, and specifi- or large deficits” (Llndbeck et al., 1993, cally against social transfers, and the p. 238). mechanisms by which it may adversely affect economic performance. I consider In the European Union, a particularly in- in the first section of the paper the ag- fliuential document has been the paper gregate empirical evidence which ap- om “Growth and Employment: The pears to underlie much of the case Scope for a European Initiative,” pre- against the Welfare State,. Countries pared by Jacques Dreze and Edmond with high spending, it is alleged, have a Nlalinvaud, on the basis of discussions poorer economic performance. However, with a group of Belgian and French not only is the evidence mixed, but also, economists. This report emphasises the such an argument is more difficult to es- positive functions of the Welfare State tablish than it may at first appear: the but lists three major objections: second section of the paper describes a number of the problems with aggregate “(i) measures of income protection or cross-country evidence. In particular, the social insurance introduce undesired interpretatilon of such stubies depends rigidities in the functionng of labour on the underlying theoretical framework. markets; Aggregate empirics of the relation be-- (ii) welfare programmes increase the size tween the Welfare State and economic of government at a risk of inefficiency; performance are open to ‘the objection their funding enhances the amount of of being “measurement without the- revenue to be raised, and so the mag- ory.” The third and fourth sections of nitude of tax distortions; the paper examine a selection of the (iii) . welfare programrnes may lead theoretical mechanisms, distinguishing to cumulative deficits and mounting between those that affect the level of public debts” (D&e and Malinvaud, output and those that influence the rate 1994, p. 95). of growth.

172 I THE WELFARE STATE AND ECONOMIC PERFORMANCE

I should stress at the outset that my noticeably between countries: in 1960 concern in this paper is with the impact West Germany had the highest spending of the Welfare State on economic per- of the countries shown, but it was over- formance and not with the success of taken by Sweden around 1975. social transfers in meeting the objectives which they are intended to perform, The availability of such aggregate data such as the alleviation of poverty, the re- on a comparable basis for different distribution of income across the life countries means that it is tempting to cycle, and the provision of a sense of se- see how far there is an association with curity. The positive contribution of the differences in economic performance. It Welfare State clearly forms part of the would be possible to regress the level of overall balance sheet. A cut in govern- GDP, denoted below by V, on the size ment spending may well reduce the ex- of the welfare state, relative to GDP, de- tent to which social objectives are noted by WS. This kind of relationship is achieved, but here I am concentrating referred to below as a levels equation. on what may be the “cost” of the Wel- Alternatively, the rate of growth of GDP, fare State in terms of reduced economic denoted by g,, could be regressed on success. VVS. This kind of relationship is referred to as a growth-rate equation. [The dis- It should also be noted that I concen- tinction between these two hypotheses trate solely on social transfers (social se- is discussed by Bourguignon (1993), who curity and welfare) and do not consider shows graphically the difference.] other elements of the Welfare State such as education or health care. In view There have been many such empirical in- of the direct role that the latter may quiries. Some simply carry out a bivariate play in human capital formation, I am analysis of economic performance intentionally tackling the areas where against the size of the Welfare State. For the critique seems most likely to apply. example, the European Commission has examined the relationship between AGGREGATE EMPIRICS growth and social protection expenditure (percent of GDP) in the 12 member states. On the basis of a graphical plot It has been argued that a large Welfare of the change in employment between State has depressed economic perfor- 1980 and 1990 and the average social mance, causing output to fall below po- protection expenditure 1980-91 (alter- tential or for the annual growth rate to natively, the change in social protection be lower than in countries without such expenditure), they conclude that a level of transfers. This argument is often supported by reference to mea- “It is clear that there is little sign of so- sures of the size of the Welfare State, cial protectton having a negative effect typically measured as a proportion of on employment creation. The graph Gross Domestic Product (GDP), as illus-. shows a wide variety of combinations trated in Figure 1, which shows the ratio between employment growth and level to GDP of spending on social security of social protection . . . The same lack transfers.’ For the United States, the ra- of relationship is also apparent if the tio in 1990 was around ten percent; in change in social expenditure is taken” Sweden it was twice that amount. (European Commission, 1993, p. 86). Transfers increased as a percentage of GDP in all countries shown in Figure 1, In a much earlier study, Smith (1975) although the rate of increase differed found that the growth rate of real GDP FIGURE 1. Social Security Transfers as Percent GDP 25 r -- 20

0 1960 1970 19x0 1990 Note: linear interpolation before 1980. per capita 1961--72 was negatively re- cial security or other government translated to public spending excluding transfer payments; it should be stressed that fers but that the effect was smaller and the authors crted are noti concerned less significant when public spending in- solely with the impact of social transfers, cluded transfers: and that in some cases it represents only a minor part of their results. “it is less economically harmful for the The results of this kind of aggregate state to raise taxes and make transfer analysis are mixed. Of the nine studies payments than to consume resources shown in Table 1, two (Landau, 1985; directly” (1975, p. 29). Hansson and Henrekson, 1994) find an insignificant effect of the Welfare State Other investigators have argued that we variable on annual growth rates, four need to control for other influences on (Weede, 1986; Weede, li991; Nord- economic performance, embedding the strom, 1992; Persson and Tabellini, statistical analysis within a fuller model, 1994) find that transfers are negatively as in the work on growth empirics by associated with average rowth, and Barro (1991) and Mankiw et al. (1992). three studies (Korpi, 198 “5 ; Castles and There have been a number of studies Dowric:k, 1990; McCallum and Blais, examining the role of social transfers, 1987) find a positive sign1 to the coeffi- and a selection are summarized in Table cient of WS, although the last of these 1. The table shows that part of the find- authors finds evidence ofIa nonmono- ings of these studies that relates to so- tonic relationship.

174 TABLE 1 STUDIES OF GROWTH RATE AND SOCIAL TRANSFERS Study Landau (1985) Korpi (1985) Weede (1986) Dependent variable Real per capita GDP Real per capita GDP Real GDP and real per capita GDP Pooled time series/cross section Mixed time series/cross section Pooled time series/cross section Period Annual growth rates Period 1950-73 and subperiods Period 1960-82 and subperiods 1952-76 1950-9, 60-6, 67-73, 73-9 1960-8, 68-73, 73-9, 79-82 Countries 16 OECD (incl Japan) 17 OECD (excl Japan) 19 OECD (incl Japan) Model and variables Controls for investment and Total effect, but controls for percent Total effect, but controls for percent education. GDP, terms of trade, labor force in agriculture or GDP per agricultural employment. country intercepts capita (catchup variable) Age of democracy Estimation method IV and HS corrected; OLS OLS; WS measured at start of period. weighted and unweighted by unweighted Also applies Cochrane-Orcutt and population Hildreth-Liu. J Definition of WS variable General government transfers IL0 social security expenditure/GDP OECD social security transfers/GDP (OECD national accounts)/GDP (from historical statistics) (different deflators) Measured in percentage points Measured in percentage points Measured in percentage points Coefficient on WS 0.004 OLS unweighted 1950-73: -0.21 (standard error) (0.031) 0.193 (0.050) (n/a) or Table 2, equation 1 or 0.012 IV, HS corrected 1973-g: -0.19 (0.037) 0.182 (0.064) (n/a) or Table 5, equation 2 excl Japan and Switzerland 0.054 OLS weighted (UVS measured at start of period) (0.035)

