PROBATE RECORDS AS A SOURCE ON PREINDUSTRIAL OCCUPATIONS

A potential source for the English occupational structure over time is statements of occupation in probate records.Footnote ¹² After 1695 parish registers of baptisms, marriages, and deaths begin to more frequently record occupations, and as a census of the whole adult population are likely a preferred way of estimating the occupational structure of the population.Footnote ¹³ But before 1695 parish records rarely note occupations. Even post 1695 the parish registers reveal little about female occupations, women being one-third of the labor force in 1861. And by 1841 and later, the female share employed in farming is much less than for men. The only alternative sporadically available occupation sources before 1695 are muster rolls and tax lists.

Probate records, however, indicate occupations for a substantial number of those probated from the 1540s onwards. Indexers of probate archives often list these occupations, making large numbers of such statements available for analysis 1540–1857. Figure 3, for example, shows the estimated fraction of adult men dying in Buckingham, Essex, Kent, and Suffolk who were probated in the various ecclesiastical courts covering these counties.Footnote ¹⁴ This is, by decade, the number of surviving probate records for men, divided by the estimated number of adult men dying.Footnote ¹⁵

Though they cover a substantial fraction of the preindustrial, male population probates are potentially misleading on the sectoral occupational distribution in that richer men were always more likely to be probated. If an industry was capital or land intensive, it would be overrepresented in probate occupations (assuming the producers tended to also own the production equipment). Thus in 1841, when we can compare the distribution of occupations in probates in Essex with those in the census, shoemakers (little capital) are undercounted in probates, and millers (much capital) are overcounted. Probates will reveal the true share in farming only if farm producers are as likely to be property owners as nonfarm producers. Fortunately, as we will see below, this appears to be the case in all periods 1540–1858. Indeed, if anything, will occupations oversample the farming sector. Nigel Goose and Nesta Evans, for example, claim that in the sixteenth and seventeenth centuries will making was relatively less frequent among men in larger towns.Footnote ¹⁶

FIGURE 3

FRACTION OF MEN PROBATED BY DECADE, 1540–1861

Sources: Clark, “Consumer Revolution.”

HOW REPRESENTATIVE WERE PROBATES OF THE FARMING EMPLOYMENT SHARE?

The key test of whether male probate occupations accurately represented the farm share is whether such occupation reports were generated proportionately to population in rural and urban areas. Suppose the proportion of men probated differed between the farm and nonfarm sectors: the fraction of the population reporting an occupation through probate was ω_f in farming, and ω_nf in nonfarm occupations. Suppose also ω_f < ω_nf. In that case, the probate occupation rate in completely rural parishes would be, at ω_f, lower than the proportion in completely urban parishes, ω_nf. In general suppose that the true fraction of a parish in farm occupations was ϕ. Then the overall rate of probate occupation reporting per person will be

(2)

FIGURE 4

HYPOTHETICAL PROBATE RATES AND NONFARM EMPLOYMENT SHARES, Ω_F ≠ Ω_NF

Sources: See the text.

If ω_f < ω_nf then the more rural the parish (the greater is ϕ), the lower the probate occupations rate. Figure 4 illustrates what the frequency of probates would look like as we move from all farm parishes to no farm parishes if ω_f = 0.2, and ω_nf = 0.4. We would see a clear gradient in probate frequencies between town and country.

Thus the key test of whether will occupations tend to over or understate the share in agriculture is the frequency of occupation reports relative to populations in rural and urban areas. If rural populations have probate occupation rates the same rate as urban rates, then probates must correctly report the farm share of the population. Only if probate occupation frequencies were lower in rural areas would probates understate the share in farming.

In England from 1540 to 1841 probate occupations tend, if anything, to modestly over sample the rural population in all periods, and thus are biased towards reporting too many farm occupations. In 1801 and 1841 probate occupation reports come in exactly the same proportion as populations in rural and urban areas. Figure 5 shows the share of probate occupations and population, by parish population density, for 1801 for Essex. The correspondence is close. A similar figure for 1841 for Essex shows exactly the same correspondence.

FIGURE 5

THE DISTRIBUTION OF PROBATES VERSUS PEOPLE, ESSEX, 1801

Note: Probates are from the years 1781–1822, parish areas from the 1851 Census.

