The Comfortable, the Rich, and the Super-Rich. What Really Happened to Top British Incomes During the First Half of the Twentieth Century?

We examine shifts in British income inequality and their causes between 1911 and 1949. Newly re-discovered Inland Revenue 1911 estimates and more detailed data from subsequent official income distribution enquiries are used to show that income was substantially more concentrated at the top of the income distribution in 1911 than previous estimates suggest, and that the top 1 per cent were the principal “losers�? in the subsequent trend towards reduced income inequality. We find that this trend reflected a sharp decline in top “unearned�? incomes - paralleling the findings of Piketty and Saez for France and the USA. This explains the paradox between the observed reduction in British income inequality and the lack of evidence for any substantial redistribution of income between salary and wage-earners.

Research on long-term changes in British income inequality has been hampered by very limited data compared, for example, to the USA or France (Pikkety, 2003;Pikkety and Saez, 2003;2013). The 1937/38 Inland Revenue Incomes Census was Britain's first published official income distribution estimate, with the exceptions of estimates for 1918/19 and 1919/20 (that are generally rejected as atypical, given the inflationary conditions around the end of the First World War). This paper focuses on changes in the incomes of the top 0.001 -5 percent of the British income distribution. Top incomes are important largely because income redistributions in western countries are typically dominated by changes in the shares of this group, especially within the top percentile (Piketty and Saez, 2006, pp. 201-2). Changes in top income shares have also been identified as key potential drivers of income inequality reduction in Britain during the first half of the twentieth century, given the very limited changes in inequality among wage/salary-earners (Lindert, 2000, p. 169;Gazeley et. al. 2017;Townsend, 1979, p. 139;Routh, 1965, pp. 51-108). Thus examining higher incomes is crucial to explaining the apparent paradox between a relatively stagnant income distribution among the bulk of the British population and the generally-assumed trend towards a more equal pre-tax income distribution (Lindert, 2000, p. 169).
This study is based on an extensive survey of Inland Revenue (hereafter IR) files held at The National Archives, Kew. We identified an unpublished IR survey of the distribution of personal incomes over £160 for 1911 and data that provide more detailed disaggregation for top incomes in 1937. These sources both enable us to examine changes in top incomes for three key bench-mark years (1911, 1937, and 1949) for the first time, and to explore the relative contributions of earned and unearned incomes to the redistribution.
We first focus in detail on the 1911 income distribution estimate and its methodologyas this data was not published, appears to have been confidential, and (to the best of our knowledge) has not been identified by any previous studies. We then outline subsequent British income distribution estimates for the first half of the twentieth century. The 1911The , 1937The and 1949 distributions (the only ones in this period that disaggregated personal income between "earned" and "unearned" components) are used to examine the contribution of falling capital income -that dominated the declining income shares of the rich -to the overall decrease in income inequality. Finally, we explore the factors behind the fall in top capital incomes. Our findings are in line with other recent studies, that the redistribution was driven primarily by shocks, policy responses, and non-market mechanisms, rather than technological change. We also find that declining capital and land factor incomes directly benefited lower-income groups (for example through lowering house prices and rents).

A RE-DISCOVERED SET OF ESTIMATES
The long-term decline in British inequality is often dated from just prior to the First World War (with a possible slight decline from 1867-1911) (Lindert, 2000, pp. 174-185).
However, there are no published official classifications of the income distribution before 1918.
IR data for the nineteenth century only measured income categories, known as 'Schedules' (property -Schedule A; profits from farming land -Schedule B; interest and dividends -Schedule C; incomes from trade or business, professions, and some miscellaneous items -Schedule D; and salaries and wages -Schedule E).