Table 1, panel C Similar with catchup variable Table 4, col 2 Effect of five percentage point Not significant at five percent level 0.9 percentage point reduction in One percentage point increase in reduction in WS annual growth rate annual growth rate TABLE 1 CONTINUED Study McCallum and Blais (1987) Castles and Dowrick (1990) Weede (1991) Dependent variable Real GDP, per capita GDP, and per Real GDP Real per capita GDP person employed Pooled time series/cross section Pooled time series/cross section Pooled time series/cross section Period Subperiods 1960-7, 67-73, 73-9, Subperiods 1960-8, 69-73, 74-9,80-S Subperiods 1960-8, 68-73, 73-9, 79-83 79-85 Countries 17 OECD incl Japan 18 OECD or 17 exe! Japan 19 DECD inch Japan Model Controls for employment growth (IV) Controls for investment and Total effect and productivity per employment (or not) person employed Catchup (log GDP per capita), Catchup (log GDP per capita), sclerosis, Percent agricultural employment, age modernization, growth of govt exp/ subperiod dummies of democracy GDP, subperiod dummies

Estimation method OLS; WS measured at start of period. OLS and test for endogeneity OLS; WS measured at start of period. Also Cochrane-Orcutt and Hildreth- Liu Definition of WS variable OECD social security transfers/GDP OECD social expenditure less health OECD social security transfers/GDP (from historical statistics) adjusted and education at constant 1970 (from historical statistics) for percent aged 65t prices, extended 1982-5 using OECD national accounts Measured in percentage points Measured as fraction of GDP Measured in percentage points Coefficient on WS 012 for 1960-79 Controlling fo: emp and inv Productivity resuits: (standard error) (0.03) 5.24 or 7.45 -0.11 (n/a) (3.54) (3.53) 0.31 ws - 0.0092 ws* 0: (0.09) (0.0031) Not controlling -0.084 (n/a) for 1960-83 -1.01 or 1.93 excl Japan (3.74) (3.45) Table 1 Tabte 5, second estimate exct Japan 1960-8 Effect of five percentage point 0.5 percentage point reduction in Controlled estimates: 0.5 percentage point increase in reduction in WS annual growth rate (1960-79 0.3-0.4 percentage point reduction in annual growth rate of productivity estimate) annual growth rate of total factor productivity zero at WS = 16.8 percent with 1960- 83 estimates TABLE 1 CONTINUED Study Nordstrom (1992) Hansson and Henrekson (1994) Persson and Tabellini (1994) Dependent variable Real GDP Real private output in 14 industry/ Real per capita GDP service sectors Cross section Cross-country/cross-industry Cross section Period 1977-89 1970-87 1960-85 Countries 14 OECD incl Japan or 13 excl Japan 14 OECD incl Japan 13 OECD excl Japan Model and variables Total effect Controls for investment and Total effect employment Growth rate related to different types GDP per capita (catchup variable), of government spending/GDP Catchup variable percent attending primary school

Estimation method OLS; WS measured at start of period OLS IV unweighted Definition of WS variable Other current transfers in OECD OECD social security transfers/GDP OECD social expenditure series/GDP National Accounts (from historical statistics) (pensions plus unemployment camp. plus other social exp) Measured as fraction of GDP Measured in percentage points Measured as fraction of GDP Coefficient on WS -0.120 (0.034) -0.063 (0.036) - 6.723 (standard error) Table 1, col 2 Table 4, equation xi for WS average (5.396) Table 8, equation iii (and similar results for other 1965-82 specifications) or -0.050 (0.035) -0.119 (0.039) excl Japan equation xii for WS average 1970- Table 2, col 2 87 Effect of five percentage point 0.6 percentage point increase in Not significant at five percent level 0.3 percentage point increase in reduction in WS annual growth rate (but significant negative coefficient annual growth rate for total transfers) The most recent (Persson and Tabellinl) Castles and Dowrick (1990), and Weede study is primarily concerned with the re- (1 991).2 Among the points identified are lation between income inequality (and the following: growth, but the authors also examine (1) sensitivity in some, but not all, cases the relation between average growth to the country coverage, notably the in- (percentage points) In real GDP per cap- clusion or exclusion of Japan, il’a 1960-85. denoted by GROWTH, and social transfers (fraction of GDP), de- (2) differences of view as to whether it noted by TRANSF, in 13 OECD countries ns appropriate to include dummy vari- They conclude that there is ables shifting the intercept for different subperiods, “some weak evidence of a negative ef- (3) different definitions of the WS vari- fect from TSANSF on GROWTH” (‘1994, able, in particular the inclusion in some p. 617). cases of other government transfers This is based on an instrumental vari- apart from social security,3 ables estimate of the coefficient on (4) distinction between studies seeking TRANSF of - 6.7 with a t-statistic Iof to explain the total growth rate, and -- 1.2, which is not significant at the five those explaining the growth of factor percent level The point estimate implies productivity, controlling for the contrnbu- that a reduction in spending from 20 tion of factor input growth (investment percent to ten percent of GDP, approxi- and employment), and mately equal to the difference between (5) different right-hand variables apart Sweden and the United States, would from WS, and factor input growth, in- increase the annual growth rate by cluding the age of democracy or ‘“insti- about 0.7 percentage points, but the 95 tutional sclerosis” variables. percent confidence Interval is from --0.4 to + 1.7 percentage points. In contrast, The next generation of aggregate empir- the earlier study by Korpi (1985), not ical studies will no doubt build on this dissimilar in structure, but covering the earlier work, and a systematic explora- period 1950-- 79, concludes that tion of the different dimensions should reduce the degree of variety in the re- “social securtty expenditures . . . show sults. At the same time, there are poten- positive and significant rel,ationships with tially problems with any empirical analy- economic growth” (19851, p. 108). sis of this kind. This is based on an estimated coefficient on the WS variable of around 19.0 (rn PROBLEMS WITH AGGREGATE terms of percentage points) with a EMPIRICAL EVIDENCE t-statistic of 3.9. This implies that a reduction in spending from 20 percent to Aggregative empirical evidence may be ten percent of GDP would reduce the questioned in principle on a number of annual growth rate by 1 8 percentage grounds, as illustrated by the following. points, with a 95 percent confidence in- Causahty terval of 1 .O to 2.9 percentage points. As it was put in an OECD study, There are a number of reasons for such discrepancies Several authors have “in the assessment of the relationship sought to reconcile the differences in between the public sector and eco- findings, including Korpi (1985), Saun- nomuc performance, it might be thought ders (1986), McCallum and Blais (1987), useful to investigate whether or not THE WELFARE STATE AND ECONOMIC PERFORMANCE

[country differences] bear any system- Dynamic Specification atic relation to differences in the size and growth of public sector activity. It is dif- The potential difficulties in interpreting ficult to believe, however, that analysis the findings have been recognized in a undertaken at this level of aggregation number of the studies, which have ap- will shed much light on what are clearly plied a variety of solutions. Some use very complex underlying relationships. the initial period value of the WS vari- . . . statistical correlations between eco- able (see Table 1) on the grounds that nomic performance indicators and pub- regressions of growth rates of GDP on lic sector involvement are not likely to initial levels of WS would not be subject be easy to establish, and even harder to to simultaneity. This, however, raises a ascribe to underlying causal mecha- fundamental issue concerning the dy- nisms” (Saunders and Klau, 1985, p. namic specification of the estimated re- 122). lationship. Suppose that there is a negative relationship between social transfers It may be poor economic performance (measured by WS) and the level of GDP. that leads to high Welfare State spend- In an econometric equation with GDP as ing, rather than vice versa. Slow growth, the left-hand variable, we might want to or output below trend, may cause re- include both current and lagged values duced employment and hence higher of the WS variable in order to allow for spending on unemployment benefit and delayed responses to changes. For in- other transfers. Alternatively, it may be stance, if higher pensions were to re- successful countries, with high income duce aggregate savings, then the capital per head, that can “afford” a more gen- stock, and hence output, would fall erous social security system. Or it may gradually to its new long-run level. But be that industrialization of the economy what long-run restrictions do we want leads both to higher living standards and to impose on the estimated relationship? to the need for social security. The mod- As has been stressed in time-series ern employment relationship, with its econometrics, it is here that economic risk of catastrophic income loss, creates theory has an important role to play. the role for social insurance. We might therefore expect more advanced coun- There are indeed two different theoreti- tries to have larger Welfare States. This cal predictions. The first is that described would predict a positive relation be- above as the levels equation, where GDP tween Yand INS, although again the depends on the size of the Welfare causation would run in the reverse State. A cut in social spending induces a direction. temporary rise in the growth rate, as GDP rises to its new equilibrium level, The same applies to the growth rate ver- but there is no permanent increase in sion of the relationship. Suppose that the rate of growth. Cast in growth-rate the growth rate is fastest during the in- terms, the growth rate is related to the dustrialization period, approaching its change in the level of WS. The alterna- steady-state value from above (as pre- tive theoretical model is that where the dicted by a number of growth models), /eve/ of transfers affects the long-run and that state spending grows as the rate of growth, referred to above as the social insurance scheme matures. The growth-rate equation. In this case, a cut higher level of Welfare State spending is in the Welfare State is predicted to raise then associated with a slowing of aggre- the growth rate permanently. gate growth, again without there being any causal connection. These two kinds of equation have quite

