Peter Lindert
University of California – Davis
How Was Life? Volume II
New Perspectives on Well-being and Global Inequality since 1820
4. Social spending and the welfare state
Abstract
The use of social spending to provide safety nets barely existed before 1820. In the next two centuries, it spread around the world. Countries now differ greatly in their commitments to social spending, which continue to take a larger share of national product in richer countries toward the north and west, and lower shares in poorer countries to the south and east. The most striking trend in the make-up of government social spending is the long drift from public investments in the young towards public subsidies to the elderly.
Introduction
Social assistance and social insurance are important indirect correlates of a population’s well-being. Spending on such programmes was tiny or non-existent until political developments assigned more and more of these tasks to government within the last hundred years.
Today’s political-historical landscape receives abstract welfare-economic support from familiar textbook reasons for tax-based social spending. There is a case for social spending based on capital-market imperfections whenever private loans could not have done the job of repairing the insufficiency of insurance, thereby smoothing consumption. Even where loans could have smoothed private consumption, externalities could justify government subsidies to income equalisation and social investments such as public health or public schooling. In addition, paternalism may justify forcing people to save more for old age, through mandatory contributions to their own pensions.
These justifications, while plausible, are static and ahistorical. Their call for social spending does not match the types, timing or geography of the social spending observed in world history. That is, the times and places with the greatest social spending are not those for which the capital imperfections and externalities should have called for it the most. On the other side of the same coin, the world’s failure to engage in any sizeable social spending until the late 18th century clashes with the prevalence of the capital-market imperfections and the externalities that should have brought massive social assistance and social insurance in earlier centuries.
This chapter fills in the history of how government expenditures on social assistance and social insurance were first introduced and then expanded. These programmes seem to be attributes of richer and earlier-developing countries. The use of social spending as safety nets continues to spread. Today at least 164 countries around the world have such programmes, delivering about 10% of world product to targeted beneficiaries as of 2017.1
The evidence reveals a striking trend in what kinds of social supports are offered. Before World War I, social spending, while meagre, focused mainly on providing mass education, a strongly pro-growth policy. Yet in today’s social budgets, the emphasis is largely, probably too largely, on non-contributory transfers to the elderly, implicitly at the expense of investments in the young.
Description of the concepts used
The central measure of social expenditures used in this chapter is the share of gross public social spending in GDP. These expenditures are “gross” in that they do not deduct any taxation, nor any present or past contributions paid by the beneficiaries. They are “public” in the sense that the government collects and distributes the funds, thus excluding private expenditures, whether or not they are mandated or controlled by government or whether or not they are stimulated by governments through tax breaks. That is, this definition is linked to the source of finance, not to the locus of control over how the funds are spent.
While it is sometimes convenient to call social expenditures “transfers”, they are not necessarily transfer payments from one part of society to another. Some of them are funded partly by the prior contributions from the recipients themselves. And some are not strictly transfer payments in the national accounting sense either. Some of these expenditures, most notably health care expenditures, are payments for final services and are, unlike transfer payments, counted in GNP or GDP.2 At the same time, public education expenditures, which should have been conceptually comparable to health expenditures, are excluded from social spending data reported here.
These arbitrary choices were made for practical reasons. To refer to all “social expenditures”, one should have included public education subsidies.3 Yet the OECD’s social expenditure series does not do so. To present a concept matching the OECD’s gold-standard practice, most of this chapter restricts its measures to social spending excluding public education – as in an earlier treatise (Lindert, 2014[1]). Still, although public education expenditures are excluded from the main aggregates presented here, their magnitudes and patterns will be noted at times, to help reveal patterns in political priorities.
Measures of social spending sometimes exclude some budgetary flows serving the same objectives as social assistance and social insurance. They omit most subsidies based on sectoral output, even though such product subsidies sometimes perform safety-net and income-equalising functions. Examples include farm subsidies and price supports, as well as Egypt’s energy subsidies, baladi bread and food ration cards, which are not transfers between groups based on their income or wealth status. Similarly, the social expenditure measures exclude payments of wages under such public works programmes as India’s National Rural Employment Guarantee. As for tax-break “revenue expenditures” such as earned-income tax credits (EITC), some partial progress has been made to include them as social expenditures. While other sources seem to omit these, the OECD’s concept of net public social expenditures now includes revenue expenditures in its updates wherever they are reported, e.g. the EITC for the United States back to 1980, and back to the first positive EITC flows in Israel and Korea.4 These deserve a place in the telling of the history of social spending, and the OECD measures including them are used in this chapter’s data for 1980 and later.
One other exclusion should be made whenever possible – yet most data sources do not allow this exclusion. In principle, our measure of public benefits should include only their “non-contributory” component, i. e. the part transferred from the rest of society. Consider public pensions, a classic social insurance programme. How should the contributory part, paid for by the beneficiaries themselves, be excluded? One way has been to subtract the targeted group’s contributions into public pension funds in the current year from the gross public pension benefits they receive in this same year. Such convenient period measures, unfortunately, compare contributions into future retirement by today’s workers with benefits reflecting, in part, the past contributions by today’s retirees. These same-year measures might work well enough in a steady state but are more problematic in the real world’s growing pension budgets for growing elderly populations. It would be better to use cohort measures, following each birth cohort over its lifetime.
Furthermore, some of the convenient same-year measures often mix the contributory and non-contributory pension revenues together. Where the two kinds of programmes are separated in the underlying data, the Commitment to Equity (CEQ) Institute offers a “mid-range” compromise: call the strictly non-contributory programmes (e.g. universal minimum pensions) “subsidies” and add to this a value equal to half the benefits of the programmes for which we cannot separate the contributory and non-contributory components.
Until the start of this century, the OECD had lumped contributory and non-contributory sources of revenue together into the same aggregate. In recent years, however, the OECD has been able to quantify contributions to pension funds for a growing number of countries, and it is identifying them separately in the SOCX series. See Adema, Fron and Ladaique (2011[2]) and OECD (2019[3]).
This chapter presents measures of social spending by major programme objective – health care, aid to the elderly, labour market subsidies, family/welfare assistance and housing subsidies – for all available benchmark years.
Whenever possible, the chapter’s measures of social spending draw on consolidated accounts for all levels of government – national, provincial and local.
Historical sources and data quality
This chapter’s tables and figures map the same historical geography offered in the other chapters of this book. The chapter covers the 20 decadal benchmarks from 1820 through 2010 for 25 countries. There are, however, some data gaps dictated by underdevelopment and by historical turbulence. Table 4.1 inventories the available direct estimates of social spending. Lack of data is most serious for the intermediate period from 1950 through 1980, when many of these 25 countries with positive social spending did not publish the relevant figures. It seems likely, in any case, that countries with the largest shares of social spending in GDP were those reporting it. Beyond our 25 countries, there was a clear rise in the numbers of countries publishing direct estimates of social expenditures. For the world as a whole, the number of published estimates rose from two countries in 1820 to at least 164 countries reporting in or around 2010 (Table 4.1).5,6 Social safety-net programmes continue to spread around the globe.
For some country/years, more than one source offers data. The international compilations can be ranked by their cross-country and inter-temporal consistency of accounting concepts, even though all of them draw on the same national sources. The source rankings can be arranged separately for the periods before and after 1945. For the latter period, the international-source rankings and “data quality” classes are as follows:
1. The OECD’s Social Expenditure database, annually from 1980 on, receives the top ranking. As noted, the present chapter follows the OECD accounting concepts, based on its editing procedures and its better coverage of public expenditures, especially those by non-central governments. The number of countries covered by the OECD has grown from 21 to 35. Data class = 1 (best).
2. The OECD’s earlier (1985) social expenditure database covered 19 countries from 1960 to 1981, though with less detail and less explanation than are offered for the new series since 1980.7 Data class = 1.
3. Lustig (2017[4]) and Lustig (2018[5]), based on the Commitment to Equity Project, covers social expenditures in 30 low- and middle-income countries for 2008-2013. Data class = 1.
4. The Asian Development Bank (2013[6]) carefully detailed social spending in 32 countries as of 2009. The set of countries covered includes many in Oceania and Asia. Data class = 1.
