"Sources of information on diet.
Historical estimates of mean caloric consumption per capita have been derived from several principal sources: national food balance sheets; household consumption surveys; food allotments in hospitals, poor houses, prisons, the armed forces, and other lower-class institutions; food entitlements to widows in wills; and food allotments in noble households, abbeys, and similar wealthy institutions. National food-balance sheets estimate the national supply of food by subtracting from the
national annual production of each crop, allowances for
seed and feed, losses in processing, changes in inventories,
and net exports (positive or negative) to obtain a residual of
grains and vegetables available for consumption. In the case
of meats, the estimates begin with the stock of livestock,
which is turned into an annual flow of meat by using estimates
of the annual slaughter ratio and live weight of each
type of livestock. To estimate the meat available for consumption
it is necessary to estimate the ratio of dressed to
carcass weight, as well as the distribution of dressed weight
among lean meat, fat, and bones (Fogel and Engerman
1974, 2:91-99).
Household surveys are based upon interviews with families
who are asked to recall their diets for a period as short as one day (the previous day) or their average diet over a
period of a week, a month, a year, or an undefined period
designated by their "normal diet." In recent times, such surveys
may be based on a daily record of the food consumed,
which is kept either by a member of the family or by a
professional investigator. Institutional food allowances are
based on food allotments for each class of individuals laid
down as a guide for provisions purchased by the institution
(as in the case of victualing allowances for military organizations
and daily diet schedules adopted in abbeys, noble
households, schools, workhouses, hospitals, and prisons)
as well as descriptions of meals actually served and actual
purchases of food for given numbers of individuals over
particular time periods (Oddy 1970; Appleby 1979; Morell
1983; Dyer 1983). Food entitlements of widows and aged
parents were specified in wills and contracts for maintenance
between parents and children or other heirs (in anticipation
of the surrender of a customary holding to an heir).
Such food entitlements have been analyzed for England,
France, the United States, and other countries at intermittent
dates between the thirteenth century and the present
(Bernard [1969] 1975; Dyer 1983; McMahon 1981; for
some studies of other countries, see Hemardinquer 1970
and Fogel 1986c).
Although these sources of information on the average
consumption of nutrients contain valuable information,
they are also fraught with difficulties. In the case of the
national food-balance sheets, for example, the accuracy of
the estimates depends in the first instance on the accuracy
of the production figures and on the various coefficients
used to transform outputs of grains and stocks of animals
into food available for human consumption. However, even
if the outputs and factors used to produce the national food
supply are accurate, the average amount of nutrients produced
is not necessarily equal to the average amount consumed.
Not only are there storage and food-processing
losses before the supply reaches the household, but there
are losses within the household as well. There is also the
question of the amount of food put on an individual's plate
that is not consumed (plate waste and scraps fed to pets).
Analysis of the estimates of average daily caloric consumption
for recent times in nations for which there exist
both national food-balance sheets (FBS) and household
consumption studies (HCS) not only indicates that the FBS
estimates generally exceed the HCS estimates of caloric
consumption, but also that the gap is positively correlated
with the level of income. As the food distribution system
becomes more complex, it apparently becomes more difficult
to correct FBS estimates for losses as food passes
through the system. On the other hand, it is possible that at
low income levels both FBS and HCS underestimate food
consumption because they do not adequately reflect such
foods as wild nuts, fish, and game (Dowler and Seo 1985;
FAO 1983; cf. Srinivasan 1992).
Household consumption surveys, especially those of past
times, have their own set of problems. They focus largely on lower-class diets and are generally judgment samples.
Hence, it is difficult to know their precise location in the
national distributions of calories and other nutrients. Because
these surveys sometimes include information on the
income of households, it is possible to relate consumption
to the income (or expenditure) of households. Such studies
for English budgets generally indicate an income elasticity
of the demand for food between the 1780s and the mid-
1850s that is at the high end of those found for lessdeveloped
nations today, which is not inconsistent with estimates
of English per capita income for that period. However,
scholars disagree over whether these households were
below or above the middle of the English income distribution
for their period or whether the reported income understates
or overstates the true household income. Although
information on the size distribution of income before World
War I is sparse, that which is available can be used to locate
households in nutrient distributions (Crafts 1981; Woodward
1981; Shammas 1983, 1984, and 1990; Fogel 1987).