179 different implications. Figure 1 shows social transfers (as percent of GDP) as spending/GDP = being broadly similar in Sweden and West Germany in 1975. In the next 15 (average benefit/aver+ge wage) years, they did not change greatly in x (average wage/GDt per worker) VVest Germany, but they increased in Sweden. Suppose that in the 1990s x (recipients/workers) transfers stabilize in Sweden at a higher (constant) percentage than in Germany. The first term is usually rkferred to as On the basis of the levels equation, we the replacement rate, thd second is the predict that GDP in Sweden would, wage share, and the thirb is the depen- when the adjustment is complete, grow dency ratio. Therefore, a Ispending ratio at the same rate as in Germany. The of 15 percent of GDP m+y correspond growth-rate equation, on the other to a replacement rate of ‘75 percent hand, predicis that growth in Sweden with a wage share of 60’ percent and a would be lower forever. Most of the dependency ratio of oneithird or to a elmpirical studies are concerned with the replacement rate of 30 piercent with a growth-rate version but the frequent ref- wage share of 75 percenk and a depen- erences to “leaky buckets” (loss of effi- dency ratio of two thirds! Put another ciency) appear to have in mind a levels way, countries may differ in the extent irlterpretatior?.4 of needs: one may have ia high spend- Measuring the Size of the Welfare ing ratio on account of ai large depen- State dent population, not on &count of a generous social security qrogram. This is A third problem concerns the rneasure- relevant if it is the generbsity of benefit ment of the size of the Welfare State, a levels that is believed to have an adverse question that has been extensively dis- impact on economic beh.&ior, since a cussed in the literature on “welfare ef- high level of WS does not necessarily fort.” Writers on soc:ial policy have imply a high level of generosity.” sought to relate this vanable to the success of different countries in reducing Of course, it rnay not be the amount of poverty or income inequality (for exam- benefit per recipient with which we are ple, Mitchell, 1991); writers on political concerned; it may be the1 cost per con- science have attempted to explain differ- tributor which is considerpd the relevant ences in the ratio of transfer spencling variable. It may simply be’the total cost to GDP by the existence of governments of the Welfare State that is a burden. of different political complexions and But in this case, a second1 objection other variables (see Wilensky, 1975, and comes into play, which is that the effective cost to contributors i the net effect the subsequent literature). 4 after allowing for taxatioi. In many But, it has been recognized in this litera- countries, part or all of sqcial transfers ture that there are serious problems with are subject to income tax1 and while measures of the size of the Welfare many beneficiaries may bf below the tax State. Statistics like those shown in Fig- threshold, some part of the gross outlay ure 1 can be quite misleading. To begin returns to the government via increased wlith, the level of spending relative to income tax receipts-to d~ifferent de- GDP does not necessarily provide an in- grees in different countrie/s. dication of the level of benefit per recipient, as is demonstrated in the following Taxation also comes into \he picture on decomposition: account of tax expenditur@. Allowances

180 1 IHt WtLkAKt >IAlt ANU tCuNuMI~ rtnrunwwCt

against income taxation may play the the person is making genuine efforts to same role as cash transfers. A higher tax seek employment. Benefit may be re- exemption for the elderly transfers in- fused where the person entered unem- come to those above a certain age with ployment voluntarily or as a result of in- the same effect (although a different dustrial misconduct, and a person may distribution) as a pension scheme. Re- be disqualified for refusing job offers. placing child income-tax allowances by a cash child benefit may leave the net fi- Not only do these conditions reduce the nancial position of a family unchanged. coverage of unemployment insurance, This is a further reason for considering but also they affect the relationship be- the net position, and a number of stud- tween transfers and the working of the ies have added tax expenditures to direct economy. The standard job-search social security payments when calculat- model, for example, assumes that working welfare effort (Gilbert and Moon, ers can reject job offers that offer less 1988). Moreover, we may want to take than a specified wage. Such a reserva- account of other “off-budget activities” tion wage strategy may, however, lead (Saunders, 1986) such as the regulation to their being disqualified from benefit. of the private sector or minimum-wage This institutional feature needs to be in- legislation. corporated and may change the predicted impact. A second example is pro- What both of these examples demon- vided by the contribution conditions, strate is the need to consider the pur- which may induce people to take jobs in pose for which the WS variable is to be order to requalify for subsequent bene- used. fit. Again these are often neglected.

Need to Examine the Fine Structure Disregard of institutional detail may of course be justified when it has no real The welfare effort literature has equally consequence. Thus it may be argued argued that the effectiveness of social that the limited duration of unemploy- transfers depends on the form of the ment insurance is irrelevant in many Eu- programs, and that one cannot base the ropean countries, since the person sim- analysis on a single aggregate spending ply moves on to unemployment variable. Reduction of poverty depends assistance. However, unemployment in- on the distribution of social spending, surance differs from assistance in impor- and the same is true if our concern is tant ways, such as the role of the con- with the impact of transfers on eco- tributory principle in providing an nomic performance. incentive for people to take insured employment. Another difference is that re- We have therefore to examine the fine ceipt of assistance depends on the in- structure of social transfers, to which come of other household members. This economists have in the past paid too lit- means that assistance payments affect tle attention. Unemployment benefit the incentives not just of the unem- provides an illustration, where economic ployed person but also of his or her models regularly assume that the only partner. Where a person moves from in- relevant condition for the receipt of ben- surance to assistance benefit, there may efit is that of being unemployed. In fact, be little financial advantage in the part- in the typical unemployment insurance ner continuing to work. program, benefit is subject to contribution conditions, is paid for a limited du- The significance of the fine structure is ration, and is monitored to check that that the same level of social transfers

181 may have quite different economic: impli- tionshlps. In this section, I explore a se- cations depending on the form of the lection of models of the determination transfer programs. Just what the rele- of the level of output. vant differences are depends in turn on the determinants of economic behavior. The simplest model of transfer payments is perhaps that of a recipient group, Conclusion fixed in size, and a working population on whose earnings is leviled an employer In his review of the lessons to be drawn payroll tax at rate t in orjier to finance from the aggregate empirics research for the transfer. Firms produce a single out- the future of the Swedish Welfare State, put, and for purposes of illustration, I Klevmarken concludes that take the Cobb-Douglas production “regardless of what result a cross- function: sectional regression would arrive at, it does not say much about how changes in the size of the public sector would affect growth in Sweden . . . it must be difficult to see different countries as ex- where K denotes capital, I! labor, and A perimental units which can provide in- the level of labor productivity, both K formation about one and the same pro- and A assumed constant lat present, and cess. At any rate, comparability must be p is the (constant) competitive share of clarified on a considerably more de- capital. The price of the output is taken tailed level” (1994, p. 16). as unity. Firms employ people up to the It is, however, not just the cross-section, point where the value m#ginal product but also the time-series, analysis which is is equal to the wage cost (w[l + t]), open to the objections sketched in this which generates a labor demand func- section. tron : In my view, we have to look inside the “black box” and provide an explicit theoretical structure and sufficient institutional detail. Without such1 a framework, it is not possible to interpret observed where c is a constant. aggregate relationships. Theory is neces- Workers are all equally productive in sary to specify the form of econometric market work but differ in their produc- relationships; the choice of indicators of tivity in horne employment (home out- the scale of the Welfare State depends put is valued at the same, price as mar- on the purpose for which they are to be ket output). There is a maximum total used; and it is theoretical models of economic behavior that identify the relevant institutional features of the transfer system.