The 1906 and 1910 Liberal government tax system reforms introduced differentiated (according to income source) and graduated taxation, together with tax allowances for dependent relatives and certain other expenses. These, together with the super tax, provide much better information on personal incomes for tax-payers; especially for people at the lowerend of the income-tax scale and for the top end (super tax payers) (Daunton, 2001, pp. 361 & 367). Income tax data relate to tax years, starting in April. However, given lags between the receipt and reporting of incomes for tax purposes, almost all the income recorded was for the calendar year when the tax year began or, for some kinds of income (such as Schedule D), even earlier (Atkinson, 2007, pp. 128-134). Therefore, we will refer to income tax data as covering the year in which the tax year of assessment began. IR personal tax estimates typically show taxable income -net of any charges on that income (such as loan interest or ground rent) and excluding depreciation; part of government transfers; the investment income of life assurance and superannuation funds, plus not for profit bodies; employers' and most employees' contributions to national insurance and private pensions (though income from all retirement pensions was included); most income in kind; part of the imputed rent of owner-occupied houses; and interest on National Savings Certificates (Lydall, 1959, pp. 28-29;UK, Inland Revenue, 1946, pp. 28-29).
The first official income distribution estimate was an unpublished exercise by the IR for 1911, for use, "in making confidential estimates, especially in connection with any legislation... Estimates affecting particular ranges of income can only be satisfactory when it is possible to see how they fit in with all other incomes dealt with." 1 Unlike later official estimates, the 1911 estimation was kept confidential. The reasons for this are not discussed in the surviving records, but probably reflected the extreme political sensitivity of Britain's high concentration of income and wealth, in the wake of the new land taxes introduced in Lloyd George's 1909 "People's Budget" and the political storm and constitutional crisis this created (Offer, 1981, pp. 317-400).
The 1911 estimate covered people above the income tax threshold (£160 per annum).
The estimation was based on the income tax and super-tax returns (and, for unearned income, estate duty, settlement estate duty and probate data) with a series of adjustments to take account of estimated incomes that fell outside the tax data. Total taxable income comprised £322,531,000 of earned income and £543,923,000 of unearned income -from which was deducted an estimated £1,000,000 of unearned income for people below the tax threshold and £65,454,000 of "impersonal income" for companies and similar bodies. Total personal incomes over £160 thus amounted to approximately £866,454,000 minus £66,454,000, i.e. £800,000,000. 2 The IR regarded their income estimates between £160 and £700 to be "based on sufficiently accurate income tax figures to be beyond question", as they were calculated using income tax liabilities net of tax abatements. 3 However, classifying incomes between £700 and £5,000 was acknowledged to be more problematic, as this could only be done by taking a curve between these two points. When this was done for earned and unearned income the curves seemed implausible, as the unearned line sloped gradually, while the earned line dropped sharply. 4 Moreover, the unearned income line cast doubt on the accuracy of estimated incomes over £5,000, derived from the super tax data. Starting from a total unearned income of £525 million, it was noted that the income distribution should broadly correspond to the capital disclosed by each income group. In addition to around £280 million of capital declared for estate duty, there was an estimated £80 million not declared, as probate and settlement estate duty had previously paid in respect of it. 5 The original "red line" estimate was based on the assumption that this hidden capital was distributed by income in the same proportion as declared capital. However, as officials noted, "In reality… the proportion of Settled Capital is higher in the larger estates, where great blocks of land etc., pass under settlement." 6 Further support for this correction was drawn from the fact that, while the line of total income ought to gradually approach the "unearned" line, the unearned line actually crossed the red total income 2 Ibid. 3 Ibid. 4 Ibid. 5 Annual data for both of these duties were collated by the IR and published in their annual reports. Emphasis in original. 6 TNA, IR64/28, "Income tax. Classification of taxable income -year 1911-12," n.d., c. 1914. line at £12,000, "thus giving the result that the unearned hypothetical income left after this point exceeds the corresponding total income declared for Super-Tax." 7 Thus a corrected "blue-line" estimate for unearned incomes was calculated, which was extended to all incomes over £700.