5. Espuelas (2012[7]) and Espuelas (2013[8]) for Spain and Portugal in all years up to 2000 wove annual estimates from national sources. Data class = 2.
6. Hedberg, Karlsson and Häggqvist (2018[9]) offer annual series for the period 1920-1959 for Australia, Canada, France, Germany, Italy, the Netherlands, Sweden, the United Kingdom and the United States. This compilation draws from Flora (1983[10]) for continental Europe and from national sources for several other countries. Data class = 3.
7. For several other European countries in 1950 and 1960, we draw on Flora (1983[10]), which offers annual estimates, with gaps, up to 1978. Some categories are over-aggregated. Data class = 3.
8. For some non-OECD countries in the period 1990-1996, the ILO attempted to report “social security” expenditures, before yielding the estimation of social expenditures to other agencies, though it has recently shown signs of returning to this task.8 Data class = 3.
9. The World Bank’s “ASPIRE” database (The World Bank, n.d.[11]), concentrating on government spending targeted at poor populations in the 21st century, for developing countries all over the world. This omits social insurance expenditures and the often-large shares of benefits delivered to the non-poor. The low ranking of this series stems from our decision to target the broader social-expenditure concept now used by the OECD. Data class = 3, for present purposes.
10. Guesses based on other expenditure series, such as parts of social spending or over-aggregations such as “social and economic” spending. Data class = 4.
Table 4.1. Data quality of estimates of social expenditures, 1820-2010
Note: Data categories: High quality: the product of official statistical agency (national or international); 2 = Medium quality: the product of economic-historical research using the same sources and methods as applied by official statistical agencies; 3 = Moderate quality: economic historical research, but making use of indirect data and estimates; and 4 = Low quality: estimates based on a range of proxy information. In case of multiple sources, the lowest quality source is given.
Source: Aside from the country-specific sources identified in Annex 4.A, the sources are international data compilations. The preferred international sources are as follows: For 1880-1930, Lindert (1994[12]), Table 1A; For 1940 and 1950, Hedberg, Karlsson and Häggqvist (2018[9]); For 1960 and 1970, OECD (1985[13]); For 1980-2010, OECD SOCX database (n.d.[14]), last accessed on 12 May 2017. The 2010 estimates shown in this chapter do not reflect the revisions included in OECD (2016[15]), Table 5.9. From around 2010, a second preferred (class 1) database is the set of Asian Development Bank (2013[6]), covering the year 2009. For 2016, data for OECD countries are from SOCX, while those for Japan refer to 2015. See Annex 4.A for detailed country sources. The data in OECD (1985[13]) seem to exclude government spending on housing, an exclusion that is sizeable for some countries such as New Zealand and United Kingdom. For the latter country, the older OECD estimates for the period 1950-70 have been replaced with those from Thomson (1996[16]), p. 50, which include housing. No splicing is applied to link the older OECD’s series for the period 1960 to 1981 with the new ones, following Ortiz-Ospina and Roser (2019[17]).
Each country’s official sources are potentially the most reliable, as international compilations depend on these national sources. It may be the case, however, that a government’s official data publications can be less reliable than the OECD or other publications, which have edited the nation’s data for accounting consistency.
Reaching back to the pre-1945 era, the international compilations play out before 1880, and even the national statistical series fade away, as sketched globally in Table 4.1. For this earlier era the international-source preference rankings are as follows:
1. A peculiarity of the social spending measure for the pre-WW1 and interwar eras is that we often know from institutional history that the value must have been zero, either since the later-independent country did not exist yet or because there were no social programmes that would have generated such expenditures: so value = 0 and data class = 1. Care must be taken, of course, not to infer the lack of social spending from the lack of data, as signalled by Table 4.1’s cautionary use of the probably zero symbol (“0?”). Searching through national legislative histories clarifies the dawn of spending when the compilations and natural statistical annuals fail to do so.
2. Lindert (1994[12]) unveiled social spending for 30 countries at the six decadal benchmarks 1880, 1890, … , 1930, starting from ILO studies in the 1930s (International Labour Office, 1933[18]; International Labour Office, 1936[19]), and mining national government sources back to 1880. These are now supplemented by his coverage of poor relief expenditures back to 1820. Data class = 2, except when this study confirmed true zeroes (so that class = 1).
3. Hedberg, Karlsson and Häggqvist (2018[9]) have pushed 18 countries’ annual series on social expenditures, with and without education, back to 1920. For their 13 countries, they drew on Flora (1983[10]), supplemented by Espuelas (2012[7]) and Espuelas (2013[8]) for Spain, a study that extends back to 1850. For five non-European countries (Australia, Canada, Japan, New Zealand and the United States) they used national sources. Data class = 3.
4. Flora (1983[10]) sketched aggregate social expenditures for about 20 European countries back to the earliest statistical annuals in the 19th century, and Flora and Heidenheimer (1981[20]) gave starting-legislation years for a similar set of countries. Data class = 3, again excepting cases in which this study confirmed true zeroes (so that class = 1).
5. For six Latin American countries – Argentina, Chile, Costa Rica, Colombia, Peru and Uruguay – Arroyo, Lindert and Lindert (2017[21]) exploited the available statistical yearbooks back to their earliest publications in the 19th century. Data class = 3 (except = 1 for confirmed zeroes).
6. Van Bavel and Rijpma (2016[22]) have ventured earlier benchmark estimates for England-Wales, Italy and the Netherlands from 1850 all the way back to 1427, weaving together clues from both the revenue and the expenditure sides of government budgets. Data class = 3.
Supplementing these compilations for the pre-war and interwar periods will require applying the same procedures to individual-country sources that their compilations omitted.
Main highlights of social spending trends
Only very recently in the long sweep of world history has social spending, as defined here, absorbed a large share of the economy. So say Table 4.2 and Table 4.3 and Figure 4.1 for our 25 selected economies. If we were to take 20% of GDP as an arbitrary threshold defining the large-budget “welfare state,” then no country had become a welfare state any time in world history before the 1960s. The main ascent took place between the 1930s and 1980s. By 1940 the United States had begun spending over 5% of GDP on New Deal Programmes, and a decade later Britain’s Labour government was spending over 10% on public health, pensions and other social spending. On the 20% definition, the welfare state was ushered in first in Germany and the Netherlands around 1967, with the lead in spending share passing to Belgium in the 1970s, to Sweden in the 1980s and to France in the early 21st century.
While the rise in social spending shares has been accelerating over the long run, there were reversals for some of the top-spending countries. In fact, the 200-year span used in this volume happens to begin with a decline from a local peak in public aid to the poor, in England and Wales around 1820, after which poor relief was slashed, especially by the Poor Law Reform of 1834. Throughout northwest Europe, poor relief, the main form of social spending, remained low.
The late 20th century also saw some national retreats from generous social spending. Sweden’s sweeping economic reforms launched by the Economics Commission’s report (Lindbeck et al., 1994[23]) in 1993 included cutbacks in social spending, and a 1998 pension reform also introduced a systemic restraint on pensions in an aging world. In the mid-1990s, the Netherlands made deep cuts in certain kinds of spending, especially overly generous disability payments. Neither Sweden nor the Netherlands spends as great a share today as in 1990. Nonetheless, the overall international trend has been toward channelling higher shares of GDP into social spending.
Within the overall rise in social spending as a share of GDP, which kinds of spending rose the most, and what are the trends in the composition of social budgets? Of course, if we wanted to know only about the composition of the increase in the entire social budgets over the last 200 years, starting from near-zero spending, then we need to look only at the latest breakdown by type of spending, to see the net changes. Table 4.4 does this, by showing the composition of social spending as of 2010, a composition that remains roughly the same today.