WELFARE STATE AND THE LEVEL OF OUTPUT

In considering the theoretical structure, I foAlow the distinction drawn earlier between levels and rate of growth rela-

182 1 THE WELFARE STATE AND ECONOMIC PERFORMANCE

In this situation, we have a simple sup- duction. The Scandinavian professors ply and demand model of the aggregate who paint their own houses rather than labor market-see Figure 2. The effect write books are still contributing to out- of the social security payroll tax is to put. This is not just an accounting point. shift the demand curve to the left at Much of public debate confuses the po- every wand there is a fall in the equilib- tential damage that taxes may do by (1) rium level of market employment, and distorting the working of the market hence output (the same would happen if and by (2) reducing output (or employ- the tax were levied on the employee). ment, or investment, or some other tar- An increase in the transfer to the depen- get economrc variable). The distortion dent population (whether on account of arises, in the simple model set out a rise in the replacement rate or a rise in above, from the “wedge” between the the dependency ratio), which raises the cost of labor to the employer (41 + t]) necessary tax rate, leads to a fall in and the opportunity cost to the em- measured output. In this case, we have ployee (h). Distortion would be elimi- a negative levels relationship between nated if t were zero. On the other hand, INS and GDP. this would not maximize market output. It may be convenient to use observed It is of course open to question whether GDP as an aggregate indicator of well- GDP is really the appropriate measure in being, ignoring nonlabor time, but the this context. Along with the reduced la- distinction is important. If it is being ar- bor supply comes increased home pro- gued that the Welfare State is driving

FIGURE 2. Competitive Labor Market and Payroll Tax

market employment

183 people out 01 Ihe market economy, then penditures were contracted. A tax we should be told whether this is unde- concession to encourage private pension sirable because it leads to an inefficient provision may have the sdme conse- allocation of resources or because It re- quences for the public-sector deficit as duces GDP. The numerical measure of the direct payment of pefisions. A switch the cost may be very different (the dis- from state to private protiision would in tortionary loss from a small tax, for ex- this case have no impact. ample, is only second order, whereas the output effect is first order). ‘More interesting in the pttesent context are arguments pointing to specific fea- This example is highly stylized but captures of Welfare State spending that tures, I believe, the kind of relationship have an impact on economic perfor- that people have in mind when consid- mance, as Illustrated by l+ze and Mal- ering the economic burden of the Wel- invaud’s first criticism of the Welfare fare State. At the sarne time, it raises a State that number of issues, in addition to the ob- “(i) measures of income protection or vious one of the quantitatfve magnitude social insurance introduce undesired of the costs. rigidities In the functionling of labour Tax Cost versus Specific Impact markets” (1994, p. 95).

First, the cost in lost output, or reduced We are now concerned with the relative welfare, arises in the model described on desirability Iof different types of govern- account of the existence of taxation. The ment spencling. The quesdion is one of fact that the lax is necessary to finance differential expenditure arjalysis, to use transfers is not, as such, material. The Musgrave’s terminology (Musgrave, Welfare State may represent a partlcu- 1959). larly large item in the budget, but the In order to explore such specific features tax cost is the same dollar for dollar as if of social transfers, we need to elaborate the spending were on overseas aid or the model. Suppose that the size of the defence. dependent population is now influenced It IIS important to distinguish this general by the payment of the transfer. More tax cost argument from arguments that precisely, let us suppose that a fraction are specific to the particular form of of people c’an receive the transfer while spending. Going back to the quotation engaged in home productlion. (As al- from DrPze and Malinvaud at the start ready emphasized, the rules of transfer of this paper, we car1 see that their sec- programs may place obstacles in the ond criticism of Welfare State programs way of such behavior.) As a result, the IS that they supply curve shifts to the left, the level of rnarket output falls further than if “increase the size of government at a there were simply the tax cost. The risk of inefficiency; their funding en- wedge between the value of market hances the amount of revenue to be output and the net benefit to the raised” (1994, p. 95) worker widens for those able to claim while working at home, and there is a (and that the third is that they increase further cost in reduced welfare. public deficits). Cuts in benefits would allow the tax t-ate to be reduced, but As soon, however, as we begin to ana- the same would be true if other forms lyze the speciftc impact of transfer pro- of government expenditure or tax ex- grams, we discover the inadequacy of

184 1 THE WELFARE STATE AND ECONOMIC PERFORMANCE

the economic model for the task since it function is given by equation 3. (This is does not incorporate the contingencies a “right to manage” model where firms toward which transfers are directed. determine employment.) At the same Benefit is indeed paid to people of time, unions look further ahead than the working age but in order to provide for wage; they recognize that there is a sickness, disability, unemployment, and probability, S, that a job will be involun- other contingencies, none of which are tarily terminated. The value of a job, de- modeled. The whole purpose of such noted by a,, takes account therefore of provision is missing from the theoretical the probability that the worker will be- framework. This is related to a second come unemployed. Workers are assumed objection to the theoretical model-that to be risk neutral and to have an infinite it incorporates none of the imperfections horizon (both unsatisfactory assump- that characterize actual economies. The tions) and to discount future income at simple model is a miniature Arrow- an exogenously fixed interest rate, r. In a Debreu general equilibrium system in stationary equilibrium, the expected which the no-government state corre- present value of a job paying wage w is sponds to a first-best situation. The Wel- such that fare State must necessarily have an economic cost since it has only a distributive function to perform. The choice of q ro, = w - s(n, - 0,) model itself precludes the possibility that social transfers may be justified on efficiency grounds. This is a major limitation where 0, is the value placed on the on much welfare economic discussion of state of being unemployed. Equation 5 the redistributive role of the state. shows that the value of a job is atten- uated by the risk of job termination. An imperfect Labor Market If not in market work, the person may Let us now introduce two features that be engaged in home production or may have so far been missing from the story: be unemployed. In order to simplify the unemployment, which provides a ration- analysis, strong (and not necessarily real- ale for social insurance, and trade istic) assumptions are made about the unions, who represent a departure from possible labor market transitions. It is as- the assumption of perfect competition. sumed that recruitment by firms takes The assumed structure is necessarily place only from the stock of unem- highly simplified. Unemployment takes ployed; there is no recruitment of those one of two forms: frictional unemploy- engaged in home production (who are ment resulting from imperfect matching out of the labor force). People may of jobs and vacancies, and wait unem- move out of home production into un- ployment as people queue for jobs at employment, so that the present value the union wage rate. Trade unions have of home production (equal to w,,/r in a stylized objective function, and bar- stationary state for the marginal person) gaining is assumed to take a specific is equal to the value placed on being form. Nevertheless, the model is un- unemployed : doubtedly closer to a real-world labor market.

