1 Introduction

Recent writings now offer us a clearer view of trends in economic inequality within the now-industrialized countries since the early 1300s. A new consensus suspects that the intra-national inequalities of income and wealth never declined significantly across these centuries, except in the wake of the Black Death and during the turbulence of world wars and depression between 1910 and 1970. Throughout Europe, inequalities probably widened between 1400 and 1800.Footnote 1 Across the long nineteenth century, by contrast, the literature sees mixed trends, apart from a clear widening of income and wealth gaps in France and the USA.Footnote 2 For the twentieth and twenty-first centuries, the greatest achievement has been the unlocking of the mystery of top nominal incomes in more than two dozen countries since 1900.Footnote 3

This emerging consensus, like the estimates that produced it, has looked only at the inequality of nominal income or wealth.Footnote 4 This gives the correct inequality result if we seek to know how unequal are people’s abilities to buy any one bundle of goods and services. Yet to compare an income class’s real purchasing power across generations or across countries, it may help to consider the cost of the kind of consumption bundle that class finds relevant. As such, a measure of “real” inequality—one that would measure inequality based on more than a single basket—could act as an important complement to the existing literature. Indeed, it is quite probable that the adverse effects of inequality are linked through real income effects, where the contrasts that matter are contrasts in individuals’ abilities to buy what they care to buy, or need to buy, and not the (nominal) inequality in their ability to buy the same common bundle as some other class could buy. For example, the literature showing that inequality can worsen aggregate health and life expectancy is based on mechanisms of nutritional status that relate to effects on income-class-specific real consumption, especially for those in the lower-income ranks.Footnote 5

If one sought only to contrast the real purchasing powers of poor people across time and space, setting aside the gaps between rich and poor, then one could contrast the abilities of poor populations to buy either a “bare-bones” (subsistence) or a “respectability” bundle, as introduced by Robert Allen. An expanding literature has done so, contributing greatly to the debates over Eurasia’s Great Divergence, Europe’s Little Divergence, and the trans-Atlantic gaps in welfare ratios. For this purpose, real wages and incomes have been mapped and compared around the world (Allen 2001, Allen et al. 2011; Pomeranz 2000, 2011; Arroyo Abad et al. 2012, Geloso 2016 and 2019a). Even the basic income estimates of Angus Maddison are now being reshaped by the Maddison Project (Bolt and Van Zanden 2014) and by new direct comparisons of countries’ purchasing power in the past (Ward and Devereux 2003, 2004, 2006; Lindert and Williamson 2016; Lindert 2016; Panza and Williamson 2017a, b; and Geloso 2019b). Allen himself has continued to pioneer in subsistence comparisons, improving on international agencies’ concepts of the poverty line (Allen 2017).

Knowing how real inequality differed between countries or continents, and how it moved over time, requires that we go beyond the emerging consensus about the real welfare of the poor alone, and try to map what the rich consumed, and what prices they paid for it. In so doing, it will be possible to enrich existing estimates of inequality based on nominal incomes. On this more daunting task we have made enough initial progress to support a new historical geography of real income inequality within countries, combining the expenditure patterns of rich and poor with prices from different continents over more than 200 years. More specifically, we focus on England, Canada, the USA, and Australia from 1688 to 1913.

We find that the relative costs of different lifestyles differed both over time and across continents. First, the movements over time showed pronounced, and reversing, trends. In the early modern setting, here meaning the period up to about 1815, the relative rise of staples prices, especially food grain prices, imposed a rising relative burden on those in the lower-income ranks, relative to those affluent enough to enjoy the declining relative cost of luxuries. In this early modern setting, the documented rises in nominal-income inequality were reinforced by the rise in the relative cost of buying bare necessities, which raised real inequality even more. Thus, up to about 1815, the recent literature’s findings of rising inequality based on nominal estimates have been confirmed, although Hoffman et al. (2002, 2005) found that real inequalities widened even more in that early modern era.