Incomes exceeding £5,000 were also adjusted by the deduction of life assurance premium tax allowances. Total insurance premium income allowed to income tax payers was £11,882,213, of which £1,500,000 was attributed to taxpayers with total incomes exceeding £5,000. Of this sum, £200,000 was estimated to apply to incomes of £5-6,000, some of which were exempted from super-tax by the deduction of insurance premiums. Assuming the true income of the £5-6,000 group to be around £25,000,000, a "liberal estimate" of £13,500,000 was taken as exempted from super-tax by the deduction of insurance premiums. The published super tax incomes were therefore adjusted as follows: (1) Published total £145,000,000 (2) Insurance premiums to incomes over £6,000 £1,300,000 (3) Insurance premiums to incomes of £5,000-6,000, used to secure exemption from super-tax £200,000 (4) Income exempted from super-tax under (3) £13,500,000 (5) Total super-tax adjusted to income-tax basis £160,000,000 Earned income for the over £5,000 group (£49,231,000) was estimated by subtracting earned income for income tax-payers below this threshold from total earned income; though direct estimates from the tax schedules produced a similar figure (around £50,000,000 and settlement estate duty had previously been paid in respect of it -concentrated among the larger estates. This revised "blue ink line" estimation raised incomes over £5,000 from being represented by log 8.204 = £160,000,000 to log 8.333 = £215,300,000. 9 This revision may still have under-estimated the income share of the very rich, as settlement estate duty -on which the revised calculations were made, was said to be commonly avoided (Mandler, 1997, p. 174). Table 1 shows the IR's corrected "blue line" 1911 income distribution estimate, together with the total UK population of "tax units" -either a single adult (or a single minor with income in his or her own right), or a married man and wife, together with their dependents -including those who did not pay any income tax (Lydall, 1959, p. 6). We follow the approach of the World Top Incomes Database and earlier studies in using tax data for top incomes and national accounts data for aggregate personal incomes (Piketty and Saez, 2013, pp. 457-8). In the following tables we use Tony Atkinson's annual estimates for both the tax unit and personal income totals. This produces plausible figures for average incomes per tax-unit equivalent for the residual income range below the minimum classified income in all cases. Atkinson's tax unit data are based on males and females aged 15 or over, minus married females (and ignoring minors under 15 with income). 10 His total incomes data are based on national accounting data on relevant personal incomes. As Atkinson acknowledges, the early figures in particular are subject to a significant margin of error, though we do not consider that we could improve on them (Atkinson, 2005, p. 331). Table 2 shows an alternative 1911 income distribution estimate, also mainly drawing on IR data, developed by Lindert and Williamson (1983), with subsequent corrections by Lindert (2000), which builds on earlier work by Arthur Bowley,Josiah Stamp,and Guy Routh 9  (hereafter BSR estimate). These data are based on households rather than tax units, though for incomes over £160 the two sources give very similar numbers of incomes. The two estimates show major differences for income classes over £700, reflecting the IR's reallocation of income from the £700-5,000 range to classes over £5,000. This raised the over £5,000 group's share of personal incomes over £160 from 17.8 per cent in the BSR estimate to 26.9 per cent. However, aggregate income for tax units with over £700 is roughly similar for the two estimates.

INCOME ESTIMATES FOR 1918-1949
The First World War is believed to have significantly reduced income inequality; including both a redistribution from the upper and middle-classes to the working-class and from skilled to less-skilled manual workers (Routh, 1965, p. 104). IR estimates of the income distribution for tax-payers in 1918 and 1919 were produced for evidence in two official enquiries. These covered just under a quarter and just under a third of all tax unit equivalents respectively, compared to only 4.9 per cent in the 1911 classification. Table 3 shows the income distribution for 1918 and 1919, together with the tax levied at each income band (after allowances etc.).