Unions and employers bargain over the R” = WJf wage rate, w, in the market economy, in the knowledge that the labor demand The value of being unemployed, in the absence of unemployment benefit, is the market, which affects the extent of fric- expectation of being recruited into a tion. market job at the union wage. There is equilibrium wait unemployment. The The wage differential itself is the subject probability of moving from unemploy- of bargaining. Following the standard ment to paid work is equal for all unem- assumption in the labor economics liter- ployed and depends on the number of ature (for example, Booth, 1995, p. vacancies and on the matching of the 125), the outcome is assumed to be the unemployed to vacant jobs, which is as- generalized Nash bargaining solution, sumed to be imperfect so that not all where employers and unions maximize jobs are filled instantaneously. I assume that the matching function, with U un- 81 ernployed and V vacancies, takes the H = Z7” {L(i), - Q,)} special form such that the number of matches is where T denotes profits and 8 is a positive parameter measurings the relative q bargaining power of the employers, and M = m d(W) where the union maximizbs the difference in total expected present value so that the rate of outward flow is from the employment of 1 workers at wage W, compared with their being unemployed. From this, one~obtains the first-order condition (it m+y be verified M/U = m v(V/U> that the second-order conditions are sat- It follows that in stationary equilibrium isfied) the valuation placed on the state of unemployment is m w/(w -. w/J= 1 /p + I9 (1 - /w/3 rf&, = [LJ, - Q,] m v( V/U) (see Booth, 1995, p. 1 25,i and using the Cobb-Douglas production function). This The probability of getting a union job is can be rewritten the only reward at this stage for the unemployed (unemployment benefit is in- m troduced below). w = Wb [I -t (P/(1 - p)>/(l + e)] From the three equations 15, 6, and 9, we can obtain the relationship which so that the negotiated differential is, as must hold in equilibrium between the we might expect, larger, the larger is the wage rate and the marginal value of share of capital (/3> and the smaller is home productlon, wb (eliminating f2, and the relative strength of employers (0). fA,> : In equilibrium, the U/V ra/io must be such that (combining equations 10 and 13) w = wb[l + ((r -t- @/m)v(U/V)]

The necessary wage differential depends m on the degree of pressure in the labor VW/V>=m/k + S)(p/(l - @)I/( 1 + 0) = A

186 I THE WELFARE STATE AND ECONOMIC PERFORMANCE

Since in equilibrium the number of va- Institutional Structure cancies is equal to the number of job terminations, 6L, we can express U as a The standard labor economics textbook proportion of L (8 times the square of treatment of unemployment benefit as- the right-hand side of equation 14). This sumes that it is paid unconditionally, so gives the augmented labor demand that we simply add the benefit, b, to curve, including the queue unemploy- the right-hand side of equation 9 for the ment, shown by L + U in Figure 3.6 valuation placed on the state of unem- There is an equilibrium with employment ployment. If the replacement rate is p, in both market and home production so that the benefit received is pw, then lower than if the labor market cleared equation 9 becomes without friction and there were no union power. The model described above serves to il- r.f2, = [cl, - n,] m A&/U) + pw lustrate how even a relatively limited modification of the assumptions intro- We now obtain duces significant complexity. It is, however, the greater richness of the model that allows us to examine the impact of wll + (dr + ~)lm)~WlV)l social transfers in a way that recognizes their key institutional characteristics. = wi# + ((r+ s)lm4u/v)l

FIGURE 3. Union Bargaining and Wait Unemployment

market employment

187 As one might expect, the existence of is equal to the marginal value of home unemployment benefit makes waiting production. Ftnally, there is the question more attractive. The introduction of ben- as to whether it is 0, or & that enters efit of this form shifts the “total” de- the union objective function. Any short- mand curve including those In the term reduction in labor force may give queue. The equilibrium value of U/V rise to benefit entitlement, but in the rises. Employment in the market sector is long term it is achieved by natural was- reduced, as is home production. tage, and the size of the,insured labor force is scaled down. In \ivhat follows, I However, as has been stressed in the assume that the fallback iposition has previous section, the typical unemploy- value &, as before. rnent insurance benefit does not take the form assumed above. Unemploy- If we write the replacement rate as p”, ment insurance is subject to contribution then solving equations 5’, 6, 9, and 15 conditions, is paid for a limited duration, for a,, L?,, and &, we arrive at and is monitored to limit coverage to in- voluntary job loss. In order to incorporate these institutional features, we need toI distinguish between insured and unin- vvfl + ISp’/(r + mV(V/U) + $1 sured unemployment, the value of these = WJl + ((r + s>/m>tl(u/v>J two states being denoted by LJ, and 0,,. People working in the market sector whose jobs are terminated are assumed Qualitatively, the effect of unemployment insurance works in the same direc- to be entitled to benefit on the basis of tion, making (Insured) unemployment past contributions, so that they enter the state of insured unernployment This more attractive, and raising the equilib- means that the value of a job in the rium U/V ratio. But the qluantitative im- market sector becomes pact is potentially quite different. The key difference is in the extent to which benefits make working in the market sector more attractive, which is given by the square brackets on the left-hand side of equation 10” for unemployment At the same time, those in receipt of insurance and of equation 10’ for the thle insurance benefit face a probability y hypothetical benefit imagined by econo- that the benefit expires. This means that mists. Suppose that the o 1 tflow rate is in stationary equilibrium thie valuation 25 percent per quarter, that the job loss placed on the state of insured unern- rate is five Ipercent and thk interest rate ployment is five percent. A replacement rate of 50 percent then raises the square bracket in equation 10’ from 1, with no benefit, to 1.2; the sarne replacemenlt rate with the insurance scheme, and an outflow rate from insuraince of 20 percent, raises the where it is assumed that the rate of square bracket from 1 to 1.05. The dis- flow out of unemployment is the same incentive effect of the ins rance benefit Y for the insured unemployed as for the is less serious because it is tied to pre- uninsured (again a strong assumption). vious ernployment record.” The fine R,, gives the valuation of the state Iof structure can make a considerable quan- uninsured unemployment, and again this titative difference,, and needs to be

188 THE WELFARE STATE AND ECONOMIC PERFORMANCE

taken into account when specifying rate (there is no home production) and econometric relationships. At an aggre- to be growing over time at rate n. In gate level, account has to be taken not growth rate form, we have just of replacement rates but also of the extent of benefit coverage; in micro- econometric studies, individual benefit gv = PgK + (1 -- PQIA + 4 entitlement and conditions need to be modeled. where gx denotes the proportionate It is not within the scope of this paper growth rate of the variable X. to examine the redistributive benefits of How may the Welfare State affect the the Welfare State, but it should be growth rate? The first possible mecha- noted that the institutional features just nism is via a reduction in savings and considered turn on benefit coverage the rate of capital accumulation.8 How- being less than complete. The contribu- ever, as is well known, in the (Solow) tion conditions associated with unem- neoclassical growth model a reduction in ployment insurance reduce its effective- saving would lower the level of output, ness as a social safety net. At the same but not affect the steady-state rate of time, it is not a simple equity/efficiency growth. The steady-state growth rate at trade-off. For instance, disqualification which output and capital are growing at provisions for job refusal may deter such the same rate is equal to the rate of refusal without anyone actually being population growth plus the rate of tech- disqualified, so that benefit coverage re- nical progress (setting gv = gK in equa- mains complete. tion 16). In the long run (and the speed WELFARE STATE AND ECONOMIC of convergence may be slow), any decline in savings induced by the Welfare GROWTH State does not affect the growth rate. This may be seen by rewriting equation I turn now to the possibility that the 16 as Welfare State may adversely affect the rate of growth of the economy: to provide theoretical justification for the growth-rate hypothesis, in contrast to gv = pwvlKl~> + (1 - p)(gA + d the levels hypothesis of the previous section. The competitive general equilibrium where 5 denotes aggregate savings, as- model used at the beginning of that sumed equal to investment. If S/Y were section may be given a dynamic inter- to fall, then over time the capital output pretation, with a full set of futures mar- ratio falls and in steady state the fall in kets, but this neither coincides with the (K/Y) fully offsets the fall in the savings reality of existing markets nor captures ratio, leaving the growth rate un- the interesting features of a dynamic changed. economy. Here, I start instead from the theory of economic growth, in which If, however, the rate of technical prog- there has been a resurgence of interest ress is treated as endogenous, rather in the past decade. than exogenous, then the transfer system may affect the long-run growth The point of departure is again the ag- rate. Suppose that we take the simple gregate production function equation 1, version of the Arrow (1962) learning by although the labor supply is now as- doing model where productivity A de- sumed to be unaffected by the wage pends on experience, which is propor-