Within the nineteenth century, however, this paper reveals new evidence of a major trend reversal. An egalitarian movement from 1815 to 1913 was shared by all four countries on which this paper concentrates, but with some differences in timing. In Canada, the ratio of the class costs of living remained trendless 1815–1850, followed by a pronounced egalitarian rise in the relative cost of living between 1880 and 1913. In England, the egalitarian drift was spread across the entire prewar century between Waterloo and the First World War. For the USA, the rise in the relative cost of higher living began even earlier, around 1790, i.e., immediately after Independence, and persisted to 1913. Australia showed the same egalitarian trend from 1870 onward.Footnote 6

The two main reasons for this widespread nineteenth-century reversal toward a more egalitarian trend in the cost of living were (1) the decline in the relative prices of staple grains, and (2) the rise in the relative price of services and labor-intensive luxuries, which outweighed the continuing decline in the relative prices of many luxury goods. The strong effect of service prices is confirmed by relative-price movements not only in England and North America, but also in Australia.

Second, across continents, the cost-of-living contrasts reveal that Canada, the USA, and Australia continued to be better lands for poorer households than for richer ones, relative to the corresponding lifestyle costs in France and Britain.

Our progress toward these tentative findings begins with the examination of different income ranks’ preferred consumption bundles, in Part II. Part III assembles our comparisons of the price-driven relative costs of living well versus barely getting by, and Part IV shows which differences in commodity sub-group prices seem to account for these contrasts. Part V uses an accounting exercise to show how the new trends in real top-income shares differ from the now-well-known trends in nominal top-income shares.

2 Consumption bundles for rich and poor

Different income ranks consume different budget shares, even within a given price environment, so that the same prices or the same movement in prices may affect them differently.Footnote 7 Thus, there is a need for more disaggregated measures of the cost of living (Pollak 1980, 1981). That need is especially strong for the pre-1914 era. Prior to 1914, trade was limited to a few (luxury) items, the prices of many goods responded to local market conditions (Williamson 2011), and purchasing power parities differed importantly over space. We already know that the rich and poor consumed very different bundles before 1914, and that such differences were greater than they have been since. Therefore, differences in relative prices would clearly have differing effects on households at higher- and lower-income ranks.

The daunting task is to capture the expenditure patterns of the rich, and to find price series for the goods and services that took a much greater share of budgets for the well-off. Several such studies do exist, usually for an individual household. Unfortunately, their expenditure categories are too broad for easy attachment to price series, and they have large “other expenditure” shares representing heterogeneous luxuries. We find that their overall expenditure patterns contrast with those of poorer households in broadly the same ways for different countries and eras, as shown by comparing the 1688 English and early-nineteenth-century American budgets featured in this article.Footnote 8

One might think that scholars have recently made good progress up the income ranks here. Yet those prototypical households consuming the “respectability” bundles pioneered by Allen (2001), and also used by other scholars, were not far above the bare-bones consumers in the income ranks, and not so different in the shares of expenditures devoted to staples.Footnote 9 While these baskets have been used heavily to compare living standards over space, they are largely geared toward capturing the realities at the bottom of the income distribution. For example, consider the households consuming the respectability versus bare-bones bundles hypothesized for “1801-03” in Broadberry et al. (2015, p. 339). If we equate their would-be expenditures with their would-be incomes, then even the households consuming the respectability budget would be in the bottom decile of the income distribution.Footnote 10 Nor do the bare-bones and respectability expenditure patterns differ enough to capture the stark contrasts that almost surely prevailed over the higher-income ranks. The share spent on staple grains, which was 61.5% of the bare-bones survival budget, was still as high as 50.8% of the “respectability” budget. One needs to find, and use, expenditure patterns that include household much further up the income ranks. Absent such information, an international appreciation of inequality in the cost of living will be elusive.

Fresh answers emerge from a new collection of consumer household budgets for different income ranks, plus new collections of price data, now including luxury expenditures and luxury prices, which were not covered by the historical literature comparing workers’ bare-bones expenditures. To begin with those income-class household budgets, Table 1 introduces two new historical contrasts between richer and poorer households in the same economy and same era. One contrast is documented for the USA in the early nineteenth century, and one reports what Lindert and Williamson (1982) and Arkell (2006) has distilled from Gregory King’s guesstimates of English consumer behavior in 1688.

