August, 1990 publication (minus footnotes to be found on Internet Archive): The Counting of Manors Professor Stone's reply refuted

 THE COUNTING OF MANORS: PROFESSOR STONE’S REPLY REFUTED

In February, 1972, the Economic History Review published an article of mine examining the estimates of landed income and of the manorial holdings of the peerage presented by Professor Lawrence Stone in his book, The Crisis of the Aristocracy 1558-1641. I argued that, on the basis of his statistics, wide variations could be detected in the average annual values of manors held by different groups of peers in 1558/9, 1602 and 1641 and that, as a consequence, manorial counts could not be relied upon as indicators of changing landed wealth. My argument was met by vigorous objections to my procedures from Professor Stone and by revised figures which, prima facie, eliminated the discrepancies I had revealed. I shall show here that Professor Stone’s objections to my procedures were fallacious and that his apparent elimination of variations in average annual manorial values rests upon a straightforward assumption that they do not exist.

The Counting of Manors Reconsidered

The principles upon which my article was constructed were relatively simple. First of all, I accepted the figures for the landed income and for the manorial holdings of the peerage presented by Professor Stone in 1965 with all their limitations. I explicitly acknowledged that, in his study, he was dealing with a restricted group and that there were serious gaps in the economic evidence at his disposal. These gaps on the financial positions of individual families had been filled by judicious estimates and with the aid of corroborative evidence. Professor Stone’s “rough calculations” of the landed income of the peerage showed its comparative distress at the beginning of the seventeenth century and its failure, despite a recovery, to regain its former position in landed society by 1641. These calculations, which relied partly upon guesswork, were supported by firmer testimony. Professor Stone’s two manorial samples indicated that the dispersion of manorial values relative to the mean remained much the same between 1535 and 1597-1608 and that manorial counts were valid if they could be shown to relate to large numbers of representative manors. His analysis of the manorial holdings of the entire peerage, of those of groups within its ranks and of its holdings within the area then covered by the Victoria County History showed how the landed resources of the peerage had been depleted. Professor Stone had been admirably frank in describing the difficulties he had faced in making his estimates. For all the documentary deficiencies and the margins of error in the manorial calculations, the two largely independent sets of figures appeared to corroborate one another.

It was clearly no simple matter to attack this case. The manorial samples could not be directly challenged. Professor Stone had, however, explained that his figures for the manorial holdings of the entire peerage in 1558, 1602 and 1641 were subject to a margin of error of plus or minus 10 per cent owing to uncertainty about the data. He had further explained that the estimates he offered for the manorial holdings of the 41 peerage families extant from 1558 to 1602 and for the 33 such families extant from 1558 to 1641 were subject to a sampling error of plus or minus 6 per cent and a margin of error of plus or minus 4 per cent: in other words, they, too, had an overall margin of error of plus or minus 10 per cent. It followed, by definition, that the manorial holdings of the families outside the ranks of the 41 and 33 families were also subject to a margin of error of plus or minus 10 per cent. The possibility of comparing the average annual values of manors held by different groups was opened up provided calculations could be fairly made of the overall landed incomes of the families involved.

Professor Stone had certainly been extremely explicit in 1965 on the hazards of placing families in categories of rental income for 1559, 1602 and 1641. A few, as he pointed out, had no doubt been placed one above or one below their true position. There was clearly no possibility of extracting exact figures for the gross rentals of the 41 and 33 families whose manorial holdings he had analysed in some detail. Nonetheless, an attempt could legitimately be made to arrive at an overall estimate of their gross rentals “by dividing the total gross rental for the group in which they appear by the number of families in that group” and adding the estimated figures for each group together. The Rich family, for example, appeared in Professor Stone’s Group V in 1559 amongst the 15 families with a total gross rental of £37,500: its gross rental at that date could be estimated as £37,500 divided by 15 = £2,500; in 1602, the Rich family was apparently amongst the 10 families in Professor Stone’s Group V with a total gross rental of £43,000 and its gross rental then could be estimated as £43,000 divided by 10 = £4,300; by 1641, the Rich family (Earls of Warwick) was amongst the 8 ‘old’ peerage families in Group V with a total gross rental of £44,000 and its gross rental may be estimated to have been £44,000 divided by 8 = £5,500. The method was time-consuming and involved calculating the mean gross rental of the category of income in which each of the 41 and 33 families were to be found in Professor Stone’s tables in his Appendix VIII. But it did have the advantage of taking account of any bunching within each income group. It provided the most accurate gauge of the overall total gross rentals of the 41 and 33 families that the imperfections of Professor Stone’s original estimates permitted. Once casualties had been added to the 1558/9 and 1602 estimates (but not to that for 1641), it was possible to produce an estimate of the landed income the 41 and 33 families received from their manorial holdings at the relevant dates: by dividing these estimates by the number of manors held, a calculation could then be made of the average annual value of one of their manors: a simple process of subtraction and division permitted this figure to be compared with the estimates for the average annual value of manors held by the rest of the peerage.