189 tional to cumulated past investment, or If we follow the herd in making this as- K. This gives a production function for sumption, then the impact of social the economy as a whole (the unsatisfac- transfers can only operate via the net tory features of this formulation are rate of return (bearing in mind that the clearly brought out by Solow, 1994): gross rate of return is fixed at ap). The payment of a state pension financed by a payroll tax which does not affect r,, has no impact on desired~ growth rate of Y=aK capital. In this respect, it ~resembles the extreme Kaldorian model ~(Kaldor, 1956), and the economy is in instantaneous where savings are proportionate to capi- steady growth at rate tal income, and the rate of growth of capital is equal to the savings rate times the rate of return. gy = gK = S/K Neither the Ramsey nor the extreme Kal- dorian models seem partikularly appeal- where 5 denotes net savings. A rise in ing as explanations of savings in modern the savings rate leads to a permanently economies. More commonly used in increased rate of growth, with the rate studies of the impact of @ensions have of technical progress being correspond- been models of life-cycle ~savings with a ingly increased. On this steady growth finite lifetime and no bequests (so that path, the private competitive return to there is no Ricardian equivalence). One capital, r, is equal to a&’ such is the discrete time model of Dia- In this endogenous growth model, can mond (1965), where peoqle, identical in social transfers reduce the long-run rate all respects apart from their date of of growth? In particular, does the exis- birth, live for two periods working for a tence of a state pay-as-you-go pension wage w during the first aind living off scheme reduce the growth rate, as com- their savings in the secono.‘” Capital mionly alleged? To consider this, we available to the next generation is equal need to investigate the determinants of to the savings of the preceding genera- saving behavior. Muc:h of recent growth tion of workers. Suppose ~that they theory assumes that this can be modeled choose to c:onsume in the first period a in terms of a representative agent maxi- fraction (1 - U) of their net present dis- mizing the integral of discounted utility counted receipts, which are equal to the over an infinite horizon. This “Ramsey” wage net of payroll tax at rate t plus the formulation requires that the rate of pension received next per od discounted growth of consumption, and hence the by (1 -t- r), since the net I ,eturn is equal steady-state growth rate of caprtal, to the gross return. This may be seen as equals the result of maximizing the Cobb- Douglas utility function m U(c,, CJ = c/l-“) c; where r, is the return to rndividuals net of any taxes, p is the rate of discount, and 1 /E the rate of intertemporal substi- (where 0 < CT < 1) subject to the bud- tution in the utility function. get constraint I THE WELFARE STATE AND ECONOMIC PERFORMANCE

adverse impact on the long-run growth rate. There are, however, a number of Cl + c*/(l + r) = w - tw + tw(1 + g)/(l + r) important considerations that are miss- = w - tw(r - g)/(l + r) ing. As in the previous section, neither the economic model nor the treatment The pension scheme is assumed to be in of the Welfare State is wholly satisfac- steady state with a constant tax rate, so tory. that the pension received per head is the contribution of the current genera- Institutional Structure tion (tw) increased by a factor (1 + g) since the wage bill is higher by this First, we need again to examine the in- amount. As is well known (Aaron, stitutional fine structure, as becomes ap- 1966), the pay-as-you-go scheme makes parent when we consider the alterna- people worse or better off according to tives to the pay-as-you-go state pension whether the rate of interest obtainable analyzed above. Those advocating cuts on private savings is greater or less than in state pensions do not usually propose the rate of growth. It follows that the that nothing take its place. Critics wish capital carried forward is to see either a better targeting of state spending, for example with universal pensions being replaced by income- tested benefits, or state provision being SW’ w(1 - t) - (1 - &I41 - rtr- g)/(l + dl replaced by private pensions. Both of = [CT- t + (1 - &o - g)/(l + r)lw these changes in policy would, however, have economic consequences. Combined with the learning by doing Suppose first that the level of state pen- model used above, this yields a rate of growth [using the fact that w = (1 - sion provided to those with no other resources is left unchanged but that the PWI state benefit is withdrawn progressively from those with other sources of income. The pension ceases to be univer- (1 + g) = s(l - p)a sal and becomes an “assistance pension.” In a limiting case, the state (It may be noted that g appears on the benefit represents a minimum income right-hand side of equation 24 via s.) If guarantee, and is reduced dollar for dol- we were to start from a position where lar of other resources. Such a reform the rate of growth equals the rate of re- promises to reduce total public expendi- turn, then from equation 23 we can see ture while still meeting the antipoverty that the payroll tax would have a pure objective (providing the guarantee is set pay-as-you-go effect, with state contri- at a sufficient level). But the test of re- butions displacing private savings dollar sources changes the inter-temporal bud- for dollar, and hence reduce the rate of get constraint faced by the individual. People who prior to retirement foresee growth. Where the initial rate of growth is less than the rate of return, the effect that increased savings lead to a reduced is smaller, but the savings rate is still state transfers may adjust their savings reduced. behavior. In the case of the minimum income guarantee, they in effect face an We have therefore described a situation either/or choice. Either they save suffi- in which the Welfare State can have an cient to be completely independent in

191 old age or they reduce their savings to pares the highest level of utility obtain- zero and rely solely on the state benefit. able on AB with that obtainable at point 0 consuming the entire het wage in the Such a policy move toward assistance first period and the minimum pension in pensions, while it would reduce total the second. From the utility function LYelfare State spending, creates a “sav- equation 21, we can calculate that the ings trap.” The potential impact may be minimum pension is preferable where seen in the earlier Diamond model. Fig- ure 4 shows the choice now faced by the individual when there is a minimum income guarantee. Suppose that the w< t/(1 -- T)‘w,,h*(l + g)/(l + r) minimum guarantee is set at the level of the previous pay-as-you-go pension, where h is a constant greater than 1. tw,,(l + g): i.e., a proportion t of the average wage, allowing for the fact that In order to understand the implications this rises at rate (1 + g). The switch to of this proposal, we can no longer rely an assistance pension allows the tax rate on the assumption of representative levied on earnings, T, to be less than the identical individuals but have to treat ex- previous value t, since the guarantee is plicitly distributional differences. For peo- paid to only a fraction of pensioners. As ple with wage rates above the critical shown in Figure 4, the opportunity set is Ivalue in equation 25, savings rise Ion now nonconvex, and the consumer com- two counts. First, the tax rate is lower.