In earlier years, probate occupation reports cover a larger share of the male population, including poorer men. But this does not seem to bias them in any way towards nonfarm populations. Now, indeed, they tend to modestly oversample rural areas. One earlier measure we have of parish populations, and population densities, is the Hearth Tax of 1662–1689. The Hearth Tax was an assessment collected on every fire hearth or stove. The records list both the numbers of 1841 hearths and the houses in each parish, even the hearths of those so 1841 poor they were excused from the tax. We measure parish population density 1662–1689 through hearths per acre, rather than houses, since richer houses would have multiple hearths, which represent their likely Tax large population of servants. For example, Ampton in Suffolk in 1674 had just eight houses. But one of these, that of Sir Algernoone May, had 21 hearths, so that the parish had 38 hearths, which likely better indicates its true population.

FIGURE 6

PROBATE OCCUPATIONS VERSUS HEARTHS PER ACRE, ESSEX, 1670

Sources: Ferguson et al., Essex Hearth Tax. Parish areas, 1851 Census.

Parishes were divided into six groups based on the number of hearths per acre: from the most rural with 0.00–0.06 hearths per acre, to the most urban with more than 0.3 hearths per acre. Figure 6 shows for Essex the distribution of population across parishes of the various levels of hearths per acre in 1670, compared to the distribution of probates reporting occupations in 1650–1699.Footnote ¹⁷ In 1670 the probates tended to modestly oversample occupations in rural areas. A similar pattern is revealed in Suffolk in 1674. Figure 7 shows the distribution of parishes across different population densities in Suffolk as revealed by hearths rural parishes with less than 0.2 people per acre, to the most urban with more than 7 people per acre.

We get an even earlier estimated parish population density for Suffolk from the records of the Lay Subsidy of 1524. These lists of taxpayers by parish, which included as well as property holders also wage earners earning at least £1 per year, allow an approximation to parish population densities in Suffolk in 1524.Footnote ¹⁸Figure 8 shows that even as early as 1540–1599 will occupation reports are distributed across rural and urban parishes in proportion to their relative populations in 1524. From at least 1540 the share of people reported as farming in probate records will provide an unbiased estimate of the farm share in the population.

FIGURE 7

PROBATE OCCUPATIONS VERSUS HEARTHS PER ACRE, SUFFOLK, 1674

Sources: Hearths, Hervey, Suffolk in 1674. Parish areas, 1851 Census.

FIGURE 8

PROBATE OCCUPATIONS 1540–1599 VERSUS TAXPAYERS, 1524, SUFFOLK

Sources: Lay Subsidy payers, Hervey, Suffolk in 1524.

TABLE 1

SHARES IN FARMING AND FISHING, 1841

Notes:

Occupation shares based on all probates for Essex and Surrey, 1824–1857.

Thus probate occupations in preindustrial England, if anything, tended to modestly overstate farm occupation shares in both the sixteenth and seventeenth centuries. The likely reason for this earlier bias is that farming in preindustrial England was even more property intensive than town production activities, because of the value of land and farm inventories.

The second test of will occupation reports as a source on the general occupational split between farming and fishing and other occupations comes by comparing these reports for the years 1824–1857 with occupations in the 1841 Census. The 1841 Census has the advantage that occupations were recorded in a terms very similar to those of occupation statements in probates. Table 1 shows the census share in farming in Essex and Surrey, counties with very different degrees of urbanization, compared to the shares among men's probates stating occupations.Footnote ¹⁹ The Essex shares match almost exactly. In Surrey, the probates report too many men in farming. But this stems largely from there being disproportionately many probates from rural areas. Once we correct for that, the farming share in probates is within 2 percent of the census share. While the probates show accurately the farm— nonfarm division of the population, they are not so useful in revealing the occupational structure within the nonfarm sector, because, as noted, more capital intensive occupations produce more probates.

FIGURE 9

THE PROBATED SHARE IN FARMING BY PARISH POPULATION DENSITY, ESSEX

Sources: Parish populations, 1640–1669, Ferguson et al., Essex Hearth Tax, and the 1801 and 1851 Censuses. Parish areas, 1851 Census.

The reason we find below that the share of farming nationally was only 60 percent even before 1660, despite low levels of urbanization, is that even in completely rural areas a significant fraction of men reported occupations outside farming. Figure 9 shows the share of testators reporting farming or fishing as their occupation in Essex in three periods—1640–1679, 1780–1819, and 1820–1858—where parishes are grouped by their estimated numbers of people per acre in each period.Footnote ²⁰ Even in the most rural parts of Essex in 1640–1679, those with more than 6 acres per person, only 75 percent of the probated reported farming occupations. And in a broad swath of parishes which would not be counted as urban—they had between 1 and 5 acres of land per person—only a minority of the probated engaged in agriculture.