Sources of information about food allotments in institutions
and about food entitlements in wills often suffer from
a common problem: lack of information on the age and sex
of the recipients. As table 2 indicates, caloric requirements
vary so significantly by age and sex that failure to standardize
for these characteristics may cause misleading interpretations
of the adequacy of diets, and shifts in the age-sex structure over time may bias the estimated trends in nutrition.
Food wasting varied greatly by institutions so that the
proportion of the food supply actually consumed was much
lower in noble households than in poor households. No
one, for example, could have consumed regularly the daily
allowance at the royal households in Sweden of foods containing
6,400 calories (Heckscher 1954, 21-22, 68-70).
Even allowing for heavy work and cold climate, one-third
to one-half of the allowance must have been wasted in
storage, in preparation, on the plate, or it was dispersed.
These and other problems make it clear that sources of
evidence on nutrient consumption are strewn with pitfalls,
but problematic sources are not unique to energy cost accounting.
As with national income accounting, energy cost
accounting provides a systematic framework for bringing
together the diverse pieces of evidence bearing on energy
intake and output, for examining the consistency of the various
bits with each other, and for making informed judgments
on how best to interpret the available evidence.
Size distributions of calories. Size distributions of caloric
consumption are one of the most potent instruments in assessing
the plausibility of proffered estimates of average
diets. They not only bear on the implications of a given
level of caloric consumption for morbidity and mortality
rates, but they also indicate whether the calories available for work are consistent with the level of agricultural output
and with the distribution of the labor force between agriculture
and nonagriculture (Fogel and Floud 1991; Fogel
1991b). Although national food-balance sheets, such as
those constructed by Toutain ( 1971) for France over the period
1781-1952, provide mean values of per capita caloric
consumption, they do not produce estimates of the size distribution
of calories. In principle it is possible to construct
size distributions of calories from household consumption
surveys. Inasmuch as most of these surveys during the nineteenth
century were focused on the lower classes, in order
to make use of them it is necessary to know from what
centiles of either the national caloric or the national income
distribution the surveyed households were drawn.
Three factors make it possible to estimate the size distributions
of calories from the patchy evidence available to
historians. First, studies covering a wide range of countries
indicate that distributions of calories are well described by
the lognormal distribution. Second, the variation in the distribution
of calories (as measured by the coefficient of variation
[siX] or the Gini [G] ratio) is far more limited than
the distribution of income. In contradistinction to income,
the bottom tail of the caloric distribution is sharply restricted
by the requirement for basal metabolism and the
prevailing death rate. At the top end it is restricted by the
human capacity to use energy and the distribution of body
builds. Consequently, the extent of the inequality of caloric
distributions is pretty well bounded by 0.4 ::::: (siX) 2:: 0.2
(0.22 2:: G 2:: 0.11) (FAO 1977; U.S. National Center for
Health Statistics 1977; Lipton 1983; Aitchison and Brown
1966).
HISTORICAL METHODS
Third, when the mean of the distribution is known, the
coefficient of variation (which together with the mean determines
the distribution) can be estimated from information
in either tail of the distribution. Fortunately, even in
places and periods where little is known about ordinary
people, there is a relative abundance of information about
the rich. Although much remains to be learned about the
ultra poor, much has already been learned about them during
the past quarter century, and such information is also
helpful in resolving the identification problem. However, at
the bottom end, it is demographic information, particularly
the death rate, that rather tightly constrains the proportion
of the population whose average daily consumption of calories
could have been below BMR or baseline maintenance.
The bottom end is also constrained by the requirement that
the energy available to the agricultural labor force is sufficient
to produce the agricultural output.