FiGURE 4. Budget Constlalnt with Mintmum Pension

consumption in first period

192 I THE WELFARE STATE AND ECONOMIC PERFORMANCE

Second, the contribution is a pure tax, we need to distinguish between the rate so that they reduce present consump- of interest, here denoted i, and the rate tion: the savings rate is reduced not by t of profit, denoted by r as before. but by UT. On the other hand, for those Consideration of the nature of the in- with wage rates below the critical value, vestment function leads naturally to the savings are reduced to zero. Whether or introduction of the corporate sector. As not aggregate savings increase depends suggested in Atkinson (1994), it may be on the number of people above and below the cutoff, their relative wages, and useful to view the investment rate in an endogenous growth model as being the other parameters. The net impact is governed by the choice of growth rate unclear. by firms that face costs of adjustment. Those making private provision for old This draws on the early literature on the age may do so through individual sav- growth of the firm (Penrose, 1959; Mar- ings but in many cases there are special ris, 1964) and follows the work of private pension institutions, and this in- Uzawa (1969) on the Penrose effect and troduces a further institutional feature of Odagiri (1981) on corporate growth. that is often ignored in the theoretical The key element in the growth theory of analysis. In order to qualify (for example the firm is the stock market valuation, V, for reduction in state contributions), pri- which is assumed to equal the present vate provision typically has to be in value of future dividend payments, some protected form, either an occupa- where the discount rate is equal to the tional scheme or one operated by a pen- interest rate i (possibly plus a risk pre- sion institution. Employer-operated mium, although uncertainty is not schemes may affect the financing of the treated explicitly). Assuming that all in- company sector, since the employer is li- vestment is financed out of retained able for any deficit. Pension institutions earnings, dividends are equal to profit acquire substantial weight in the capital less the cost of expansion at rate g, market, and again may influence the given by c(g)K, so that working of the company sector. We cannot simply suppose that a switch to private pension provision would be neutral as far as the capital market is con- v = [rK - c(g)K]/(i- g) cerned. A situation where savings are in the hands of pension funds is different since dividends grow at rate g. from one where they belong to individ- The firm may maximize its stock market ual savers. However, in order to explore value, in which case the desired growth the implications, we need to enrich the rate depends on i and on the internal treatment of the capital market. costs of expansion. Equilibrium of sav- Investment and Firm Behavior ings, which depend also on i, and investment is achieved by variation in the To this point, it has been supposed that rate of interest. In Figure 5 the invest- changes in savings are automatically ment function for a firm maximizing translated into changes in investment. It stock market value is shown by the is assumed that investment can be car- curve labeled I, and the savings rate is ried out of an amount equal to the level assumed to be proportional to the inter- of savings, without consideration of the est rate, generating the equilibrium be- underlying mechanism. As noted by fore any change marked by the dot. Al- Hahn and Matthews (1964, pp. 1 l-l 5) ternatively, in the managerial version,

193 FIGURE 5. Caprtal Market and Effect of Increase in Savings

rate of interest firms maximize the rate of growth sub- move from state to private pensions. The ject to a takeover constralint. The con- first effect is an upward shift in the sav- straint may take the form of limiting the ings function, as analyzed above. This stock markel value to some fraction of tends to raise the equilibrium rate of the “break-up” value of the assets: growth, for both profit-maximizing and growth-maximizing firms, as shown in Figure 5.

VL mK There is, however, a second possible ef- .fect. As already noted, private pension In this case, managers choose the high- .funds come to play a more important est rate of growth consistent with this role in the capital market. In the case of constraint, which yields a different, Sweden, such a development is wel- higher equilibrium rate of growth (and comed by the Lindbeck Commission: interest rate). This is shown in Figure 5 “It is also important tQ stimulate the by the intersection, marked by a cross, emergence of a larger number of insti- of the “5 before” line with the IM curve. tutions that not only hold shares, but are also willing to play Ian active Iown- Capital Mad-ets and the Welfare State ership role” (1994, p. 96). The elaboration of the capital market The precise nature of the takeover con- model allows us to see that impact of a straint, equation 27, has not been

194 I THE WELFARE STATE AND ECONOMIC PERFORMANCE spelled out, but there are good reasons considerable concern about the influence to expect that the larger the fraction of of financial institutions on investment shares owned by pension funds, the decisions. In the United Kingdom, the tighter is likely to be the constraint. (An Goode Committee noted that there had argument may be developed along the been’ ’ lines of the shirking models in the labor market.) If this is the case, then a switch “widespread discussion of the ‘short- in pension from unfunded state to termism’ of pension funds. Those who funded private may lead to a rise in the identified this as a problem saw it as making long-term investment decisions savings rate but a fall in the desired growth rate of managerially controlled in research and development or capital firms. The net effect may be to either projects impossible for company man- raise or lower the rate of growth, or to agements to pursue”(l993, p. 159). leave it the same, as illustrated in Figure They went on to point out that 6, where the IM curve shifts from “IM before” to “IM after.” “Perhaps it did not matter whether the institutions were ‘short-termist’ or not; The existence of such an effect operat- the critical question was whether it ing in the opposite direction from that changed the behaviour of company usually treated may be considered by management to the detriment of the some readers to be mere academic long-term prospects of the economy as theorizing. There has, however, been a consequence of the mere belief that

FIGURE 6. Effect of Move to Private Penslon

rate of interest

195 institutions were likely to behave in this private provision may replace one set of manner” (1993, p. 159). disinc:entives by another. Economists cannot ignore what may appear to be Conclusions Issues of detail.

From this paper, three main conclusions ENDNOTES lrnay be drawn: Paper prepared in memory of Morris Beck I ‘1 Study of the aggregate relationship am grateful to the Editor, Joel Slemrod, and between economic performance and the to I’rat-qois Bourguignon, John Hills, and An- size of the Welfare State is unlikely to ders Klevmarken for their very helpful com- ments on the first version of this essay. I yield conclusive evidence. While popular would also like to thank Steve Dowrick, Mag- argument often refers in a casual way to nus Henrekson, Walter Kqrpi, Dan Landau, the experience of Sweden or other H&an Nordstriim, Torsten Persson, Guido Ta- countries with sizeable levels of spend- bellini and Erich Weede for information about ing, the results of econometric studies the studies quoted in Table 1. None of the above is to be held in any way responsible for are mixed, and provide no overwhelming the opinions expressed in ‘the paper. evidence that high spending on social ’ There is 3 profusion of statistics comparing transfers leads to lower growth rates. social transfer spending in different countries. Nor is it evident that firm conclusions The figures in Figure 1 are from the OECD could be drawn from such an approach, Hislorical Statistics (OECD, 1992, Table 6.3, p. 67) and relate only to social security transfers, which poses serious problems of Inter- excluding other government transfer pay- pretation. ments. They are broadly similar to the figures for “cash benefits” published by the IL0 in * In order to understand the relationship The Cost of Social Security (1992). between the Welfare State and eco- On the other hand, the figures in Figure 1 nomic performance, the theoretical differ from the statistics for income transfers framework needs to be set out explicitly. also produced by the OECD (see, for example, Barr, 1994, Table l-3), which include, in the I have given examples where, to explore case of the United Kingdam, payments under the implications of existing social trans- private occupational pensipn schemes. The fers, and of possible reforms, we need figures in Figure 1 differ allso from those for to enrich the model to Introduce the social protection expenditvre published by Eu- considerations that are central to the rostat (for example, Europiean Commlssion, 1993, p. 42) which includb benefits in kind policy issue. At the same time, the and expenditure on publid health services. rnodels used are far from fully satisfac- The choilze between these different statistics tory and are in need of development. depends on the purpose for which they are Ilnderstanding the impact of the Wel- to be used- a point develioped below. fare State is a challenge to economic ’ See also two reviews in Swedish, which reach theory and not just to applied econome- rather different conclusions from each other: SiSderstrOm et al. (1994) alnd Agell, Lindh, tricians. and Ohlsson (1994). I owe these references to Klevmarken (1994). l l An important role is played by the institutional structure of the Welfare State. 3 Hansson and Henrekson (1994), for example, find a significant negative coefficient on total The form of benefits, and the conditions transfers but a smaller anal less significant under which they may be claimed, can coefficient for social security alone. Since total c:hange their impact on economic behav- transfers include subsidies to firms and inter- ior. The same level of total spending est payments on the natioinal debt, this seems a less rellevant variable for the present pur- rnay have different implic:ations for the pose. level of GDP or the long-run growth rate ‘* Hansson and Henrekson (1994) try also enter- depending on the entitlement structure. ing both level and change in total transfer Switching to “targeted” benefits or to payments.