The data were acknowledged to be imperfect, especially given the inadequate information available for estimating non-personal income and income accruing to residents abroad -collectively estimated at £230,000,000 for 1918 and £260,000,000 in 1919. 11 As profits under Schedule D were then assessed based on the average over the previous three years, the 1918 and 1919 data also partially reflect the very high profit rates of the War years and the wider inflationary environment (Lydall, 1959, p.2). Like the 1911 estimate, the data representing the super-tax income bands show substantially larger numbers of individual incomes than the super-tax data, suggesting that the figures were adjusted to take account of settled estates and similar distortions (UK, Inland Revenue, 1920, p. 85).
[ Table 3 near here] No similar data are available until the late 1930s, as changes introduced in the 1920 Finance Act made income distribution estimates using IR tax data impracticable. The only usable data for intervening years cover the super-tax income brackets. 12 Figure 1 shows Atkinson's estimates of the income shares of the top 0.05 and 0.01 per cent of the income distribution from 1911-1949, based on super-tax data (except for 1918, 1919, and 1937, and 1943-49, where he used the IR income distributions). Despite the 1918 and 1919 estimates typically being dismissed as irrelevant owing to the inflationary and turbulent conditions of these years (e.g. Bowley, 1942, p. 113), Atkinson's data suggest that they were not atypical of the longer-term super-tax/surtax income trend.
[ Figure 1 near here] The next IR income distribution estimate, for 1937, was based on a special investigation of all tax returns for incomes of £200 or more (16.53 per cent of all tax units); the first of a series of what came to be known as "Surveys of Personal Incomes" (hereafter "SPI's") (UK, Inland Revenue, 1946, pp. 28-29). Published data from the investigation provides very limited disaggregation for incomes over £20,000 -with only three income classes, the highest covering incomes over £50,000. Fortunately, the final working sheets from the survey have survived, enabling us to replace the three highest income classes with seven income classes, the highest of which covers £100,000 or more, in Further estimates were made for 1938, 1941, 1947, and 1948, based on the 1937 SPI, up-dated by the annual statistics of assessments and other data (UK, Parliamentary Debates, 1942;UK, Inland Revenue,1949, p. 34;UK, Inland Revenue, 1950, pp. 83-87). In 1949 a new SPI was conducted, based on a 10 per cent sample survey of all income taxpayers (Lydall, 1959, p. 3). The IR found that the 1949 SPI had important discrepancies when compared with other evidence. There was a considerable deficiency in income from interest and dividends taxed at source (mainly affecting income ranges below £2,000); plus an apparent omission, as compared with National Insurance statistics, of over a million women in paid employment. The IR (1952, pp. 96 & 117) produced a corrected distribution, to include these incomes. Further revisions were made when the 1949 data were published in the 1954 "Blue Book" (Britain's main annual national accounting publication). These appear to involve an adjustment raising the aggregate value of real property (Schedule A) income, to take account of the average rise in rents since the last revaluation in 1935/36 (Lydall, 1959, p. 26).
We have compiled a composite series, using the official "Blue Book" figures for incomes from £250-£20,000, together with incomes over £20,000 from the original 1949 SPI.
No disaggregated data for incomes over £20,000 were available in any tabulation other than the SPI, though total numbers of incomes, and their amounts, for this range change very little between the different estimates. Data for incomes of £135-150 and 150-250 are from the corrected (Table 110)  [ Table 5 near here]

UNEARNED COMPONENTS
We present the five 1911-49 British income distribution estimates based on direct data, (rather than adjustments to previous years' estimates) in Table 6. For three of these, 1911, 1937, and 1949, the data are disaggregated into earned and unearned components, enabling us to explore the relative importance of capital and labour income in the declining incomes of the rich. Our analysis is restricted to the top five percent of the population, as the 1911 survey does not classify lower incomes (which were not then subject to income tax). However, we are still We considered two alternative approaches for estimating income shares. The first is the standard method in the literature, of using the Pareto distribution (e.g. Atkinson and Piketty, 2010), and the second is to assume a linear approximation between the tabulated intervals. The Pareto approach has been found to be an accurate approximation towards the upper end of the income distribution (particularly for the top 1 to 0.1 percent), but performs poorly for levels above the 0.1 percent (Stamp, 1914, pp. 200-204;Feenberg and Poterba, 1992, pp. 172-73).  Tables 1, 4, and 5), we adopt this second method.