Table 1 Expenditure shares for richer and poorer families in two settings
Full size table

The American and English richer classes whose budgets are portrayed in Table 1 were probably in the top deciles (top 10%) of the respective income ranks, and the lower-income budgets characterize poorer groups. For the English poor in 1688, Stone’s grouping of King’s estimates offers an aggregate sketch of the bottom 41% of the income ranks. In the American case, we can rank the two household positions more specifically. The richer-household expenditure pattern was recorded by a high-level federal clerk in Washington DC in 1816–1817 (Woodhouse 1929). Several years later Matthew Carey (1833) produced a credible expenditure budget for a Philadelphia canal worker in 1831. The ratio of their household expenditures was nearly 8:1. These are only two households, or course, but the detail offered by Woodhouse seems to be a fair representation of somebody much higher in the ranks. As best we can tell by comparing the Washington clerk and the Philadelphia canal worker with the income ranks estimated for 1850 (Lindert and Williamson 2016, esp. p. 115), the clerk’s income would have ranked somewhere in the 90th–95th percentile of the nation’s households, while the canal worker probably ranked down around the 15th percentile. Their expenditure patterns contrasted strongly, as shown in Table 1. The canal laborer’s household spent 31.3% of its budget on staple grain products, versus only 10.4 for the clerk’s family.

The expenditure patterns in Table 1 naturally conform to Engel’s Law, with the better-off spending lower shares of their overall budgets on foods of all types. Similarly, within the food and beverage category, staple grains feature prominently in the baskets of the poor, while meat, alcohol, and luxuries take bigger monetary shares in the baskets of the rich. Among the non-food categories, our sources offer only hints about the luxury status of different commodities. Fuel appears to have been a luxury good in both the USA and Canada.Footnote 11 Potentially the ultimate luxuries are the entries shown at the bottom of the US expenditure list (Table 1, Panel (B.)). Personal services, home furnishings, and retail clothing are likely to offer the greatest luxury category. By their nature, these heterogeneous products are hard to capture in consistent price data. For now, the last column of Panel (B.) has decomposed such products into input shares representing labor, wood, and cloth components, so that the next section can convey how the missing prices of such retail products might have behaved.Footnote 12

3 How relative costs of living shaped real inequality over time and space

Combining the consumer expenditure shares of Table 1 with time series on prices reveals how the cost of each consumer bundle, or each lifestyle, moved over the decades and centuries in four settings—southern England, eastern Canada, eastern America, and urban Australia. For each setting, we have spliced long-run price series for twenty or more goods and services. The number of goods and services must be trimmed slightly for international comparisons, since a few goods are available only for some countries and not for others.

Table 2 provides the evolution of such index-number time series for the different countries using English 1688 weights, and Table 3 does the same using American 1816–1831 weights. The price series for Canada through 1850 are taken from Geloso (2019b), with an extension to 1913 assembled from a variety of well-known sources in Canadian economic history and described in the appendix to this paper (see Appendixes 1 and 2 in the supplementary materials).Footnote 13 For the USA, the prices are mostly urban, except where the main series is from T.M. Adams’s (1944) monumental price history based on Vermont farm record books. The data for Australia comes from the work of McLean and Woodland (1992), while the data for Britain comes from the work of Clark (2005). In Fig. 1, the weights used are those for England and Wales in 1688, which can also be called the King-Stone expenditures shares. In Fig. 2, the same can be seen but with the American expenditure shares instead.

Table 2 Cost-of-living movements for richer and poorer income groups, 1690–1912, using consumption bundles from England 1688 Five-year averages (“1850” (i.e., 1848/52) = 1.000)
Full size table
Table 3 Cost-of-living movements for richer and poorer income groups, 1690–1912, using consumption bundles from USA 1815–1831 Five-year averages (same country’s “1850” (i.e., 1848/52) = 1.000)
Full size table
Fig. 1
figure 1