Could the rise in gross public social spending have been offset by a decline in private charity, so that there was no true rise in overall social spending? Historical evidence, and simple arithmetic, reject this suspicion of a full “crowding out”. Back in 1820, northern Europeans and North Americans surely did not privately transfer more than 20% of national product to vulnerable populations, as their welfare states are doing today. The best spotty evidence available for dates between 1500 and 1820 suggests that private and religious charities, aggregated to the national level, always gave less than 1% of national income, with the brief exception of the Netherlands in the mid-18th century. Furthermore, US data show that before 1927 poor and disabled people received little in private transfers, as in public transfers, and that since 1927 both private and public transfers have risen together, contrary to the crowding-out hypothesis.9
Another suspicion that can be set aside is that the dramatic rise in gross social spending might have been cancelled by an equal rise in the “clawback” of taxes paid by the recipients themselves. This may be partially true for social insurance programmes, given the difficulties of sorting out the prior contributions and insurance premiums paid by the current cohort of recipients. Still, the literature quantifying fiscal redistribution finds that most of the gross transfers between income ranks are also net transfers after taxes.10
Table 4.2. Total social spending since 1820: The first 100 years, 1820-1910
Percentage shares of GDP, excluding public education spending
Note: Blank = not a sovereign state at that time (e.g. colonies); “0” = known to have none; “0?“ = probably zero or below 0.1%; and “?” = no estimate yet, though true value > 0.
Source: Aside from the country-specific sources identified in Annex 4.A, the sources are international data compilations. The preferred international sources are as follows: For 1880-1930, Lindert (1994[12]), Table 1A; For 1940 and 1950, Hedberg, Karlsson and Häggqvist (2018[9]); For 1960 and 1970, OECD (1985[13]); For 1980-2010, OECD SOCX database (n.d.[14]), last accessed on 12 May 2017. The 2010 estimates shown in this chapter do not reflect the revisions included in OECD (2016[15]), Table 5.9. From around 2010, a second preferred (class 1) database is the set of Asian Development Bank (2013[6]), covering the year 2009. For 2016, data for OECD countries are from SOCX, while those for Japan refer to 2015. See Annex 4.A for detailed country sources. The data in OECD (1985[13]) seem to exclude government spending on housing, an exclusion that is sizeable for some countries such as New Zealand and United Kingdom. For the latter country, the older OECD estimates for the period 1950-70 have been replaced with those from Thomson (1996[16]), p. 50, which include housing. No splicing is applied to link the older OECD’s series for the period 1960 to 1981 with the new ones, following Ortiz-Ospina and Roser (2019[17]).
Table 4.3. Total social spending since 1820: Developments since 1920
Percentage shares of GDP, excluding public education spending
Source: Aside from the country-specific sources identified in Annex 4.A, the sources are international data compilations. The preferred international sources are as follows: For 1880-1930, Lindert (1994[12]), Table 1A; For 1940 and 1950, Hedberg, Karlsson and Häggqvist (2018[9]); For 1960 and 1970, OECD (1985[13]); For 1980-2010, OECD SOCX database (n.d.[14]), last accessed on 12 May 2017. The 2010 estimates shown in this chapter do not reflect the revisions included in OECD (2016[15]), Table 5.9. From around 2010, a second preferred (class 1) database is the set of Asian Development Bank (2013[6]), covering the year 2009. For 2016, data for OECD countries are from SOCX, while those for Japan refer to 2015. See Annex 4.A for detailed country sources. The data in OECD (1985[13]) seem to exclude government spending on housing, an exclusion that is sizeable for some countries such as New Zealand and United Kingdom. For the latter country, the older OECD estimates for the period 1950-70 have been replaced with those from Thomson (1996[16]), p. 50, which include housing. No splicing is applied to link the older OECD’s series for the period 1960 to 1981 with the new ones, following Ortiz-Ospina and Roser (2019[17]).
Figure 4.1. Social transfer in seven countries, 1820-2016
Within the overall rise in social spending as a share of GDP, which kinds of spending rose the most, and what are the trends in the composition of social budgets? Of course, if we wanted to know only about the composition of the increase in the entire social budgets over the last 200 years, starting from near-zero spending, then we need to look only at the latest breakdown by type of spending, to see the net changes. Table 4.4 does this, by showing the composition of social spending as of 2010, a composition that remains roughly the same today.
Over all recent centuries, the kinds of government social spending that have grown the most as shares of GDP in Europe, Japan and the Americas are pension-and-survivor transfers to the elderly, followed by health subsidies and public education. Means-tested “welfare” payments and family benefits are a very small share of any welfare-state social budget, contrary to their dominance in media discussions.
Table 4.4. Government social spending in 25 countries in 2010, by type
Percentages of GDP
Source: For 2010 data for OECD countries, including Mexico, Poland and Turkey, the source is the OECD’s Social Expenditure series, available through the OECD iLibrary as accessed in 2017. The 2019 version of the same dataset from the same source, used in updating Table 4.2 and Table 4.3, gives some different GDP shares for total social spending from those shown here. As of this writing, the breakdown in Table 4.4 cannot be updated yet from 2017 estimates to 2019 estimates. For the 5 Asian countries, the source is the set of Asian Development Bank (2013[6]) for 2009. For Argentina in 2009, the values refer to all levels of government consolidated; see Lustig (2017[4]) and the time series in Arroyo, Lindert and Lindert (2017[21]). For Brazil in 2009, the source is Higgins and Pereira (2014[24]), Table 1, which “includes spending at the federal, state and municipal levels” (p. 3). For South Africa in 2010-2011, the source is Inchauste and Lustig (2017[25]), Table 8.2, drawing on (Statistics South Africa, n.d.[26]), 2012a. ”Financial Statistics of Consolidated General Government.” Annual statistical release, Stats SA, Pretoria; the title of this publication implies that the numbers refer to consolidated general government, not just central government; contributory pensions are not included, nor are public sector pensions. The “(≥)” for China 2009, Egypt 2010, Kenya 2010 and Nigeria 2012 signifies that the social spending is likely to have been understated by these “All Social Assistance” numbers from the World Bank’s ASPIRE database (The World Bank, n.d.[11]), since they exclude social spending of the “social insurance” variety.
Table 4.5. Government social spending in 13 OECD countries in 1910, by type
Percentages of GDP
Source: For education, Lindert (2004[27]) see App. C, except for Japan, New Zealand and Spain, as noted below. For all other social expenditures, Lindert (1994[12]), Tables 1A-1E and its working-paper appendices. Data on public education expenditures for Japan in 1910 were kindly provided by Yuzuru Kumon; those for New Zealand in 1910 are from New Zealand (1912); those for Spain in 1910 are from Espuelas (2013[8]).
Yet within the last 200 years, there has been an important shift in the composition of social expenditures. To highlight just the largest of several shifts, we can compare the 2010 expenditures, just summarised in Table 4.4, with those of 100 years earlier, in Table 4.5. For a better sense of the timing of these changes, Figure 4.2 adds two benchmarks between 1910 and 2010, the first in 1960 (to highlight the progressive decline in the share of public education) and the second in 1980 (in correspondence to a break in the series). Before World War I, social spending was directed primarily toward education. Most advanced countries had public subsidies for mass schooling. Investing in schooling other people’s children through taxes was a sign of an advanced country. Even Adam Smith and Thomas Jefferson, and later Milton Friedman, approved of subsidising mass primary education. Granted, the amounts of taxpayer money spent on public education were still below 2% of GDP as late as 1910 (Table 4.5). Yet advanced countries clearly invested in the young in the pre-war era. On the other hand, most of them had no unemployment compensation nor housing subsidies, nor even public pensions. Public education led the way.
Figure 4.2. Types of social expenditures, average of 13 OECD countries, 1910-2010
Note: The vertical line indicates a break in series. The 13 countries covered are Australia, Canada, France, Germany, Italy, Japan, the Netherlands, New Zealand, Norway, Spain, Sweden, the United Kingdom and the United States. The weight given to each country is its real gross national expenditures at international prices (rgdpe = real GDP in the Penn World Tables) for the year 1960. The data for public housing before the 1980 in the new series were too close to zero to show up on the figure. “Welfare” is the sum of family assistance, labour-market assistance and public housing subsidies. In the new OECD series for 1980 and 2010, pensions are the sum of all “elderly” benefits plus survivor benefits.
Over the hundred years between 1910 and 2010, and especially in the half-century before 1960, what grew most were public payments to the elderly. On first thought, this might seem to be an automatic by-product of population aging – surely society pays a greater share of GDP to the elderly as their share of the population expands. Yet the shift toward pensions has gone beyond that. Indeed, what have expanded fastest, even faster than public education, have been pension subsidies per person of the targeted age group, as shown in Table 4.6 and Figure 4.3. That is, social expenditures have shifted toward the elderly and (relatively) away from the young or the poor.