[ Table 6 near here]  [ Tables 7 and 8 near here] Table 7 shows ratios of earned to total incomes by income range for our three benchmark years, and Table 8

WHAT CAUSED THE FALL IN TOP INCOME SHARES?
Disaggregating the earned and unearned components of top incomes is important, as nineteenth century Britain's extreme income inequality compared to other developed countries was believed to be driven primarily by inequality in wealth and, therefore, investment income (Lindert, 1991, pp. 220-24). More generally, as Piketty and Saez (2006, p .200) have noted, decomposing incomes into earned and unearned components enables analysis of the economic mechanisms underpinning changes in the distribution of labour, and capital, incomes, which can be very different. Lindert (1991, p. 225)  Wealth inequality is both a major factor in current income inequality and a generator of longer-term income inequality and social immobility -by differentiating families' abilities to make substantial investments in their own, or their children's, education and training (often a critical factor for entry into higher professions such as law and medicine) (Campion, 1939, p. 118). Inequality of incomes flowing from wealth typically exceeds wealth inequality, as the upper ranks of the wealth distribution achieve higher yields on their capital -owing to higher returns for larger holdings in the same asset class (for example bank accounts); lower proportional transactions costs; greater possibilities for diversification to achieve higher yields at any given level of total portfolio risk; and a weaker preference for liquidity (Daniels and Campion, 1936, pp. 60-62;Atkinson and Harrison, 1978, p. 173;Lydall and Tipping, 1961, p. 95).
Our three benchmark years shows a marked decline in the contribution of unearned, to total, income. The drivers of this process appear to be broadly similar to those identified for  (Mandler, 1997, p. 228). Some studies argue that land-owners shrewdly disposed of land in the early post-Armistice period, using the proceeds to diversify their asset base, and/or shift into safer securities, thereby maintaining their nominal wealth (Howkins, 2003, p. 58;Mandler, 1997, pp. 242-3;Rothery, 2007). There was also downward pressure on interest and dividends, which dominated unearned, and total, top incomes by 1937 (see Figure 2). Prior to 1914 Britain devoted a higher proportion of savings to capital export than any other major country has ever done (Matthews, Feinstein, and Odling-Smee, 1982, p. 353 (Kynaston, 1991, p. 143).

[Figures 2 near here]
There had also been periodic pressures to divest of securities on unattractive terms.
During the First World War the Treasury sought to acquire dollar securities and sell them in New York. Patriotic appeals were followed by a penal tax on their dividends/interest in the 1916 Budget, while from January 1917 the Treasury had powers to requisition securities for selling (Morgan, 1952, pp. 326-331). British overseas investments, valued at almost £4,000 million on the eve of the First World War, are estimated to have declined by something in the region of 15 -25 percent owing to these measures (Feinstein, 1990;Hardach, 1977, p. 289-90). Then in the 1930s "cheap money" policy led to a boom in conversion issues, replacing high-yielding government and corporate securities with lower-interest ones, led by the June 1932 conversion of the 5 per cent 1917 War Loan stock to 3.5 percent undated stock (Kynaston, 2000, pp. 365-368 The dominance of negative capital income shocks as drivers of income reduction for the rich is also consistent with Atkinson's annual super-tax estimates of the top 0.1 and 0.05 per cent income shares for 1911-1949 in Figure 1. In addition to a general downward trend, the two World Wars and the 1920-21 and 1929-32 recessions stand out as periods of particularly rapid declines in top income shares (Atkinson, 2005, pp. 335-6).