Cost-of-living movements for richer versus poorer income groups, 1690–1912, using consumption bundles from England in 1688. Expenditure shares See Table 1. Price series The main price sources for 1850 are available and documented at gpih.ucdavis.edu. We used Gregory Clark’s series for England. The Canadian series are from Geloso (2016, 2019a, b). The main American series are those supplied by Bezanson et al. (1936) for Philadelphia, Carroll Wright for Massachusetts (1885), and T. M. Adams (1944) for Vermont. Those for Australia are from data underlying McLean and Woodland (1992), helpfully supplied by Ian McLean. In all four countries, most of the prices used were recorded in cities, rather than on farms. The exceptions were some of Clark’s English price series, a few price series from farm households in Vermont, and the data for Canada relating to rural areas that are in close proximity to urban centers in the province of Quebec. The 1850 benchmark All countries’ ratios of (COLrich/COLpoor) are related to the same ratio for the single year 1850 in England, to assure comparability. These differ from the country-specific ratios in Tables 2 and 3, because the latter are all standardized separately so using the base the five-year period 1848–1852 for the same country

Full size image
Fig. 2
figure 2

Sources and notes: see Fig. 1

Cost-of-living movements for richer versus poorer income groups, 1690–1912, using consumption bundles from USA 1816–1831.

Full size image

Shared patterns can be perceived. Both Canada and the thirteen American colonies experienced important shocks during the eighteenth century. For Canada, the most prominent shock was a relative grain scarcity during the Seven Years’ War (1756–1763), which especially hurt lower-income workers more than the better-off because of the British invasion of the colony. However, war shocks in Canada did not always strike in the direction of hurting the poor. During the War for the Spanish Succession (1701–1714), the British caused an important disruption of foreign shipping into Quebec which affected the supply of imported luxuries (whose prices rose faster than that of necessities). For the thirteen American colonies, the main shock of the eighteenth century also came during the Seven Years’ War, as in Canada.

Beyond those North American shocks, what emerges from Figs. 1 and 2 for the eighteenth century, and indeed up to the end of the French Wars in 1815, is the lack of any egalitarian shift in the relative costs of living for Britain, for Canada, or for America. The poor got no relief from the trend toward dearer necessities relative to luxuries, the trend that their ancestral counterparts in the sixteenth and seventeenth centuries had suffered.Footnote 14

Then, the “nineteenth-century” period 1815–1914 brought a clearly egalitarian shift in the price structure for all four countries—England, Canada, the USA, and post-1850 Australia. The net change over these 100 years is unmistakable in Figs. 1 and 2, supported by Tables 2 and 3. That is, something brought relief in the price of necessities relative to staples over those 100 years. This relative improvement for the poor did not advance evenly, however. In all four countries, there was a brief inegalitarian retreat somewhere between 1850 and 1867, though the timing of this retreat depends on which set of weights one uses. Such reversions aside, the relative costs of living in all four countries shifted unmistakably in favor of lower-income groups from the end of the French Wars to the eve of the First World War.

The new numbers unveil not only the great break in trend from the eighteenth century to the nineteenth, but also how the relative cost of living for the different income ranks differed across the oceans. Recent data improvements, particularly the accumulation of information on how to convert old local units into metric measures and in hard currency, open up such comparisons. Recall that such international comparisons have already been introduced for the lower-income ranks, particularly by Robert Allen’s work on “bare-bones” costs (Allen 2009, 2001) and by the publicly available online databases on global prices and incomes.Footnote 15 What we can now add are expenditure shares and luxury prices more appropriate to the higher-income groups. The ability to compare the cost of living of the poor relative that of the rich over space helps us to see which was the “best poor man’s country” (Lemon 1972).

Using 1850 as a benchmark year, this contrast is generated by using the double ratios of expenditures at different prices and expenditures shares

$${\text{Cost-of-living (COL) double ratio for }} 1850 = \frac{{\text{(COL, rich in this country/COL, poor in this country)}}}{\text{(COL, rich in England/COL, poor in England)}}$$
(1)

Applying separate national price levels to Table 1’s two consumer bundles for richer and poorer households in the early nineteenth century yields the results in Table 4. Table 4 shows that both the richer and the poor strata could have found their consumer lifestyle to be cheaper in eastern Canada or in the eastern USA than in southern England, whereas the same was not true in Sydney, at least not in 1850.