Table 4.6. Public pension support ratios and public education support ratios, 2010 and 1910
Source: Support ratios are defined as in Figure 4.3. Public expenditures as percentages of GDP are from Table 4.4 and Table 4.5; the age shares of total population are from (United Nations, Department of Economic and Social Affairs and Population Division, 2015[28]). Also see sources and notes to Table 4.1, Table 4.2 and Table 4.3.
Figure 4.3. Support ratios for public education, pensions and health, average of 13 OECD countries, 1910-2010
Note: The 13 countries included in the OECD average are Australia, Canada, France, Germany, Italy, Japan, the Netherlands, New Zealand, Norway, Spain, Sweden, the United Kingdom and the United States. The weight given to each country is its real gross national expenditures at international prices (rgdpe = real GDP in the Penn World Tables) for the year 1960. The data for public housing before 1980 in the new series were too close to zero to show up on the graph. “Welfare” is the sum of family assistance, labour-market assistance and public housing subsidies. In the new OECD series for 1980 and 2010, pensions are the sum of all “elderly” benefits plus survivor benefits.
There is no obvious explanation for such a shift in the make-up of social spending on either efficiency or paternalistic grounds. As these countries prospered, as poverty declined, and as capital markets and public information improved, individuals should have been increasingly able to save for their own retirement, without government subsidy or compulsion. Note in Table 4.6 and Figure 4.4 those countries in which the shift toward pensions rose most over the last century. The generosity of support for the elderly increased the most in South America, Mediterranean countries and Eastern Europe. The literature on these issues shows that these three regions, and developing countries around the globe, share a telling bias in their policy preferences. Their pension generosity is biased toward those who served in higher-paying occupations in the formal sector, including the civil service and the military. Combining their bias in favour of higher-paid occupations with an overlapping bias against targeting social spending at the young means that transfers have reduced poverty more successfully for the elderly than for children or for those of working age, as the OECD has recently shown.11
Figure 4.4. Public pension support and public education support, 2010 and 1910
Note: Data for 1910, limited to 9 countries, are shown as blue diamonds (see Table 4.6). Public elderly-support ratio is the ratio between public pension and survivorship benefit payments per person aged 65 or over divided by GDP per capita. Public education-support ratio is the ratio between public education spending per person aged 5-19 divided by GDP per capita. No data are available for China, Egypt, Kenya and Nigeria. Regional markers are defined as in Figure 4.7.
The combination of population aging and the rising generosity of annual support per elderly person over the last hundred years implies an ominous rise in the tax effort that younger age groups are paying to support the current elderly. So says this simple accounting identity:
Implied tax effort (transfers to elderly / GDP) = (elderly share of population) times (public elderly-support ratio)
where the public elderly-support ratio is as defined in Table 4.6 and Figure 4.3 and Figure 4.4. To keep the tax effort from rising indefinitely in the presence of a higher elderly share of the population requires an offsetting drop in the support ratio. Over the last hundred years every term in this equation has been rising, which seems unsustainable. What does this portend for the future of pension finance?
Fortunately, in at least some countries, the 100-year rise in the annual-support ratio has been reversed since 1980, without actually reducing the real absolute value of the transfers per elderly person. Figure 4.5 shows that the support ratio has slowly declined since 1987 in Germany and since 1992 in Sweden.12 Additional countries, not shown in Figure 4.5, have succeeded in slowly suppressing the same ratio in the face of population aging; Australia, Canada, the Netherlands and New Zealand have succeeded in doing so since 1980, and Estonia and Poland since 1995. On the other side of the coin, represented in Figure 4.5 by the extreme case of Greece, several countries have increased the generosity of support per elderly person, relative to average earnings. This has been the case since 1980 not only for Greece but also for Finland, France, Israel, Italy and Portugal, and slightly so for Denmark and the United Kingdom. These trends suggest an increasing risk for the latter group of countries, but not for the former.
Figure 4.5. The generosity of elderly support in six countries, 1980-2013
Note: The elderly support ratio is defined as old-age and survivor benefits per person aged 65 or older as a percentage of GDP per person aged 18 to 64.
Source: See Table 4.6 and Figure 4.2.
Social spending under, and since, communism
How does one devise a historical measure for “social spending” during the communist command economy – i.e. Poland 1950-1990, Russia 1920-1990 and China 1950-1970? The institutions were very different from those in the other 22 countries covered in this chapter. Yet, on reflection, one can see that the complexity of the differences does not greatly affect our choice of social spending measures.
Consider, for example, a case in which the government confiscated a private house, and then let a government loyalist occupy it rent-free. This is a stark redistribution, one that does not easily fit the social insurance motive inspiring most social spending programmes. Redistributing that house also does not fit the notion of egalitarian social assistance, unless the recipient happens to be poorer than the previous private owner. Yet on both sides of the institutional chasm, the logical choice of a measure of social spending is the same. The fact that the government is providing the home free of charge means that the rental value of that home contributes the same value to gross social spending as in a market economy. In both settings, the government’s transfers in kind (this house, in our example) or its transfers in cash are gross flows of social spending as long as they go to health, old age, disability, family needs, unemployment or housing.
How large were those expenditures as shares of GDP during the communist era? We now have at least partial benchmarks for the Soviet Union, Poland and China (Adema, Fron and Ladaique, 2014[29]).
Soviet social spending is at least partially illuminated by official numbers for 1930 and by the study (McAuley, 1979[30]) of Soviet living standards in the 1960s and 1970s (Table 4.2, Table 4.3). As of the early 1930s, during Stalin’s First Five-Year Plan, official numbers say that the Soviet Union spent what amounted to 4.31% of estimated GDP on housing and other social programmes, perhaps the second-highest share in the world, behind Weimar Germany’s 4.96%. The Soviet numbers do not, of course, reflect the unmeasured negative transfers to the peasantry during Stalin’s collectivisation campaign. By the 1960s and 1970s, under Khrushchev and Brezhnev, Soviet social spending had risen to 15-17% of GDP, roughly comparable to that in the Western European countries that were about to become welfare states on the 20%-of-GDP criterion. Its mix of social expenditures was also similar to Western European practice, if we follow the usual OECD practice of excluding public education subsidies. However, the Soviets spent a larger share of GDP on public education – 6.7% in 1960 and 7.7% in 1970 – than any OECD country other than the United States, Canada and Sweden. The commitments to social spending and to education retreated somewhat, however, in the post-Soviet republics, including the Russian Federation. Broadly speaking, the shares of GDP they devoted to social spending plus public education settled back to the average practice of Western Europe.
For Poland in the communist era, the available estimates in Table 4.2 and Table 4.3 imply less social support than those in the Soviet Union, and much less than the generous supports offered by the Polish government since 1990. However, these numbers may cover social spending incompletely, as explained in Annex 4.A.
China under Mao (1949-1976) lagged behind the Soviet Union in social spending. In 1960, during the heavily communal living of the Great Leap Forward, it still spent only 5.51% of a slumping GDP, slightly above the 4.31% under Stalin during the first Soviet Five-Year Plan and well below the 1960 Soviet social expenditures (18.38% of GDP) under Khrushchev. During the post-Mao reforms, China expanded its social spending to 6.8-8.5% of GDP, though these fluctuating shares are well below those of post-communist Russia and Eastern Europe.13
Looking at all post-communist or still-officially-communist Eurasian countries as a group since the 1990s, one can see a rough geographic pattern in their government commitments to social spending. Among these countries, social spending tends to rise from east to west, and sometimes also from south to north – just as it does in their never-communist near neighbours. In East Asia in 2009, Vietnam’s central government spent less than 5% of GDP on social programmes, slightly less than in China (5.4% for central government only), whereas Mongolia’s central government spent 9.6% (Asian Development Bank, 2013[6]). Among the central governments of formerly Soviet Central Asia and Transcaucasia, Tajikistan and Armenia spent less (1.2% and 2.2%), whereas the central governments of Kirghizia, Uzbekistan, Azerbaijan and Georgia spent somewhat more (6.1-10.2%), still less than the Russian Federation. Moving further west, Russia’s spending was exceeded by its nearest neighbours to the west, and the highest spending shares in the former Soviet bloc are those of its western-tier states of Poland, the Czech Republic, Hungary and Slovenia, all of them around the arbitrary 20% threshold used in this chapter to define a welfare state.14
Correlation with GDP per capita and related social indicators
The fact that social spending has claimed its highest share of GDP only in recent times means that social spending must have been positively correlated with GDP per capita over time. Is it also positively correlated with GDP per capita across countries in a given year, like some other indicators surveyed in this volume?