This operation rested on a recognition of the problems posed by the form of Professor Stone’s 1965 estimates of gross rental and by the margin of error allowed for in his manorial calculations. The results were striking. My estimate of the total gross rentals of the 41 families with 1,982 manors (subject to a margin of error of plus or minus 10 per cent) in 1558/9 was £73,605 4s 8 1/2d; with casualties added, their total landed income was estimated to have been £88,365 4s 8 1/2d; at the mean holding of 1,982 manors, the average annual value of one of their manors was £44.58. If Professor Stone’s contention that his largely independent estimates of the peerage’s landed income and of their manorial holdings were broadly compatible had been right, it might have been expected that the average annual value of a manor held by the rest of the peerage would have fallen within plus or minus 10 per cent of £44.58, i.e. between £40.122 and £49.038; in fact, my calculations indicated a figure of £33.12. The manors held by the 22 families with a mean holding of 1,408 in 1558/9 appeared to be over 25 per cent poorer on average than those held by the 41 families. In 1602, the manors held by the 16 families with a mean holding of 801 manors were apparently almost 23 per cent poorer on average than those held by the 41 families. If the same comparison was made between the estimated average annual value of a manor held by the 33 families and that of one held by the rest, the latter were poorer by over 26 per cent in 1558/9 and by over 13 per cent in 1602 but were richer by almost 23 per cent in 1641. Professor Stone could have doubled his margin of error to plus or minus 20 per cent and only have saved his argument in one of the five comparative cases I set out.

Variations of this order, let alone those disclosed between the average annual values of manors in the area covered by the V.C.H. and the rest of the country, undermined Professor Stone’s argument. His original estimates of landed income and of manorial holdings were clearly incompatible with one another and with the claim that manorial counts could be used as a reliable index of changing landed wealth.

Professor Stone’s Methodological Objections

In his reply, Professor Stone faced the difficulties posed by my article directly. He very properly stressed the uncertainty of his original estimates of the income of all peerage families and the margin of error to which his statistics on manorial holdings were subject. With equal propriety, he declared his intention to “look very closely at the way he [Thompson] has arrived at his statistics” before seeing if another explanation was possible for the discrepancies revealed and finally asking what difference my analysis made anyway.

This exercise began with an examination of the way in which I had estimated the “mean income” of the 41 and 33 families “from the eight very rough income groups” into which Professor Stone had originally placed them. I had carefully explained in a footnote in my article that “for each of these families the gross rental has been calculated by dividing the total gross rental for the group in which they appear by the number of families in that group” and gave as an example the calculations for the Rich family which I cited above. I had also taken the further precaution of supplying Professor Stone with my tables showing the estimates of gross rental for each of the 41 and 33 families and my other calculations in July, 1970. As he recognised in his reply, this had entailed using his data “to work out the figures per family down to the last halfpenny” in many instances. There was thus no lack of clarity about the method I had adopted in estimating the gross rentals of these families.

Professor Stone commenced his criticisms by repeating the warnings he had given in 1965 about the difficulties of placing families in categories of income for gross rental and went on quite rightly to point out that his groups “cover a very wide income span indeed and the number in each is very small.” He continued his criticism, however, by ascribing to me a procedure quite different from the one I had actually adopted. “By taking the average in the middle of the group as the income for each family, Mr Thompson has exposed himself to the possibility of arriving at a family income which may be very seriously wrong in any given case. For example, Group VII in 1641 spans £1,100 to £2,199. Mr Thompson presumably put any family in that group at £1,600. A family might, however, have had an income of only £1,100, which means Mr Thompson’s figure would be wrong by nearly 50 per cent. This is admittedly an extreme case, but errors of up to plus or minus 20 per cent would be common in the procedure he has adopted, particularly since the few families in a given group may easily bunch at the top or at the bottom.”