Table 4 Relative prices and costs of living in 1850
Full size table

On the other hand, this cost-of-living advantage varied between the two income strata, and is sensitive to the choice of household budget weights. The contrast between American and English price structures makes the American prices look more egalitarian using either set of weights. Not so, however, for the other two countries. To be sure, sticking with English weights would suggest that all three newly settled countries were blessed with more egalitarian prices structures. Yet using the weights from early nineteenth-century America yields nearly a tie game, in which the rich and poor look similarly disadvantaged by higher Australian prices. In Canada using American weights makes the price structure in 1850 look inegalitarian relative to Britain. Thus, Table 4 finds a more egalitarian prices structure in four or five of the six intercontinental contrasts.

Thus far, our results offer new perspectives on both history and geography. Over the decades between the late seventeenth century and the early twentieth, the new cost-of-living history seems to be suggesting a reversal. In the eighteenth century and the start of the nineteenth, we see no improvement in cost pressure on the common-labor subsistence lifestyle relative to the living style of those better off—yet such an improvement arrived unmistakably in each country in the nineteenth century, with the common folk now getting the better break from cost-of-living trends. As for geography, our transoceanic contrasts for 1850 tend to find that price structures in the countries of new settlement tended to favor both income groups, but especially the common folk toward the bottom of the income ranks. Why?

4 Which component price movements shifted the relative costs of living?

What price movements explain the nineteenth-century egalitarian movements, and the present section “accounts for” the observed movements and international contrasts, by breaking down each contrast in the cost of living into its component parts. Starting with this step helps direct our search for deeper, more exogenous, causal forces by providing more detailed results that any causal idea must also explain.

Four explicit accounting exercises are presented in Table 5. The first and third use a price-index identity to “account for” a movement in income classes’ costs of living over more than a century, and the other two use the same identity to account for an international difference in the middle of the nineteenth century. Each panel of Table 5 uses one of the two different sets of expenditure shares from Table 1.

Table 5 Accounting for some differences in the relative cost of living (COL), rich versus poor
Full size table

For each of the four countries, and using either set of expenditure weights, the time-series challenge is the same: To account for those sizeable egalitarian gains, the advance in the relative cost of living a richer lifestyle, unveiled for the nineteenth century back in Figs. 1 and 2 and in Tables 2 and 3. Two positive historical forces combine to account for most of this egalitarian shift in price stricture in all eight cases.

The relatively cheaper cost of grain-based food items accounts for a sizable share of the evolution, regardless of the basket used. The opening of the Canadian west especially with the “wheat boom” of the 1890s (Russell 2012) and a previous agricultural expansion in the province of Ontario (McCallum 1980) meant that the land-rich frontier was being harnessed to increase food production. The same was true for the USA and Australia (Federico 2005: 31–40; McLean 2013: 58–63). This combination of a land-rich frontier, which could be utilized to increase food production, and a growing access to international markets, especially after 1850, thanks to reductions in the costs of transport and communications,Footnote 16 would have magnified the contribution of these food items to the changes in real inequality. For as long as grains were key staples, with income elasticities below one, a relative decline in grain prices was egalitarian. Its contribution assumes great weight here because Engel effects were so powerful in the less prosperous world before 1914, as we reaffirmed back in Table 1.

The other strong influence on changing costs over this century appears to have been the rise in the relative price of services, here proxied by the rise in the wage rate for common labor, given that the rich consumed services more than the poor—especially when the American weights are used. One would expect services, which account for a sizeable share of expenditures by rich households, to depend heavily on unit labor costs, or the wage rate times average labor productivity in the service sector.

We know from the international trade literature that the wage rate tends to be driven by productivity in the whole economy, while non-tradable services tended to have more stagnant productivity before the revolutions in finance, communications, and retailing accelerated after the 1970 s. This means that middle- and upper-class consumers of non-tradable services may suffer more upward price pressure when labor productivity is advancing rapidly economy wide. It is therefore easy to understand the contrast in the role of service prices, tied to the unskilled wage rate, in Table 5’s first and third panels. By contrast, the relative prices of non-grain foods and of the residual category (non-foods other than services) have no consistent impact on the relative costs of living. Not all staple goods were traded internationally, even if some like wheat were. As such, goods like meats, dairy products, and potatoes were not traded internationally (at least until the late decades of the nineteenth century) and they remained noticeably cheaper in the New World. For example, beef in 1850 was 41% as expensive in Canada as it was in England, while that proportion for potatoes stood at 38%. In the USA and Australia, the proportions for meat stood at 37% and 30%, while those for potatoes stood at 38% and 22%. Differences like these made the New World an attractive place for the poor, even in the case of Canada (Geloso 2016, 2019a, b).