Using just the 25 countries that are the focus of this book yields the correlation trends shown in Figure 4.6. The correlation between GDP per capita and the share of social spending in GDP was generally positive, even though errors in GDP measurement should have imparted a negative bias. Both at the start and at the end of our 200-year span, social spending tended to be significantly greater among the richer countries. This is what Figure 4.6 shows us for the early 1820-40 benchmarks and for the years since 1960. In between, however, the positive correlation vanished. This appears to have occurred because of the relative rise in GDP per capita of four rich frontier economies (Argentina, Australia, Canada and the United States) that were low social spenders in that interim period and began to catch up after 1890.
Figure 4.6. The correlation between social spending as a share of GDP and real GDP per capita among several countries, 1820-2010
Today, richer countries still commit a larger share of their resources to social spending, even if we look beyond our set of 25 countries to a larger group of countries supplying good data. Among the 70 countries shown in Figure 4.7, the correlation between the share of social spending in GDP and GDP per capita is 0.77.
In the early 21st century, the share of social spending in GDP is correlated not only with GDP per capita but also with several other societal outcomes with direct relevance to aggregate well-being. We would expect social spending to reduce income inequality, and it clearly does so, according to recent international compilations. Among 53 countries around the year 2013, social spending is highly correlated with the progressive fiscal redistribution of income (0.63), reducing the inequality of final incomes after taxes and transfers enough to reverse the fact that social spending tends to be lower in countries where pre-tax inequality is greater.15 Spending a larger share of GDP on government social programmes is also associated with lower gender pay gaps.16 Furthermore, large welfare states, particularly in northern Europe, have some of the world’s cleanest and least corrupt governments, as surveyed by Transparency International, with lower budget deficits than the United States, Japan and other rich countries. And, for what it is worth, their populations express greater happiness in international surveys of public opinion.
Figure 4.7. Social spending as a share of GDP and GDP per capita in 2010, and paths followed by three leading countries since the 19th century
Priorities for future research
The available evidence invites a wide range of research projects that should help to understand trends in social spending. In particular, one cannot help noticing in Figure 4.7 that the levels of social spending can be much higher in one country than in another with a similar average income. Why is it that across Eurasia, social spending looms larger toward the northwest and much smaller toward the southeast even if we compare countries with about the same level of GDP per capita? The differences in behaviour are also great even among those countries that experienced communism and central planning between World War II and the start of the 1990s. For example, contrast in Figure 4.7 the generous social supports in Hungary with the near-zero supports in Azerbaijan. Why?
References
[66] Adema, W. et al. (1996), “Net Public Social Expenditure”, OECD Labour Market and Social Policy Occasional Papers, No. 19, OECD Publishing, Paris, https://dx.doi.org/10.1787/122001486025.
[29] Adema, W., P. Fron and M. Ladaique (2014), “How much do OECD countries spend on social protection and how redistributive are their tax/benefit systems?”, International Social Security Review, Vol. 67/1.
[2] Adema, W., P. Fron and M. Ladaique (2011), “Is the European Welfare State Really More Expensive?: Indicators on Social Spending, 1980-2012; and a Manual to the OECD Social Expenditure Database (SOCX)”, OECD Social, Employment and Migration Working Papers, No. 124, OECD Publishing, Paris, https://dx.doi.org/10.1787/5kg2d2d4pbf0-en.
[44] Alexander Eckstein, W. (ed.) (1968), Quantitative Trends in the Economy, Social Science Research Council (U.S.), Committee on the Economy of China.
[57] Ansell, B. and D. Samuels (2014), Inequality and Democratization: An Elite-Competition Approach, Cambridge UK: Cambridge University Press.
[21] Arroyo, A., L. Lindert and P. Lindert (2017), “Fiscal Redistribution in Latin America since the Nineteenth Century”, Has American Inequality Changed Direction? Looking over the Long Run.
[6] Asian Development Bank (2013), The Social Protection Index: Assessing Results for Asia and the Pacific, Asian Development Bank, Mandaluyong City, Philippines, https://www.adb.org/sites/default/files/publication/30293/social-protection-index.pdf.
[48] CAPMAS, A. (2012), The Statistical Yearbook. September.
[51] Causa, O. and M. Hermansen (2017), “Income redistribution through taxes and transfers across OECD countries”, OECD Economics Department Working Papers, No. 1453, OECD Publishing, Paris, https://dx.doi.org/10.1787/bc7569c6-en.
[67] Cook, L. (2007), Postcommunist Welfare States: Reform Politics in Russia and Eastern Europe, Cornell University Press, http://www.jstor.org/stable/10.7591/j.ctt7zdbd (accessed on 12 January 2020).
[42] De Oliveira e Silva, P. (2017), “Social policy in Brazil (2004-2014): An overview”, International Policy Centre for Inclusive Growth Working Paper, No. 155, http://ipcig.org/pub/eng/WP155_Social_policy_in_Brazil.pdf.
[32] Dieguiez, H., J. Llach and A. Petrecolla (1990), El Gasto Público Social, Includes central administration, provinces, municipalities, state-owned enterprises, and social security. Ministerio de Economia, Buenos Aires.
[34] Dirección General de Estadística, Argentina (1915), Extracto estadístico de la República Argentina, Compañia Sud-Americana de Billetes de Banco.
[33] Dirección Nacional de Política Económica (n.d.), , http://www.mecon.gov.ar/peconomica/basehome/serie_gasto.html.
[52] Dostal, J. and G. Nakoshi (2017), “The Indonesian welfare state system: with special reference to social security extension in the development context”, in Aspalter, C. (ed.), The Routledge International Handbook to Welfare State Systems, London, New York: Routledge Taylor & Francis Group. 2017.
[49] Egypt Network for Integrated Development (2014), “Subsidies and the social safety net in Egypt”, Policy Brief, No. 012, https://socialprotection.org/discover/publications/subsidies-and-social-safety-net-egypt.
[62] Elveren, A. and T. Agartan (2017), The Turkish Welfare State System: With Special Reference to Human Capital Development, London: Routledge.
[8] Espuelas, S. (2013), “La evolución del gasto social público en España, 1850-2005”, Estudios de Historia Economica, Vol. 63.
[7] Espuelas, S. (2012), “Are dictatorships less redistributive? A comparative analysis of social spending in Europe, 1950-1980”, European Review of Economic History, Vol. 16/2, pp. 211-232.
[40] Ferreres, O. (2005), Dos Siglos de Economía Argentina, 1810-2004: Historia Argentina En Cifras, Editorial El Ateneo, Buenos Aires, Argentina.
[10] Flora, P. (1983), State, Economy and Society in Western Europe, 1815-1975, Frankfurt, Campus Verlag.
[20] Flora, P. and A. Heidenheimer (1981), The Development of Welfare States in Europe and America, London, Transaction Books.
[65] Garfinkel, I., L. Rainwater and T. Smeeding (2010), Wealth and Welfare States. Is America a Laggard or Leader?, Oxford University Press.
[60] Group, G. (ongoing), Global Price and Income History Group: A Research Team Measuring Prices, Incomes, and Economic Wellbeing around the World Before 1950, http://gpih.ucdavis.edu. (accessed on 12 January 2020).
[47] Gu, E. (2001), “Dismantling the Chinese mini-welfare state? Marketization and the politics of institutional transformation, 1979-1999”, Communist and Post-Communist Studies, Vol. 34, pp. 91-111.
[55] Hagen-Zanker, J. and H. Tavakoli (2012), An analysis of fiscal space for social protection in Nigeria, Overseas Development Institute Research Reports and Studies, https://www.odi.org/sites/odi.org.uk/files/odi-assets/publications-opinion-files/7580.pdf (accessed on 12 January 2020).
[9] Hedberg, P., L. Karlsson and H. Häggqvist (2018), How does openness to trade really affect the welfare state? Evidence from social spending in 21 countries, 1920-2000, Conference Paper, XVIII World Economic History Congress, Boston, Massachusetts.