It is at once clear that Professor Stone’s account of my procedure for estimating the gross rentals of the 41 and 33 families is completely unfounded. There is no way in which it could have been derived from the description given in my article or from the tables which I had supplied to him. What I had calculated was the mean of each income group for gross rental: what he mistakenly attributed to me was the method for calculating the median of each of these income groups. It is astonishing that Professor Stone should have made such an error. The difference between my actual procedure and his account of it can be illustrated by re-examining the example of a family in Group VII in 1641: I should have divided the total gross rental enjoyed by that group by the number of families in that category to obtain an estimate of mean gross rental: there were, in fact, three such families – Eure, Hastings (Huntingdon) and Wentworth (Cleveland) – in my calculations for the 33 families in 1641 and, in each case, the gross rental was estimated as £7,000 divided by 4 = £1,750. This procedure took direct account of the bunching of incomes towards the top or bottom of an income group for gross rental. I did not suppose when I made these calculations nor do I suppose now that the method tells us what the gross rentals of such families actually were. What it did permit was as accurate an estimate of the total gross rental of the 41 and 33 families to be compiled as the nature of Professor Stone’s original estimates allowed. This was the key point: the figures for the estimated total gross rentals plus casualties for these 41 and 33 families in 1558/9 and 1602 and for the estimated gross rental of the 33 families in 1641 were not challenged at all by Professor Stone in 1972: on the contrary, he considered the figures plausible enough to use as the basis for his own revised figures. His criticism, therefore, of my method of estimating the mean gross rentals of the 41 and 33 families was entirely unfounded and scarcely consistent with an attempt to look very closely at the way in which I arrived at my statistics.

Of course, Professor Stone had a second objection to make over the relationship between manors and gross income. He had argued in The Crisis of the Aristocracy that “up to the Civil War most manors included not only the by now financially almost valueless rights of overlordship over freeholders, but also substantial ancient demesne lands, mills, rights over the waste, and the power to extract fines from the copyholders, often at the lord’s discretion.” Indeed, he gave the further assurance that “non-manorial agricultural property cannot seriously have affected the landed fortunes of most of the nobility, at any rate before the 1620s.” But he was careful to explain how non-manorial property came to play an increasingly important part in the finances of some peerage families by 1641.

These assurances about the source and composition of landed income and his warning about the position by 1641 both had to be taken into account in my calculations. Figures about casualties were accordingly included in my tables for 1558/9 and 1602 on the basis of Professor Stone’s statements. In the light, however, of his admonitions, I wrote that “there are also difficulties to be faced in making calculations for 1641. For it is by no means clear how large a proportion of the estimated casualties at this date is derived from Irish property, industrial profits, and London building rents. Prof. Stone does, however, argue that by 1641, there had been “a marked shift on most estates from large fines to large rents”, so it may not be unreasonable to consider the gross rental for that date as the equivalent of the 1559 and 1602 gross rentals plus casualties.” My reason for excluding casualties (which might have been derived from Irish property, industrial profits, or London building rents) entirely in 1641 was clearly explained: no calculation for casualties appeared in my Table 7 for that date although I did take the precaution of specifying in a footnote that, “if casualties were included, the figures [for the average annual value of a manor held by the 33 and 88 families in 1641] would be £79.09 and £99.39 respectively at 1559 prices.”

Professor Stone’s comments on this point are highly illuminating. “By 1641”, he wrote, “ gross landed income included, to a degree which varied widely from family to family, both non-manorial property and income from specialized sources. I [i.e. Stone] offered examples in which non-manorial property varied from 11 to 20 per cent of the total value. In addition, in some cases the total income included very large revenues from industrial activities or from urban rents. For some families this might amount to as much as a quarter or more of their gross income. For all these reasons not very much confidence can be placed in Mr Thompson’s attempts to relate numbers of manors to gross income for all peers.” It is not clear whether this criticism is meant to apply to my calculations for 1558/9 and 1602 as well as those for 1641. If it is supposed to apply in 1558/9 and 1602 – which seems doubtful – then it represents a change of view from his original position which I could not have divined without telepathic powers. If the objection is meant to apply only to 1641, then it is completely false and unfounded since my reasons for excluding casualties at that date were openly explained. It is, in any case, refuted by Professor Stone’s subsequent (and equally unfounded) claim that “Mr Thompson mysteriously fails to include casualties in Table 7” for 1641.

Professor Stone’s description of the method I had used to estimate the mean gross rentals of the 41and 33 families, his claim that non-manorial sources of income had been improperly included in 1641 and his contradictory assertion that casualties had been mysteriously excluded in 1641, can all be seen to have been fallacious.