It turns out that these two positive forces—cheaper staple grains and more expensive labor-intensive luxury services—also help to explain those intercontinental contrasts in living costs in 1850, introduced in Table 4. As the spatial contrasts for 1850 in Table 5 show, both the grain effect and the services effect stand out as features making the price structure more egalitarian in the three countries of recent settlement than in England.Footnote 17 Here again, the roles of non-grain foods and of non-service luxuries are less consistent.

5 Revising the trends in top income shares

How and when did the trends in the relative cost of a richer lifestyle affect the magnitude of the now-famous rise of inequality before 1914? Mild egalitarian trends in the cost of living, such as those we find for the four countries we consider, attenuate increases in income inequality. In other words, adjusting for those trends in the cost of living allows us to capture “real” inequality. For two of our four countries, we can use the measures developed in Part III to convert the conventional measures of nominal income inequality into corresponding real-inequality measures. For Britain between 1688 and 1911, we can apply our cost-of-living adjustments to the latest estimates of nominal top-income shares, and we can do the same for America between 1774 and 1914. The top-income shares most conveniently available for these full-time spans are the top-one-percent share and the top-ten-percent share.Footnote 18

To attach our cost-of-living measures to movements in nominal income shares requires a simple transformation of movements in nominal income shares into movements in ratios of average real incomes. We offer two such transformations in Table 6. First, to follow the literature’s frequent focus on the relative income of the top 1 percent, we contrast the real income movements of the top 1% with those of the bottom 99%. Second, to remain truer to the income ranks of our government clerk versus laborer, we contrast the real income movements of the top 10% with those of the bottom 40%.

Table 6 What the cost-of-living ratios imply about real household-income inequality trends in Britain and America, 1688–1914
Full size table

The resulting movements are sketched for benchmark years in Table 6 and in the two-part Fig. 3a, b. Our derived real inequalities confirm some qualitative impressions from the current literature on nominal inequalities in Britain and America, yet seem to affect the long-run trends in, and the timing of historical peaks in, income inequality. For Britain, Fig. 3a, b starts with nominal rich/poor income ratios from Robert Allen (2018). The richest 1% had their peak advantage in nominal income in the 1867 benchmark (Fig. 3a), whereas the whole top 10% had their peak advantage in nominal incomes come earlier, at the 1798 benchmark (Fig. 3b). Then, both top groups lost ground between 1867 and 1911. Adding the social differences in the cost of living does not change the directions of movement, but it magnifies the leveling of incomes after 1867.Footnote 19 Whichever estimates one prefers, Britain’s real inequality between the top 1% and the rest came sometime back in the nineteenth century, and not as late as the early twentieth.

Fig. 3
figure 3

What the cost-of-living ratios imply about real income inequality in Britain And America, 1688–1914 (See Table 6)

Full size image

For America, the real inequality movements from 1774 to 1914 departed more visibly from the nominal movements. In nominal terms, the rich kept steadily pulling ahead of the poor, aside from a pause during the Civil War decade (Lindert and Williamson 2016). By the eve of the First World War, in nominal terms, the Americans were as unequal in their incomes as the British. In real terms, however, relative price movements eliminated either half (Fig. 3a) or all (Fig. 3b) of the rise in inequality over those 140 years.

6 Remaining caveats and new frontiers

Our results offer important insights into the study of inequality. The first is that it reinforces the contention that more attention must be given to price structures in order to reflect “real” inequality differences between rich and poor. There are many factors related to prices that can render incomplete any nominal-income based portrait of inequality. This is true over time (as the relative price of certain goods changes over time) and over space (as certain goods tend to be cheaper in certain areas—especially when these goods are not internationally traded).

However, there are caveats to be underlined in order to properly expand this research frontier. We have already mentioned some of them. Ideally, as we have noted, one should complement the constant elasticity measures used in this paper with fixed quantities of goods (like poverty baskets and respectable-living baskets), to see whether the elasticity assumptions make any great difference. This would call for the design of baskets, along the methodological lines drawn by Allen (2001), but for the very richest in each society, as we do have well-established baskets for those near the poverty line.