To some extent the decline in unearned top incomes can be directly linked to improvements in incomes and living standards for the bottom 90 percent of the population. For example, rent control, introduced in 1915, depressed the incomes of landlords, but substantially reduced the real value of a major household expenditure burden, in a country where around 90 percent of households were private tenants (Merrett, 1982, p.1). Rent control also subsequently led to extensive sales of house property portfolios, mainly to sitting tenants, at prices reflecting 18 Rutterford, et al., 2011 pp. 179-80; data read from graph.  (Speight, 2000, pp. 39-40). Meanwhile, the scarcity of low-risk, higher yielding assets during the inter-war years led to substantial deposits in building societies (mutual savings and loan institutions for house purchase) by high-income individuals, facilitating an increase in building society mortgage debt from £120 million in 1924 to £636 million in 1937. This underpinned the owner-occupier house-building boom of the 1930swhich produced Britain's greatest recorded proportional housing stock increase, together with the lowest recorded ratios of weekly house mortgage costs to average incomes (Scott, 2014, pp. 107-8). These conditions also made it easier for local authorities to raise loans for a series of inter-war social housing programmes, cumulatively creating around 1.3 million new homes (Speight, 2000, chapters 4-5;Scott, 2013, pp. 98-127). Meanwhile restrictions on overseas new issues led the City of London to become increasingly involved in British industrial financeexpanding industrial growth and employment -despite protests from City-insiders that domestic industrial issues involved more work and less profit than the foreign loan stock that merchant banks had hitherto focused on (Kynaston, 2000, pp. 131-137 & 295).
While capital incomes dominate the top income decline, our estimates also show a substantial decline in top earned incomes. This is more surprising, especially given that the ratio of earnings for professional and managerial occupations, compared to all workers, remained relatively stable between 1913/14 and 1935/36 (Routh, 1965, p. 107). This reflects a decline in Schedule D incomes -which were classed as earned income, but included a substantial element of profits. Schedule D mainly covers profits from businesses and professions (including employers' salaries). While self-employment incomes are commonly, but not universally, categorised as earned income in national personal income series (Bengtsson and Waldenstrom, 2018, p. 720), this schedule also includes returns on capital invested in unincorporated businesses by proprietors and partners, together with some items of pure investment income -for example colonial and foreign securities (other than government securities) and interest on War securities not taxed at source (Atkinson, 2007, p. 161;UK, Inland Revenue, 1912, pp. 111-113;idem, 1939, p. 56).
In 1911 Schedule D accounted for 61.8 percent of all taxable income. However, its contribution fell to 42.6 percent in 1929;31.9 percent in 1937, and 26.2 percent in 1949(UK Inland Revenue, 1920idem, 1940, p. 56;idem, 1953, p. 42). This is probably mainly due to the growth of incorporation, which shifted profits from personal earned incomes into the investment incomes of shareholders (or retained corporate incomes). By the 1940s tax avoidance was probably also significant in reducing Schedule D incomes, as employers and partners could easily find ways to, for example, convert income into capital gains (which were not subject to income tax in Britain). Given that Schedule D incorporated a substantial element of "profit", that might more properly be regarded as unearned income, the underlying collapse in top unearned incomes is thus even greater than the data suggest.

CONCLUSIONS
Our re-discovered 1911 estimation reveals a markedly more unequal distribution of income than previous estimates, driven primarily by extreme inequality in unearned income.
This is in line with research showing that Britain's peculiarly sharp pre-1914 income inequality reflected its extreme inequality of wealth (Lindert, 1991, pp. 220-224 (Piketty, 2003(Piketty, , p. 1037Piketty and Saez, 2003, pp. 8-9). The following decades witnessed a sharp reduction in British factor incomes for rent, interest, and dividends, which substantially reduced unearned income inequality -a trend which dominated the reduction in overall income inequality. This explains the paradox between the observed reduction in income inequality and the lack of evidence for any substantial redistribution of earnings between salary and wage-earners. However, despite having closed much of the relative gap with America, British incomes remained more unequal than for the USA and France even in 1949.