Looking at parishes with less than 0.3 hearths per acre in Suffolk in 1675, for the years 1600–1702 the main occupations were as follows: farming and fishing, 0.666, clothing production and distribution, 0.128; clergy, 0.049; food production and distribution, 0.046; building trades, 0.036; transport, 0.027; and metalworking, 0.019. Rural preindustrial England thus had both significant manufacturing activity, but also trade and transport. The share of the probated in the most rural parishes of 1670 reporting farming occupations circa 1800 and 1840 declined only modestly—to 69 percent in 1800 and 61 percent in 1840. The preindustrial countryside in England had many manufacturing and service workers.

We see exactly the same phenomenon of significant numbers of rural workers in nonfarm employment in the occupation reports from the earlier Poll Tax returns of 1377–1381.Footnote ²¹ This is why urbanization rates in preindustrial England provide a very poor guide to the share of the population employed in farming.

NATIONAL FARMING SHARE, 1652–1660

Before 1858 estates in England were probated in a great variety of ecclesiastical courts. Thus getting a representative sample of probate occupations for every county in England would be an enormous task. However in 1652 to 1660, in the Interregnum, the great majority of probates were filed in the newly established national Court for Proving of Wills and the Granting of Administrations, which sat in London. The index of this court in these years is available to us from the National Archives, filed with the Prerogative Court of Canterbury probates. We thus have for 1652–1660 an estimate by county of the farming share in employment.

Table 2 shows for each county in England the number of houses in 1689 estimated from the hearth tax.Footnote ²² This measures the likely relative populations of these counties in 1652–1660. In the third column in the table are shown the number of interregnum probates giving occupations for each of these counties in 1652–1660. The probate frequencies relative to the likely populations of the counties are shown in column 4, with 100 set as the average. The closer is a county to London, the greater the probate numbers relative to estimated population. Thus the most northern counties, Cumberland, Northumberland, Durham, and Westmorland, all show only 21–29 percent of the average number of probates per person. To get the national share in farming and fishing we thus have to reweight the share in farming in each county by the relative county population, to correct for the underrepresentation of the north of England.

TABLE 2

FRACTION IN FARMING AND FISHING BY COUNTY, 1652–1660

Sources:

Houses, 1689. Glass, “Gregory King's Estimate,” p. 372, Houghton estimate.

The fifth column shows the fraction of will makers in each county in 1652–1660 who reported their occupation as being in farming or fishing. This varied from a low of 9 percent in Middlesex to a high of 79 percent in Cheshire. For 1652–1660 the population weighted average for the country as a whole was 59 percent. Thus even one hundred years before the Industrial Revolution, a substantial fraction of the English labor force, 41 percent, was engaged in other than agricultural occupations.

Interestingly, in 1652–1660 there is little evidence from the probates of the later industrialization of the north. By the 1760s Leigh Shaw Taylor and E. A. Wrigley report that about 65 percent of the occupied of Lancashire and the West Riding of Yorkshire were employed in manufacturing. As Table 3 shows, on average 68 percent of the probated in the north were engaged in farming, compared to 56 percent in the south. Lancashire, in particular, reports 77 percent in farming, well above the national average.

Table 2 also shows the fraction of those probated engaged in the production and distribution of textiles and clothing, shown in the last column.Footnote ²³ Lancashire is below the national average of 13 percent, and Yorkshire just at this average. Concentrations of such workers appear in Middlesex, followed by Worcester and Gloucester. However, if we look just at the production of cloth (such occupations as weaver, fuller, dyer, shearman, comber, clothworker, and clothier) the county with the largest percentage was Suffolk (9.2), followed by Worcester (8.1), Devon (8.1), Gloucester (7.8), Norfolk (7.3), Somerset (7.3), Essex (6.9), and Wiltshire (6.9). Cloth production was concentrated in East Anglia and the South West, as would be expected from the history of these industries.

TABLE 3

EMPLOYMENT COMPOSITION 1652–1660, AND 1560–1579

Sources:

Tables 2 and 4.

NATIONAL FARMING SHARE, 1560–1579

From probates, the earliest broad representative estimation of occupational structure we can get is 1560–1579. Before this few probates state occupations, and those that do are potentially unrepresentative, since they seem to contain too many clergy. Probates fell under ecclesiastical courts, and clergy may have had easier access to these courts in early years.