Table 3 presents three possible size distributions of calories
in France circa 1785 (see the Appendix for procedures
employed in construction of these distributions). They are
all lognormal distributions and are denominated in daily caloric
consumption per consuming unit. They all have the
same mean (2,290 kcal) but differ in their coefficients of
variation. For reasons that will become clear in the following
discussion, distributions B and C are more egalitarian
than the distribution of calories that now exists in the
United States, although the U.S. distribution of calories is
far more egalitarian than that of its income or of the income
of any major nation in the world today (Paukert 1973; Sawyer
1976; Kuznets 1966). The degree of egalitarianism is
measured by the coefficient of variation and the Gini ratio, which are closely related to each other. Distribution A is
the least egalitarian in its distribution of calories, although
with a Gini ratio of 0.22 it is far more egalitarian than the
income distribution of any major nation. Distribution B,
which has about the same coefficient of variation as the
Philippines in 1965 (G = 0.17), is one of the most egalitarian
of the known caloric distributions for less-developed
nations today. Distribution C, with a coefficient of variation
of 0.20 and a Gini ratio of 0.11, is considerably more egalitarian
than any of the national distributions of calories currently
available (FAO 1977; U.S. National Center for
Health Statistics 1977; Lipton 1983).
Consideration of table 3 makes it possible to illustrate
how the patchy evidence can be brought to bear in choosing
which of the three distributions comes closest to representing
the situation in France circa 1785. Even before we consider
its consistency with the available historical evidence,
table 3 yields an important implication, one that is robust
to any plausible assumption about the egalitarianism of the
French calorie distribution on the eve of the French Revolution.
The bottom tenth of French households lacked the
energy to participate regularly in the labor force. Another
robust point is that the average caloric consumption of the
middle classes (fourth through the eighth deciles) is largely
independent of the assumption about the egalitarianism of
the caloric distribution. Their mean consumption is virtually
the same under distributions A and C (2,263 and
2,276) and only slightly higher (2,296) under distribution
B. The assumptions about the degree of egalitarianism in
the diet of the late ancien regime have their cutting edge on
the two highest and the two lowest deciles.
Allowing for adaptation of the two lowest deciles to their
status, which includes absolutely no waste of any food, and
assuming energy balance at exceedingly low body masses
(average body mass index [BMI] = 15),5 an average of
about 1 ,240 calories would still have been required for
BMR and about 1 ,570 calories for long-term survival of
inactive equivalent adults. 6 Consequently, distribution A
not only implies that the poorest 21 percent of French
households had no energy available even for minimal sustained
work, but that the majority of those in the two bottom
deciles were starving to death-about a third of them
quite rapidly because their intake was below basal metabolism.
7 Such high proportions of starvation diets during normal
times is inconsistent with what is known about the condition
of the French lower classes during this period
(Goubert 1973; Jones 1988; Dupaquier 1989).
Distribution C, on the other hand, implies levels of consumption
in the highest decile that are inconsistent with
what is known about the conditions of rentiers as well as of
the nobility and their retainers who made up that decile of
consumers of calories. Not only did the form of food consumption
lead to significant losses of nutrients, but it is also
likely that plate waste was high. Such losses probably reduced
actual consumption by about 10 percent (to about
2,860 kcals per equivalent adult). Thus, distribution C im-
II
plies that France's richest tenth had only enough energy for
about 5 hours of moderate activity per day (Goubert 1973;
Quenouille et al. 1951; FAO/WHO/UNU 1985). 8 Thus, because
of its implications about the diet of the rich, high
egalitarianism is as implausible as low egalitarianism because
of its implications about the diet of the poor.
We are left with moderate egalitarianism (distribution B)
as a plausible assumption. Under that distribution only the
bottom 15 percent of households is entirely unproductive,
and those subject to rapid starvation are less than 2 percent
of the population, a finding consistent with what is known
about mortality rates during the ancien regime (Goubert
1973; Bourgeois-Pi chat 1965; INED 1977; Flinn 1981;
Weir 1984 and 1989; Galloway 1986; Dupaquier 1989).
Similarly, under distribution B, given the same assumptions
about waste made previously, the richest decile of the population
has on average enough energy for nearly 9 hours a
day of moderate activity: about the same amount of energy
available for a moderately active adult male from the prosperous
classes in developed nations today (Lipton 1983;
U.S. National Center for Health Statistics 1977).