Professor Stone’s Revised Figures

Professor Stone nevertheless considered it “just possible” that I had “hit on a method to improve on my [i.e. Stone’s] guesses of total manorial counts and total income.” For the 41 and 33 families about whom most was claimed by him to be known, my figures for the average annual value per manor were taken to be:

Date Average Landed Income per Manor Ditto at 1559 prices

£ £

1559 45 45

1602 82 46

1641 142 65

These results did not seem at all implausible to Professor Stone: the disturbingly large increase in the real value of manors between 1602 and 1641 on which Professor Coleman had commented in 1966 had been, so Professor Stone claimed, “reduced to more modest and now easily credible proportions.” It should be pointed out, however, that my calculation for 1641, which Professor Stone was in the process of adopting, explicitly excluded casualties while Professor Coleman’s calculation included casualties. If casualties were to be included in 1641, the figures for that date should have been £173.21 and £79.09 respectively. The broader problem of the remarkable apparent increase in manorial values between 1602 and 1641 suggested by Professor Stone’s estimates had not, in truth, been resolved.

Professor Stone then invited his readers to make a series of assumptions. First of all, he asked them to consider the data on the manorial counts for the 41 and 33 families as being “as secure as I [i.e. Stone] think they probably are”: in other words, that they were subject to an overall margin of error of plus or minus 10 per cent [which was no different to that for his estimates of the manorial holdings of the entire peerage and, by implication, for those peerage families not amongst the 41 or 33]. Secondly, he suggested that his “allocation of families into income brackets [for gross rentals] is a reasonable approximation to the truth.” His third assumption was “that Mr Thompson’s method of matching manorial holdings to income by using these income brackets does not introduce major errors in the procedure”, a charge that he had himself just attempted to sustain while erroneously describing my procedures on the previous page. Finally, he asked his readers to assume “that Mr Thompson has done his sums correctly.” Since he had received a copy of my tables in July, 1970, he was in an excellent position to point out any errors in them had he wished to do so.

Professor Stone went on to argue that I had certainly demonstrated that there were serious discrepancies between the average landed income per manor for the well documented [41 and 33] families and that for the less well-documented families. He rejected the possibility that there might have been an in-built bias that made the mean incomes per manor of families “with good data consistently higher or lower than those with poor data.” It seemed clear to him that his “rough estimates of total manorial holdings of all families must be somewhat faulty, and need to be adjusted to bring the two means together – in other words, to adjust the mean for the families with poor data to equal the mean for the families with good data.”

It is essential to be clear here. In conceding that there were serious variations in the average landed income per manor of the groups in my Tables, Professor Stone admitted that my criticism was valid. But there was no statistical difference in the reliability of the manorial counts for the entire peerage, for the 41 and 33 families within its ranks, and for the remainder of the peerage: all were subject to a margin of error of plus or minus 10 per cent. The degree of documentation for these families’ manorial holdings was immaterial. There was nothing in my argument to justify a move “to adjust the mean [manorial values] for the families with poor data to equal the mean for the families with good data.” In plain English, this is to assume that the average value of a manor held by peerage families outside the 41 and/or 33 families is identical to the average value of a manor held by them. If one accepts this assumption, then the discrepancies revealed in tables 3 to 7 of my article are ipso facto wished away. Alterations to the number of manors held by the peerage families outside the ranks of the 41 and 33 families are irrelevant to this process: no matter what the number of manors attributed to the rest of the peerage, the adjustment of the mean value for their manorial holdings to that for the 41 and/or 33 families ensures that any divergences between them are removed. The elimination of the discrepancies in value per manor revealed in my Tables on which Professor Stone congratulated himself is the result of simply assuming that they do not exist. It is an intellectually breath-taking manoeuvre. But this statistical contrivance cannot establish the contention that the counting of manors is a reliable method of measuring changes in landownership.

Conclusion

In 1972, I argued that Professor Stone had taken an intellectual leap in the dark by deducing from an apparent demonstration that his manorial samples approximated about a mean value that those of the peerage did so too. When I showed that there were significant variations in the average values of manors held by different groups of peers in 1558/9, 1602 and 1641, variations well outside the margins of error allowed for in his manorial calculations, and even larger regional divergences, Professor Stone responded by misrepresenting the procedures I had followed and by adopting the assumption that the average value of all manors held by the families outside the ranks of the 41 and 33 families in my Tables could be adjusted to equal the value of those held by the 41 or 33 families. One assumption succeeded another. The proposition that manorial counts can be used as a reliable indicator of the peerage’s landed wealth still remains to be demonstrated: simply assuming that they do does not constitute a valid argument.

Christopher Thompson