[24] Higgins, S. and C. Pereira (2014), “The Effects of Brazil’s High Taxation and Social Spending on the Distribution of Household Income”, Public Finance Review, Vol. 42/3, pp. 346-367.
[53] IBP Kenya (2015), The Right Priorities? What Kenya’s National Government Spends Money On, https://www.internationalbudget.org/wp-content/uploads/What-Does-the-Kenyan-Government-Spend-Money-On.pdf (accessed on 12 January 2020).
[37] IEERAL, I. (1986), “Estadísticas de la evolución económica de Argentina, 1913-1984”, Estudios, Vol. 9/36.
[41] ILO (2020), Cost of Social Security 1990-96, https://www.ilo.org/secsoc/areas-of-work/statistical-knowledge-base/cost-of-social-security-1990-96/lang--en/index.htm.
[59] IMF (1996, 2006), International Financial Statistics, International Monetary Fund, Washington DC.
[25] Inchauste, G. and N. Lustig (2017), The Distributional Impact of Taxes and Transfers: Evidence from Eight Low- and Middle-Income Countries, Washington: World Bank.
[36] InfoLEG (2014), Presupuestos Nacionales, http://www.infoleg.gob.ar/?page_id=837 (accessed on 12 January 2020).
[19] International Labour Office (1936), International survey of social services, ILO Studies and Reports, Series M, No.13, Kraus Reprint 1975, ILO, Geneva.
[18] International Labour Office (1933), International survey of social services, ILO Studies and Reports, Series M, No.11, Kraus Reprint 1975. ILO, Geneva.
[63] Kenworthy, L. (2019), Social Democratic Capitalism, Oxford University Press.
[58] Kuklo, C., J. Łukasiewicz and C. Leszczyńska (2018), Historia Polski w Liczbach (History of Poland in numbers), Warsaw, https://stat.gov.pl/en/topics/other-studies/other-aggregated-studies/history-of-poland-in-numbers,8,1.html.
[68] Ladaique, M. (2014), Measurement of Social Protection in OECD Countries. Re-thinking Social Protection in a Changing World, ILO Regional Seminar.
[23] Lindbeck, A. et al. (1994), Turning Sweden Around, Cambridge: MIT Press, https://mitpress.mit.edu/books/turning-sweden-around.
[69] Lindert, K., E. Skoufias and J. Shapiro (2006), Redistributing Income to the Poor and the Rich: Public Transfers in Latin America and the Caribbean, The World Bank, Washington DC.
[70] Lindert, P. (2017), “The rise and future of progressive redistribution”, Commitment to Equity (CEQ) Institute Working Paper, No. 73.
[1] Lindert, P. (2014), Private Welfare and the Welfare State, Cambridge University Press.
[27] Lindert, P. (2004), Growing Public. Social Spending and Economic Growth since the Eighteenth Century, Cambridge University Press.
[50] Lindert, P. (1998), “Poor relief before the Welfare State: Britain versus the continent, 1780-1880”, European Review of Economic History, Vol. 2/2, pp. 101-140.
[12] Lindert, P. (1994), “The rise of social spending, 1880-1930”, Explorations in Economic History, Vol. 31/1, pp. 1-36.
[43] Liu, D. and K. Yeh (1965), The Economy of the Chinese Mainland, Princeton: Princeton University Press.
[5] Lustig, N. (2018), Commitment to Equity Handbook. Estimating the Impact of Fiscal Policy on Inequality and Poverty, Brooking Institution Press.
[4] Lustig, N. (2017), “Fiscal policy, income redistribution and poverty reduction in low and middle income countries”, Commitment to Equity Institute Working Paper, No. 54.
[31] Macon, J. (1985), Las Finanzas Públicas Argentinas, Includes central administration, provinces, and municipalities, Ediciones Macchi, Buenos Aires.
[30] McAuley, A. (1979), Economic Welfare in the Soviet Union: Poverty, Living Standards, and Inequality, University of Wisconsin Press, Madison.
[71] Mesa-Lago, C. (2017), “The Cuban welfare system: With special reference to universalism”, in The Routledge International Handbook to Welfare Systems, London, Routledge.
[38] Ministerio de Economía y Producción, A. (2006), Presupuesto de la Administración Nacional: Gastos por Finalidad – Función y Naturaleza Económica, 1965-2006, Ministerio de Economía y Producción.
[3] OECD (2019), The OECD SOCX Manual, 2019 Edition. A guide to the OECD Social Expenditure Database, https://www.oecd.org/social/soc/SOCX_Manuel_2019.pdf (accessed on 12 January 2020).
[15] OECD (2016), Society at a Glance 2016: OECD Social Indicators, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264261488-en.
[64] OECD (2008), Growing Unequal? Income Distribution and Poverty in OECD Countries.
[73] OECD (2001), 1980-1998: 20 Years of Social Expenditure - The OECD Database, OECD, Paris.
[13] OECD (1985), “Social expenditure 1960-1990. Problems of growth and control”, OECD Social Policy Studies, OECD, Paris, http://www.oecd.org/social/soc/40836112.pdf (accessed on 12 January 2020).
[14] OECD (n.d.), Social Expenditure Database (SOCX), Covers social spending for the expanded OECD, 1980-2015, https://www.oecd.org/social/expenditure.htm.
[39] Oficina Nacional del Presupuesto, A. (2014), Series estadísticas, Administración Nacional por Finalidad-Función y Naturaleza Económica, 2007 … 2013, http://www.mecon.gov.ar/onp/html/#.
[17] Ortiz-Ospina, E. and M. Roser (2019), Government Spending, Published online at OurWorldInData.org, https://ourworldindata.org/government-spending.
[26] Statistics South Africa (n.d.), Stats SA: Improving lives through data ecosystems, http://www.statssa.gov.za/?s=database.
[54] The Global Economy.com (n.d.), Kenya: Government spending, percent of GDP, https://www.theglobaleconomy.com/Kenya/Government_size/ (accessed on 12 January 2020).
[11] The World Bank (n.d.), ASPIRE Database, http://datatopics.worldbank.org/aspire/indicator/social-expenditure.
[16] Thomson, D. (1996), Selfish Generations? How Welfare States Grow Old, White Horse Press, Cambridge, UK.
[56] Umukoro, N. (2013), “Poverty and Social Protection in Nigeria”, Journal of Developing Societies, Vol. 29/3 (September), pp. 305-322.
[61] Union of Soviet Socialist Republics (1936), U.S.S.R. Handbook 1936, Victor Gollancz Ltd, London.
[28] United Nations, Department of Economic and Social Affairs and Population Division (2015), World Population Prospects: The 2015 Revision.
[22] Van Bavel, B. and A. Rijpma (2016), “How important were formalized charity and social spending before the rise of the Welfare state? A long-run analysis of the selected Western European cases, 1400-1850”, Economic History Review, Vol. 69/1, pp. 159-187.
[35] Vasquez-Presedo, V. (1971), Estadísticas Históricas Argentinas (comparadas), Primera Parte 1875-1914, Ediciones Macchi, Buenos Aires.
[72] Wang, C. and K. Caminada (2012), “The redistributive effect of social transfer programmes and taxes: A decomposition across countries”, International Social Security Review, Vol. 65/3, pp. 27-48.
[45] Wang, Y. and A. Murie (1999), Housing Policy and Practice in China, St. Martin’s Press, New York.
[46] Zhang, X. (1998), Privatisation: A Study of Housing Policy in Urban China, Nova Science Publishers, New York.