We have thus constructed for a broad range of 17 counties and parts of counties in 1560–1579, representing 33 percent of the population of England, county-level estimates of the shares of probates listing occupations in farming and fishing. The sample includes parts of South London, so it will capture hopefully a representative sample of the effects on occupational shares of the growth of London. Because ecclesiastical jurisdictions crossed county boundaries, in some cases we only have data for part of the county in 1560–1579 (Cambridge, Dorset, Gloucester, Kent, Norfolk, and Warwick). In these cases, we compare the occupation shares in 1560–1579 with the share of the equivalent parishes later. Figure 10 shows the distribution of these counties across England. Where the partial county data was more limited, we expanded the years covered to 1550–1589 to reduce sampling errors earlier.

FIGURE 10

THE SAMPLE COUNTIES IN ENGLAND, 1560–1579

Note: Lighter-shaded counties are those where less than half of the county was included.

The farm share estimates for 1560–1579 are shown in Table 4. Column 3 reports the farming share in these counties in 1652–1660. Column 5 shows the same raw estimated farm share for 1560–1579 for these counties, and column 4 the number of observations 1560–1579. There is no shortage of observations on which to make this estimate: 16,771 occupation statements. Based on the raw numbers, the same share of the population was employed in farming in 1560–1579 as in 1652–1660 in the sample counties. The implied national population share in farming would be 59 percent in 1560–1579, the same as in 1652–1660.

As Table 3 shows for these counties, there was little change in the allocation of the probated across occupations between 1560–1579 and 1652–1660. In 1652–1660, 13 percent of the probated were in clothing production and distribution, but the comparative figure in 1560–1579 was already 12 percent. Three percent were in building in both periods. The nonfarm activity that did show a significant increase was food manufacture and distribution, up from 3 percent to 7 percent. There was a rising share across all categories of food manufacture and distribution—bakers, brewers, butchers, grocers, and maltsters—between these two dates. But this was counterbalanced by a declining share of the probated reporting church occupations. An average of 6.5 percent of men report clerical occupations in 1560–1579, nearly double the 3.6 percent clergy for these same counties in 1652–1660. Some of this higher share may reflect occupational realities in the mid-sixteenth century. The extensive holding of religious institutions were seized by the crown only in 1536–1539, with many monks and priests pensioned off, so plausibly a higher fraction of men dying 1560–1579 were indeed clergy. But some may reflect either a higher likelihood of clergy being attributed an occupation in earlier years, when few men had an occupation attributed to them, or a higher likelihood of clergy having wills proved in church courts.

TABLE 4

FRACTION FARMING AND FISHING BY COUNTY, 1560–1579

Notes:

^a Part of the county, 1550–1589; ^b Part of the county, 1550–1589; ^c Excluding Bristol; ^d East Kent; ^e Norfolk Deanery; and ^f Parishes in the Worcester diocese, 1550–1589. The farm share in 1652–1660 is calculated for the same parishes as earlier in all cases.

FIGURE 11

FARM SHARES, 1560–1579 VERSUS 1652–1660

Sources:Tables 2 and 4.

To control for possible overrepresentation of clergy in the early years, the last column of the table reports for each county in 1560–1579 the share in farming, making the share of clergy in 1560–1579 the same as that for the same county in 1652–1660. In this case, the estimated share in farming in 1560–1579 in the sample counties becomes 66 percent, and the implied national share in farming in 1560–1579 is 61 percent. Thus we can conclude that the national share of employment in farming in 1560–1579 was, at 59–61 percent, little more than in 1652–1660.Footnote ²⁴

Figure 11 shows the estimated share in farming in counties in 1560–1579 versus the share in 1652–1660. As can be seen, there is a close relationship between the shares in these two periods (the correlation coefficient is 0.7). Also shown is the 45 degree line, where the shares are the same in both periods. Most counties are very close to this line.

Though we lack information on any county in the north other than Cheshire in 1560–1579, this will not likely bias our estimated farm share then, since by 1652–1660 the farm share in the north was still higher than in England as a whole. The north became heavily involved in manufacturing only after 1660.

BIASES IN THE PROBATE ESTIMATED FARM SHARE?

Are there biases in the probate occupation statements that would bias downwards estimated farm employment in 1560 and 1650? Many men in the preindustrial era pursued a variety of occupations, such as weaver and farmer, but the probate statements normally state only one occupation. In a sample of 5,000 wills from Suffolk, Essex, and Surrey pre-1650, 21 percent of farmers, for example, left cattle. But 8 percent of craftsmen—carpenters, smiths, etc.—also left cattle. A significant fraction of craftsmen engaged in some farming on the side. But while this adds noise to estimates of the fraction engaged in farming, but there is no reason to expect that it would bias estimates against farm occupations. For just as craftsmen engaged in some farming, so did early farmers engage also in some construction, manufacturing, service, and trade. Farm households, for example, routinely engaged in the production of homespun cloth, in shoes, and in wooden objects.