Another caveat relates to urban-tenant bias in real income measures, including ours. Regions and classes differed in their tendency to supply themselves with goods and services, especially food and housing. Missing this point would mean misreading the history of real incomes. In the real-wage literature, for example, workers are supposedly non-owning tenants who pay for all their food and rent for occupying housing they do not own, an assumption that fits low-skill laborers in England. Yet if a large share of them (as in the case of Canada, the USA, and Australia in this paper) do own their own food-growing resources or own their homes, either as outright freehold or as long-term tenants paying less rent that the current market value, these implicit values of income should be included in any measure of nominal income. In the real wage literature, it has not been included. Contrary to an inference invited by that literature, when food prices rose by 30% relative to farm wages, ordinary peasants did not experience anything like a 30% drop in real purchasing power. As net sellers, many peasants actually gained from that rise in food prices.

The degree of home production also constrains Engel’s Law. For example, consider the shares of purchased food in total purchases for different classes in the USA in the 1830s, as sketched by Dorothy Brady (1972, pp. 76–83). For a farm family with annual expenditures of only $200–500, the food share of total expenditures was 34 percent, which was no higher than the mid-range share, again 34 percent, for urban families having total expenditures of $1000–3000. What Engel had in mind in his classic studies of the food share of total expenditures was its share of total consumption expenditures for families who bought all that they consumed. To restore his Law, one must compare richer and poorer with similar shares of home production. In this same case of the USA in the 1830s, Engel’s Law re-emerges when we compare the shares of all expenditures devoted to food within the cities—as high as 61% in that poorer $200–500 group versus the 34% for the $1000–3000 group. Yet if we carelessly compare farm with city, the importance of home production of one’s own food (and housing and unskilled services) was historically so great as to confound the usual assumption that food is the ultimate staple in terms of purchases (as opposed to total consumption).

Our calculations have side-stepped the home-consumption issue. Even in our frontier economies, the relevant price series are mainly non-farm, and even urban. For Canada, the prices relate to the main cities plus the proximate countryside around those cities. Some of them should be categorized as urban prices, but some are rural prices (but not hinterland prices, which would be quite problematic). For America, the series are largely from Philadelphia, except that we have (courageously) spliced Vermont farm figures onto the more urban series, to extend the indices past 1896 to 1914. For Australia, too, the series are urban, coming mainly from Sydney. One obvious solution would be to find benchmark year estimates of prices in both rural and urban areas in order to properly assess differences.

The fact that data are more available for urban markets than for the countryside means that our findings should be read as applying especially to urban households, or more broadly to those who buy all of their consumption in the marketplace, rather than producing much of it themselves. The few cases where we can observe both urban and farm-gate prices underline how the urban–rural price ratios drifted across the nineteenth century, with major implications for the study of inequality. We find that the prices of staple grains dropped much more in the cities (data from Philadelphia, New York, and Boston) than on farms (in Vermont). This movement was presumably aided by improvements in transporting grains from farms to urban markets. On the other hand, the prices of two relative luxuries, sugar and cotton sheeting, rose in the cities relative to those farms. Again, we presume that improved transport played a role, easing the shipment and sales of these largely imported goods from the cities into the hinterland. The upshot for relative living costs is clear. In relying on urban price data we, like other scholars, have quantified movements that were particularly egalitarian in the cities. We would have found much less egalitarian drift in the cost of living in the countryside—where, in any case, the staple goods in question were often home-produced. The cost-of-living story, in other words, carries particular weight in the urban part of the economy.

Each of these caveats has its positive side, in the form of a further research frontier to be conquered as historical data continue to become more available.

For the present, we have at least opened a new historical geography, suggesting when and where the cost of living moved quite differently for different income ranks. Relative to the emerging history of nominal income (or wealth) inequality, we find that not only in Western Europe but also in other continents, cost-of-living movements failed to move in favor of the poor in the eighteenth century, but did so across the nineteenth century. We have also suggested two main explanatory factors: the real prices of staple grains and of luxury services. Given the roles played by these two forces, one could tell a new story of how a delayed rise in (decline of) the real grain wage translates into an egalitarian (inegalitarian) movement in relative lifestyle costs.