Annex 4.A. Data sources for individual countries
Countries covered in the chapter
Argentina
The estimate shown for 2010 refers to 2009. Sources for social spending: (a) 1950-1970: Macon (1985[31]), (b) 1970-1979: Dieguiez, Llach and Petrecolla (1990[32]), (c) 1980-2009: Dirección Nacional de Política Económica (n.d.[33]). Data refer to spending by central administration, provinces, municipalities, state-owned enterprises and social security. The "social assistance transfers" are larger after 1980 as the new source includes labour transfers. Other secondary sources used are as follows: for social expenditure, 1900-1915: Dirección General de Estadística (1915[34]); 1929-1939: Vasquez-Presedo (1971[35]); 1940-1962: Based on national budgets, InfoLeg, (2014[36]) and IEERAL (1986[37]); 1965-2006: Ministerio de Economía y Producción (2006[38]) and Oficina Nacional del Presupuesto (2014[39]); for nominal GDP: 1900-1932: Ferreres (2005[40]); 1932-2013: Ministerio de Economía y Producción (2006[38]) various years. Comparison with the CEQ Institute estimates for Argentina in 2012 suggests that these numbers include some contributory components to pensions. The difficulty is, again, to extract the contributions and derive the non-contributory component. For 2012, the CEQ compromise ("mid-range") estimates suggest a share of social spending of only 17.0% of GDP, versus the 20.5% yielded by the official aggregates that include all benefits paid from past contributions. Still, for comparability with the earlier Argentine benchmark totals, the larger estimate is used here.
Australia
Available federal data show positive social spending as early as 1910, even though Australia was officially a dominion of the British Empire (1901-1939), and not officially independent until 1986. Public pensions began in 1908. Housing subsidies were minimal before 1919. Health and welfare had very small coverage around 1910.
Brazil
The estimate shown for 1990 refers to 1994; the ILO (2020[41]) refers to 2002 (De Oliveira e Silva, 2017[42]), Table 3, which is also the source for 2010.
Canada
For the period before Dominion status in 1867, there clearly was no independent national budget, hence the values of o shown in this chapter. However, provinces could have had their own social programmes.
China
For 1933 (here “1930”), 1952 (“1950) and 1957 (“1960”), we have measures of “communal services” expenditure shares, plus the public share of estimated expenditures on housing services. The shares are defined in current prices for 1933 and 1952, with indirect evidence on price changes between 1952 and 1957. The main sources are Liu and Yeh (1965[43]), p. 68 and Liu (1968[44]), p. 138. Indirect evidence suggests that the shares of all housing service expenditures paid by government were 1% for 1933, a very rough 10% for 1952 and 61% for 1957. On these shares, see Wang and Murie (1999[45]), pp. 46-99, esp. p. 81 and Zhang (1998[46]), pp. 18-42. For 1980-1996 (“2000”), see Gu (2001[47]), p. 94, which includes welfare spending and public housing for both danwei (workplace) and larger government units. The public housing expenditures for 1980 were extrapolated backward to 10 billion yuan from numbers given for 1981 and beyond. Estimates for the year 2012 (here shown as “2010”) are from OECD (2016[15]), Table 5.9, citing "Asian Development Bank’s Social Protection Index (SPI), World Health Organization (WHO)”. Alternatively, the Asian Development Bank (2013[6]) gives a 2009 share of 5.4% of GDP, for central government only.
Egypt
For 1820-1870, Egypt seems to have had no tax-based social spending, like the rest of the Ottoman Empire. For 1880-1910, Egypt was under receivership and its finances were run by Britain and France, with no known social spending. For circa 2010, an official publication (CAPMAS, 2012[48]), pp. 413, 451, gives the “value of pensions & compensations for pensioners & their beneficiaries by reward system in the social insurance sector" as a percentage (8.4%) of GDP. As noted in the main text, most of Egypt's social subsidy programmes, such as energy subsidies, baladi bread and food ration cards, are nearly universal and not transfers between groups based on their income or wealth status (Egypt Network for Integrated Development, 2014[49]). They are thus excluded by the OECD concept measured in this chapter.
France
Estimates for 1830 refer to 1833, those for 1840 to 1841, and those for 1870 to 1871 (Lindert, 1998[50]), p. 113, while those for 1940 refer to 1938. The Flora (1983[10]) estimates for 1940 and 1950, taken from Hedberg, Karlsson and Häggqvist (2018[9]), tend to be higher than those used here for other years (higher than Lindert series in 1930 and higher than OECD in 1959 versus 1960).
Germany
Estimates are from our preferred international sources: Lindert (1994[12]) for 1880-1930, (Hedberg, Karlsson and Häggqvist (2018[9]) for 1950, OECD (1985[13]) for 1960 and 1970 and OECD (2017[51]) for 1980-2016.
India
The estimate for 1990 refers to April 1991-March 1992 (ILO, 2020[41]). For 2010 and latest, they refer to 2012 (2016[15]), Table 5.9, citing “Asian Development Bank's Social Protection Index (SPI), World Health Organization (WHO)”.
Indonesia
The zeroes for 1950 and 1960 are confirmed by Dostal and Nakoshi (2017[52]), p. 368. The estimate for 1990 refers to April 1991-March 1992, derived from the receipts side of social programmes (ILO, 2020[41]). 2010 is from OECD (2016[15]), Table 5.9, citing "ILO.”
Japan
The 1950 estimate is from Hedberg, Karlsson and Häggqvist (2018[9]) citing Statistics Japan. The estimate for 2016 refers to 2015.
Italy
For 1790, before the period covered here, van Bavel and Rijpma (2016[22]) give 1.1% of GDP as the social spending of “formal institutions”, most of which were not tax-funded. The similarly based 1870 estimate refers to 1868, from van Bavel and Rijpma (2016[22]), p. 171. The alternative estimates from Flora (1983[10]) for 1940 and 1950, sourced from Hedberg, Karlsson and Häggqvist (2018[9]) tend to be higher than those reported in this chapter for other dates (higher than Lindert series in 1930 and higher than OECD in 1959 versus 1960).
Kenya
The estimate for 1990 refers to 1994 (ILO, 2020[41]). For 2013/14, expenditures on Security = 16.1% of government expenditures and Health = 4.0% (IBP Kenya, 2015[53]). This presumably includes benefits paid out to civil servants. In turn, government spending = 14.0% of GDP for 2012-2013 (The Global Economy.com, n.d.[54]). So Security plus Health = 2.81% of GDP for 2012-2013.
Netherlands
The estimate for 1820 is for poor relief, from van Bavel and Rijpma (2016[22]), p. 171. Estimates for 1850-1870 are from Lindert (1998[50]), p. 113. Those for 1880-1930 are from Lindert (1994[12]), and those for 1950 from Hedberg, Karlsson and Häggqvist (2018[9]).
Nigeria
The estimate for 1990 refers to 1991 (ILO, 2020[41]). For 2010, the estimate of 0.6% of GDP excludes schemes for civil servants (Hagen-Zanker and Tavakoli, 2012[55]), p. 9. The same study notes major data caveats: "This study is based mostly on budget data, not actual spending data, with the exception of data on total government expenditure… A related problem is that corruption and leakage are very high. Leakages are hard to quantify". Hagen-Zanker and Tavakoli (2012[55]), p. 7. The bleak picture of corruption and programme non-delivery is confirmed by Umukoro (2013[56]).
Poland
The 1930 estimate from Lindert (1994[12]) is used here. Ansell and Samuels (2014[57]) obtained 0.16% of GDP for 1922, and 0.65% of GDP for 1930. For 1951-1990, social spending measures are available in Kuklo, Łukasiewicz and Leszczyńska (2018[58]). I am indebted to Mikolaj Malinowsky and to Cecylia Leszczyńska for providing these source materials. However, this source lacks nominal (current-price) GDP estimates. I have found nominal GDP numbers only for the 1970 and 1980 benchmarks, in (International Monetary Fund (1996, 2006[59]). For 1970 and 1980, the available estimates may omit parts of social spending. For both years, we have only government expenditures on education (omitted here), health protection and social care (ochrona zdrowia i opieka społeczna), financing (subsidies) and social security (finanse i ubezpieczenia społeczne). For 1970, estimates identify a small amount of housing subsidy (gospodarka mieszkaniowa oraz niematerialne), while those for 1980 give communal services (gospodarka komunalna and usługi komunalne). No coverage of subsidies for disability, family aid and labour-market subsidies is made explicit.
South Africa
The 1930 estimate is from Ansell and Samuels (2014[57]), with underlying data kindly supplied by David Samuels. For 2010, the CEQ "mid-range" estimate was used. A lower number (8.7%) would obtain if we had used year 2012 per OECD (2016[15]), Table 5.9, citing "National Budget 2014, Estimates of National Expenditure, National Treasury and World Health Organization (WHO)".