196 THE WELFARE STATE AND ECONOMIC PERFORMANCE

McCallum and Blais (1987) adjust total social bor Market Transitions: A Critical Review.” Jour- security spending to allow for differences be- nal of Economic Literature 29 (1991): 1679- tween countries in the proportion of popula- 1727. tion aged 65 and over. Barr, Nicholas. Labor Markets and Social Policy Denoting the right-hand side of equation 14 in Central and Eastern Europe. Oxford: Oxford by A, the equilibrium condition is University Press, 1994. Barro, Robert J. “Economic Growth in a Cross L[l + 6h2] = Nqwh) Section of Countries.” Quarter/y Journal of Eco- nomrcs 706 (1991): 407-43. where wh is obtained from equation 13. Beck, Morris. Government Spending. New York: For fuller analysis of these institutional details, Praeger, 1981. in a different model, see Atkinson (1992); for a more general discussion of the actual fea- Blanchard, Olivier. “Debts, Deficits, and Finite tures of schemes, see Atkinson and Mickle- Horizons.” Journal of Political Economy 93 (1985): 223-47. wright (1991). Booth, Alison L. The Economics of the Trade In terms of the earlier discussion of the em- Union. Cambridge: Cambridge University Press, pirical literature, this would show up as an ef- 1995. fect on the total rate of growth, not on factor productivity. Bourguignon, Francois. “Growth, Distribution and Human Resources: A Cross-country Analy- The competitive share of capital is /3, and the sis.” DELTA Document 93-13. 1993. output-capital ratio is a. Castles, Francis G. and Steve Dowrick. “The Labor supply in the first period is assumed to Impact of Government Spending Levels on Me- be fixed. The model may be extended to al- dium-term Economic Growth in the OECD, low for variation in the date of retirement 1960-85.” Journal of Theoretical Politics 2 (which may be affected by the pension)-see (1990): 173-204. Feldstein (1976). Another version has been used by Blanchard (1985), where there is a Diamond, Peter A. “National Debt in a Neo- classical Growth Model.” American Economic Re- constant probability of death and wages de- view 55 (1965): 1126-50. cline exponentially over the lifetime. This has been used by Saint-Paul (1992) to argue that D&e, Jacques H. and Edmond Malinvaud. an unfunded social security system reduces “Growth and Employment: The Scope for a Eu- the growth rate. ropean Initiative.” European Economy No. 1 (1994): 77-106. I was a member of this committee, but did not write this passage! European Commission. Social Protection in fu- rope. Brussels: European Commission, 1993. REFERENCES Feldstein, Martin 5. “Social Security and Sav- ing: The Extended Life Cycle Theory.” American Aaron, Henry. “The Social Insurance Paradox.” Economic Review 66 (1976): 76-86. Canadian Journal of Economics and Political Sci- Gilbert, Neil, and Ailee Moon. “Analyzing ence 32 (1966): 37 l-4. Welfare Effort: An Appraisal of Comparative Agell, Jonas, Thomas Lindh, and Henry Ohls- Methods.” Journal of Policy Analysis and Man- son. “Tillvaxt och Offentlig Sektor.” fkonomisk agement 7 (1988): 326-340. Debatt 22 (1994): 373-85. Goode, Roy. Pension Law Reform. London: Arrow, Kenneth J. “The Economic Implications HMSO, 1993. of Learning by Doing.” Review of Economic Hahn, Frank H. and Robin C. 0. Matthews. Studies 29 (1962): 155-73. “The Theory of Economic Growth.” Economic Atkinson, Anthony 6. “Institutional Features of Journal 74 (1964): 779-902. Unemployment Insurance and the Working of Hansson, Par and Magnus Henrekson. “A the Labor Market.” In Economic Analysis of Mar- New Framework for Testing the Effect of Gov- kets and Games, edited by Partha Dasgupta, ernment Spending on Growth and Productivity.” Douglas Gale, Eric Maskin, and Oliver Hart. Cam- Public Choice 8 7 (1994): 381-40 1. bridge, MA: MIT Press, 1992. ILO. The Cost of Social Security. Geneva: ILO, Atkinson, Anthony B. “What has Happened to 1992. the Macro-economic Theory of Distribution?” Kaldor, Nicholas. “Alternative Theories of Distri- Gaffe Lectures. University of Rome La Sapienza, bution.” Review of Economic Studies 23 (1956): 1994. 83-100. Atkinson, Anthony B. and John Mickle- Klevmarken, Anders. “Economic Astrology or wright. “Unemployment Compensation and La- Empincal Science?” Inaugural Lecture, Depart- ment of Economics, Uppsala IJniversity, Annual OECD. Historical Statistics 196O- 1990. Paris: Report 1994, 11-22. OECD, 1992. Korpi, Walter. “Economic Growth and the Wel- Penrose, Edith T. The Theov of the Growth of fare System: Leaky Bucket or Irrigation System?” the Firm. Oxford: Basil Blackwell, 1959. European Sociological Review 1 (1985): 97-l 18. Persson, Torsten, and Gui o Tabellini. “Is In- Landau, Daniel L. “Government Expenditure equality Harmful for Growth “, ” American Eco- and Economic Growth in the Developed Coun- nomic Review 84 (1994): 601) -2 1. tries: 1952-76.” Public Choice 47 (1985): 459-- Saint-Paul, Gilles. “Fiscal Policy in an Endoge- .77. nous Growth Model.” Quarter/y Journal of Eco- Lindbeck, Assar, Per Molander, Torsten Pers- nomics 107 (1992): 1243-59. son, Olof Petersson, Agnar Sandmo, Birgitta Saunders, Peter. “What Can We Learn from In- Swedenborg, and Niels Thygesen. “Options iernational Comparisons of Public Sector Size for Economic and Political Reform in Sweden.” and Economic Performance?” European Socio- Economic Policy 7 7 (1993): 2 19-64. logical Review 2 (1986): 52-160. Lindbeck, Assar, Per Molander, Torsten Pers- Saunders, Peter and Friedrich Klau. “The Role son, Olof Petersson, Agnar Sandmo, Birgitta of the Public Sector.” OECD ~Economic .Studies Swedenborg, and Niels Thygesen. Turning No. 4 (1985): 6-239. Sweden Around. Cambridge: MIT Press, 1994. Smith, David. “Public Consumption and Eco- Mankiw, N. Gregory, David Romer, and nomic Performance.” Nation@ Westminster Bank David N. Weil. “A Contribution to the Empirics Quarterly Review (November, 1975): 17--30. of Economic Growth.” Quarter/y Journal of Eco- Soderstrom, H. Wf;irds/and i Of;irdst/d. Stock- /JO/-d 107 (1992): 407-37. holm: SNS Ftjrbg, 1994. Marris, Robin. The Economic Theory of “Mana- Solow, Robert M. “Perspectives on Growth gerial” Capitalism. London: Macmillan, ‘I 964. Theory,” Journal of Economit Perspectives 8 McCallum, John and Andre Blais. “Govern- (1994): 45554 ment, Special Interest Groups, and Economic Uzawa, Hirofumi. “Time Preference and the Growth.” Pub//c Choice 54 (1987): 3- 18. Penrose Effect In a Two-class Model of Economic Growth.” Journal of Political,Economy 77 Mitchell, Deborah. income ‘liansfers In Ten (1969): 628-52. We/fare States. Aldershot: Avebury, 1991. Weede, Erich. “Sectoral Re Ilocation, Drstribu- Musgrave, Richard A. The Theory of Public Fi- tional Coalitions and the We I fare State as Deter- nance. New York: McGraw-Hill, 1959. minants of Economic Growth Rates in Industri- Nordstrom, Hlkan. “Studies in Trade Policy and altsed Democracies.” Europedn Journal of Economic Growth.” Monograph No. 20, Stock- Political Research 14 (1986): ~501-19. holm: Institute for International Economrc Stud- Weede, Erich. “The Impact pf State Power on ies, 1992. Economic Growth Rates in OKD Countries.” Odagiri, Hiroyuki. The Theory of Growth In a Qua/Q/ and Quantity 25 (19gl): 421.-38. Corporate Economy. Cambridge: Cambridge Uni- Wilensky, IHarold. The Welfare State and Equal- versity Press, 1981. fry. University of California Press, Berkeley, 1975.