Suppose the distribution of the fraction of time spent on farm and nonfarm work across the population was as in Figure 12. The fraction of the population who spend the majority of their time in farming is 0.6, but on either side of that share there are many people who split their time between both sectors. However, if we exactly calculate the share of time that the average person spends in farming, it is also 0.6. The only way in which making the split at the point where people's time switches towards more than 0.5 in agriculture would underestimate the share in farming would be if people who spend most of their time in nonfarming pursuits engaged in more farm activities than the corresponding people who were mainly in farming. There is no evidence this was ever the case.

A second issue is the substantial numbers of will makers who instead of an occupation were described as “gentleman” or “esquire.” In Suffolk 1600–1702, for example, 9 percent of men beyond those with a specific occupation are described only by honorifics such as “gentleman” or “esquire.” If these men were really just high-class farmers, then we would underestimate the share in farming. However, if that was the case, then “gentlemen” should appear more frequently per head of the population in the most rural areas. Figure 13 shows the distribution of testators described only as “gentleman” or the like compared with the distribution of hearths by parish hearth density for Suffolk, 1600–1702. The distributions are very similar, with just a slight bias in favor of rural locations. Gentlemen as a group were not just high-status farmers.

FIGURE 12

THE ESTIMATED SHARE IN FARMING WITH MIXED OCCUPATIONS

Sources: See the text.

A third issue with probate occupation reports is that they were made at the end of the lives of testators. Before 1700 the average testator had an estimated age at making the will of 53. In contrast, the average worker would be aged about 39 years assuming that everyone aged 15–65 was employed. What if there was a life cycle where men were more likely to be employed in farming when young and in nonfarm occupations when older? In that case, probates overstate the share of the total labor force employed outside farming. Using a large sample of probates assembled from Essex, Suffolk, and Surrey in 1560–1699 we can show, however, that if anything the life-cycle path was for a larger share of men to be employed in farming at older ages.Footnote ²⁵Figure 14 shows the share of testators reporting farming or fishing as occupations as a function of their estimated age at the time of the will. The dotted line shows the expected share on the hypotheses that men began work in farming then moved to nonfarm occupations. As can be seen there instead a modest increase in the share of men reporting farm occupations with age.

FIGURE 13

DISTRIBUTION OF “GENTLEMEN” VERSUS HEARTHS, SUFFOLK, 1600–1702

Sources: Hearths, Hervey, Suffolk in 1524. Parish areas, 1851 Census.

The observed pattern of additional older men reporting farm occupations is clearly the product of those in farm occupations living longer on average.Footnote ²⁶ The gain in the farm share, for example, is as much in the years 50–59 to 80+ as it is from ages 20–29 to 50–59. Presumably changes in occupation were much more common for men of working age than for those who would be effectively retired. In that case, the rate of probates relative to those occupied in farming would be lower, all things equal, than the rate in nonfarm occupations. However, we have seen above that the probate rate relative to the farm population is if anything higher in England before 1700 than in the nonfarm population. So even though those engaged in farming lived longer, there is no evidence that probates understate farm occupations because of either life-cycle occupation changing, or differences in average years lived by occupation.

The final issue is the effect of the omission of women. Women were not included in the probates sample, since occupations are rarely recorded, and the most common one recorded “spinster” is ambiguous between being an occupation and a statement about marital status. Could this omission bias the estimated occupational shares against farming? Suppose probates really only describe the occupational split of the male population. In both 1381 and 1841 women tend to be employed significantly less in farming than men.Footnote ²⁷ So in this case the low farm shares observed in 1560–1579 and 1652–1660 would be underestimates. England in the preindustrial period was even less agricultural than the probates would suggest.

FIGURE 14

AGE AT PROBATE AND SHARE IN FARMING, 1560–1699

Sources: Clark and Cummins, “England's First Demographic Transition.”

Another possibility is that the occupations of male testators, mainly richer men and property owners, serve as an index of the location of economic activity in general. There were the same numbers of workers, both men and women, associated with each of these property owners in the farm and nonfarm sectors. Thus though women are not directly included, the index we use implicitly is counting all workers in each sector. In this case, the distribution of probate occupations will not overstate the farm share in employment. But in both cases, the omission of women will not bias us towards finding too few farm occupations.