Spain
Note that estimates cover all kinds of social spending, even for the 19th century – unlike for other European countries before 1880, that cover only poor relief. The source is Espuelas (2013[8]) and Espuelas (2012[7]). Espuelas’s work was used even up through the 1970 benchmark, since OECD (1985[13]) did not cover Spain (or Portugal).
Russia
For 1870-1910, Stephen Nafziger has kindly supplied budget numbers for the Imperial (central) government. Local governments probably allocated only negligible amounts to social spending, though possibly noticeable amounts to local public education. The small amounts shown here were rather regressive expenditures on bits of health care for the Army, Navy and a few other ministries with seemingly elite mandates. The nominal nations income denominator is available annually for 1885-1913 at Global Price and Income History Group (ongoing[60]). The Imperial government also spent a bit on education within the Army, Navy and four other ministries. These amounted to 0.47% of GDP in 1890 and 0.42% in 1900. For 1900, Ansell and Samuels (2014[57]), data provided by David Samuels, came up with the same share by different means. The 1930 estimates refer to the Soviet Union in 1931, and are derived from the 1936 Union of Soviet Socialist Republics (1936[61]). Categorical percentage shares of GDP for 1931 were welfare 1.19, pensions 1.56, health care 1.03 and housing 0.53. The housing estimate may be for expenditures building housing, and thus a capital-account entry, as opposed to an imputed rental value of housing provided or subsidised by government in 1931. For 1960 and 1970, estimates refer to social consumption expenditures (obshchestvennye fondy potrebleniya) of the entire Soviet Union, minus education and holiday pay, divided by McAuley’s estimate of Soviet GDP (McAuley, 1979[30]), p. 262. Dividing by Mitchell’s higher estimate of Soviet nominal GDP would have lowered these shares by about 30%. For 2010, our estimates use the CEQ Institute's "mid-range” measure, including half the benefits from partly contributory programmes plus all the benefits from strictly non-contributory programmes.
Sweden
For 1940 and 1950, the source is Flora (1983[10]) via Hedberg, Karlsson and Häggqvist (2018[9]). For the earlier 1930 benchmark, this source gives 3.5% instead of the 2.5% shown in this chapter, based on Lindert (1994[12]).
Thailand
The estimate for 1990 refers to 1993, and is derived from the receipts side of social programmes (ILO, 2020[41]). The estimate for 2010 refers to 2011, from the Asian Development Bank (2013[6]), Technical Report 44152 (2012, p. 27).
Turkey
Estimates for 1980-2010 use the OECD, our preferred source. The zeroes before 1950 seem quite secure. Elveren and Agartan (2017[62]), esp. p. 318, imply that even the first privileged social insurance for public sector employees was not legislated until 1946. The estimate for “2016” refers to 2013.
United Kingdom
Estimates refer to England-Wales through 1870, and to the United Kingdom thereafter. Estimates for England and Wales in 1820-1870 are for poor relief only, from Lindert (1998[50]), p. 114. Those for 1880-1930 are for all social spending, from Lindert (1994[12]). The social-transfer estimates for 1940 and 1950 are from Hedberg, Karlsson and Häggqvist (2018[9]), borrowing from Flora (1983[10]); those from 1960 on are from OECD (1985[13]) for 1960-70 numbers, and from OECD (n.d.[14]) for 1980-2016.
United States
For 1850-1870, see the sources cited in Lindert (2004[27]), Table 3.4. For 1880-1910, see Lindert (1994[12]), Table 2. For 1940 and 1950, see Hedberg, Karlsson and Häggqvist (2018[9]).
Notes
← 1. The 10% value lies between two averages computed over the 25 countries that are the focus of this book. Weighting by population, as in other chapters of this volume, yields a value of 9.2% for 2010. On the other hand, weighting social transfer shares of GDP (correctly) by each country’s nominal GDP implies that transfers took 15.8% of the 25-country GDP.
← 2. Differently from National Accounts, another type of productive service is excluded from the social expenditure measure used in this chapter. The OECD series on social expenditures, which is used as a reference in this chapter, exclude administrative costs – see page 13 of OECD (2001[73]).
← 3. Garfinkel, Rainwater and Smeeding (2010[65]) rightly argue for inclusion of public education expenditures. This author’s joining the OECD in excluding them was originally a debating tactic, as explained in Chapter 1 of Lindert (2004[27]).
← 4. See Adema and Pearson (1996[66]), Adema, Fron and Ladaique (2014[29]) and OECD (2019[3]).
← 5. Of the 164 countries yielding data on social spending in 2010, 91 gave estimates of the desired aggregate expenditures, whereas for the other 73 countries we have only the World Bank’s ASPIRE (The World Bank, n.d.[11]) estimates of the more progressive part of social expenditures defined as “social assistance”.
← 6. Presumably the data collections underlying the IMF’s series Government Finance Statistics would include details on total social expenditures for all countries. Their data on the Internet, however, fail to provide useful breakdowns.
← 7. The discrepancies between (1) and (2) for the years 1980 and 1981 are sometimes large and are not fully explained. The 1998 OECD manual (OECD, 2001[73]) notes that the new series are 2-3% lower than the old series due to a shift in the OECD accounting for GDP. This falls short of explaining why the new series tends to be 12% below the old for the median country covered in 1980.
← 8. See the PowerPoint tables Ladaique (2014[68]).
← 9. For the best available estimates on Italy, Netherlands and England before 1800, see van Bavel and Rijpma (2016[22]), esp. Table 1. See also Lindert (2004[27]), Chapter 3 and Lindert (2014[1]).
← 10. See the whole set of developing-country studies by the Commitment to Equity (CEQ) Institute (Lustig, 2017[4]; Lustig, 2018[5]), plus Wang and Caminada (2012[72]), Lindert (2017[70]) and OECD (2019[3]). See also Adema, Fron and Ladaique (2014[29]).
← 11. The geographical patterns of the regressive bias and the bias in favour of the elderly overlap but are not the same. The two occur together in Latin America (Lindert, Skoufias and Shapiro, 2006[69]; Arroyo, Lindert and Lindert, 2017[21]; Lustig, 2017[4]; Lustig, 2018[5]). The OECD study featured here (OECD, 2008[64]) (see Figure 5.12) also showed that poverty has been reduced more for the elderly than for children and those of working age, confirming an anti-young bias in policy reduction through government taxes and transfers. It did not cover Latin America or Eastern Europe, but did get this same result for Portugal and Italy, along with the United States, Japan and a few other countries.
← 12. Note that Figure 4.3 defines the support ratio as relative to GDP per working-age population, whereas Table 4.6 and Figure 4.2 defined it as relative to GDP per capita. The patterns are similar, however.
← 13. Again, see Table 4.2 and Table 4.3 and the sources cited there. Public education expenditures, still excluded here, are also lower in China than in Russia and Eastern Europe after communism. From about 1990 to about 2010, China’s public education expenditure drifted from 2.5% of GDP down to 1.9%, while the public-education share varied between 2.9 and 4.3% in Russia, and between 4.4 and 5.5% in Poland, according to UNESCO. China’s spending on higher education seems to have expanded considerably since 2010, but we lack reliable estimates for this last decade.
← 14. Cook (2007[67]), Table 4.3, Lustig (2018[5]) and Asian Development Bank (2013[6]). Extending still further west in the set of countries that was communist before 1990, Cuba has at times recorded the highest shares of GDP devoted to social spending. Even if one excludes public education spending, its social-spending share reached over 22% of GDP in 2008. Its public-education shares were also perhaps the highest in the world at 11-14% of GDP (Mesa-Lago, 2017[71]). These ratios speak not only to the achievements of Cuban public health and education, but also to the depressed state of the country’s GDP.
← 15. Lindert (2017[70]). Similarly, social spending and overall GDP per capita tend to be greater in countries where the income shares going to the top 1% or to the top 10% tend to be lower (Kenworthy, 2019[63]).
← 16. So say the underlying numbers behind the negative correlation between GDP per capita and the gender pay gap in Kenworthy (2019[63]) on page 19. Kenworthy (2019[63]) on page 55 also finds that patented innovations tend to be highly correlated, across countries, with his index of social-democratic capitalism, one component of which is social spending as a share of GDP.