To some extent Britain's income inequality reduction represented a genuine redistribution from the rich to lower income groups (even prior to income tax and fiscal transfers). The reduction in factor incomes from rent, interest and dividends provided greater scope for higher factor incomes for wages and salaries, while lower income families benefited directly from controlled rents and, to some extent, from lower interest rates and greater credit availability for house-purchases. However, the data also reflects an increase in the extent of tax avoidance and evasion, incentivised by a ten-fold increase in the top rate of income tax between 1911 and 1949, either directly, or by companies (for example by retaining profits to benefit their shareholders in the long-term, rather than incurring heavy taxes on their dividends).
While Britain started from a position of markedly greater income inequality than most developed nations, its overall trend towards reduced inequality, and the underlying causes, appear to be broadly similar. Research on France, the USA, and Japan has found that reductions in income inequality during the first half of the twentieth century were also driven by severe shocks to the capital holdings of the wealthy, including depressions, bankruptcies, war-time inflation, declining real asset prices, and the fiscal shocks of war finance (Piketty, 2003, pp. 1011-1019; Piketty and Saez, 2006, p. 203;idem, 2003, p. 12;idem, 2013,p. 474). In common with Piketty and Saez's findings for the USA and France (Pikkety, 2003(Pikkety, , p. 1011Piketty and Saez, 2013, pp. 461-2;idem, 2003, pp. 3-11 & 33-35), Britain's income redistribution appears to be driven more by political shocks and policy responses, together with non-market mechanisms such as labour market institutions, rather than technological change.
This in turn, raises the question why political shocks, policy responses, and non-market mechanisms increased income inequality during 1911-49, but appear to have acted to concentrate the income shares of the very rich since the 1970s. A partial explanation may be found in theories first advanced in the 1950s and 1960s by historians such as Stanislaw Andrzejewski, Richard Titmuss, and Philip Abrams, that the levelling tendency of wars is proportional to the extent to which low-status groups and classes become essential to the war effort -leading to policy responses and institutional changes that might persist well beyond the war period. 21 Such impacts would be reinforced by tax increases, which reduced even "pretax" personal incomes; for example higher corporation taxes reduced dividends and incentivised firms to retain profits. However, another important factor concerns the changing sources of top incomes. In both Britain and the USA the contribution of capital incomes to top incomes has declined substantially since the 1970s, in favour of salary and entrepreneurial incomes (Piketty and Saez, 2003, p. 17). Thus Marx's (1954, 585-589) prediction that shocks lead to the concentration of capital would imply that top entrepreneurs and executives might benefit from them, while rentiers, receiving incomes from more diversified portfolios of stocks, would be more likely to suffer from their aggregate negative economic impact.
Another related factor governing the impact of shocks on income distribution, under different institutional environments, concerns the ability of nation states to tax rich individuals, or the factor incomes they receive. The 1911-49 inequality reduction was driven, at least in part, by the progressive collapse of the liberal, globalised, world order, which made it more difficult for the rich to seek out more attractive overseas outlets for their investments and made policies such as capital controls more politically expedient and acceptable. Similarly, the policy liberalizations of the 1980s that heralded the start of the new globalisation (and the resumption of growing income inequality in western nations) have made it far easier for the rich to offshore their assets, or themselves, either in search of better investments opportunities, or jurisdictions more suited to protecting their wealth.      , 1911, , 1918, , 1919, , 1937, AND 1949 TABLE 7  UNEARNED INCOME AS A PROPORTION OF TOTAL INCOME, BY INCOME  CLASS , 1911, , 1937, , AND 1949 1911 1913 1915 1917 1919 1921 1923 1925 1927 1929 1931 1933 1935 1937 1939 1941 1943 1945 1947 1949 Top 0.1% Top 0.05% THE DISTRIBUTION OF PERSONAL INCOME, BY TAX SCHEDULE, FOR  DIFFERENT LEVELS OF TOTAL INCOME,  1937 (% OF TOTAL)