The French distribution is also consistent with what is
known about the English consumption of calories circa
1790. Table 4 presents the probable English distribution,
which was based on estimates of mean consumption derived
from household budget surveys, and compares it with
the French distribution B. The mean consumption of calories
was about 18 percent higher in England than in France,
which is consistent with recent estimates of the relative productivity
of agriculture in the two countries (O'Brien and
Keyder 1978; Wrigley 1987b; Hoffman 1988; Grantham
1992; Allen 1991). Moreover, the average levels are not out
of keeping with recent experiences in the less-developed
nations. Low as it is, Toutain's estimate of the French supply
of calories is above the average supply of calories in
1965 estimated for such nations as Pakistan, Rwanda, and
Algeria, and only slightly less (39 calories) than that of
Indonesia. The English estimate is above that for thirty
less-developed nations in 1965-including China, Bolivia,
the Philippines, and Honduras-and only slightly below
(37 calories) India (World Bank 1987).
The distributional implications of the two estimates are
consistent with both qualitative and quantitative descriptions
of the diets of various social classes (Hufton 1974,
1983; Goubert 1973; L. Tilly 1971; C. Tilly 1975; Frijhoff
and Julia 1979; Blum 1978; Cole and Postgate [1938] 1956;
Rose 1971; Drummond and Wilbraham 1958; Pullar 1970;
Wilson 1973; Burnett 1979; Mennell 1985). For example,
Bernard's study (l1969] 1975) of marriage contracts made
in the Gevaudan during the third quarter of the eighteenth
century revealed that the average ration provided for parents
in complete pensions contained about 1,674 calories. Because
the average age of a male parent at the marriage of
his first surviving child was about 59, the preceding figure
implies a diet of about 2,146 calories per consuming unit
(Fogel 1987). That figure falls at the 47th centi1e of the estimated French distribution, which is quite consistent
with the class of peasants described by Bernard.
So far I have assumed that Toutain's estimate of the mean
consumption of calories circa 1785 is acceptable in the
sense that its distributional implications are not implausible.
However, other estimates of mean French caloric consumption
circa 1785 may also be consistent with moderate
egalitarianism. Some of Toutain's critics have argued that
his estimate for circa 1785 is too low because it neglects or
underestimates the caloric contribution of wild foods and
wine; others believe this estimate is too high because it
overestimates production of grain or other nutrients (Le
Roy Ladurie 1979; Jones 1988; Bekaert 1991; Grantham
1992). The range of differences suggested by critics is on
the order of plus or minus 200 kcal per capita (plus or minus
260 kcal per consuming unit).
The plausibility of the outer limits of this range of potential
error can be evaluated by looking at the consistency
between the estimates of agricultural output, of labor productivity
in agriculture, and of the energy available to produce
that output. Because the procedures for such an evaluation
have been described at length elsewhere (Fogel and
Floud 1991; Fogel 1991 b) and because of space limitations,
only a brief description can be provided here. The procedure
turns on the proposition that human populations can
alter their basal metabolic rate by adjusting their body
sizes. In this way the share of available dietary energy required
for maintenance can be reduced and that available
for work can be increased. In the short run, such adjustments
can only be made by varying body mass. In the long
run, height can also be varied. 9
An implication of this proposition is that the aggregate
amount of energy available for work during a given year
depends on the aggregate supply of dietary energy available
in that year and the aggregate amount of dietary energy
required for maintenance, which is a function of the average
equivalent adult stature and BMI during that year. Hence,
when average height and BMI are known, it is possible to
determine whether the energy available for work by the agricultural
labor force is adequate to produce the estimated
dietary output (measured in kcal). One can obtain the output
of calories permitted by energy available for work by
multiplying the energy that farmers expend on their tasks
by a measure of the productivity of agricultural labor,
which in this case is the "caloric productivity" ratio. When
this procedure is applied to the lower bound (2,030 kcal per
consuming unit) suggested by critics of Toutain, not enough
energy is available beyond maintenance to produce 2,030
kcal per consuming unit, even though it is assumed that
French agricultural labor was as productive as English agricultural
labor. 10
When the same test is applied to the upper bound (2,550
kcal per consuming unit), one obtains the odd result that
the energy available for work should have yielded far more
dietary energy than was produced by the French. The result
is a consequence of having assumed that the French were
as efficient as the English in the production of dietary
energy. Thus, it is necessary to ask the following: If 2,550
kcal is the correct mean level of French dietary production,
how much less efficient than English agricultural labor
would French agricultural labor have had to have been in
order to reduce the implied level of energy production to
2,550 kcal per consuming unit? The answer is that French
agricultural laborers must have been less than one-quarter
as productive as their English counterparts. This figure is
too far below current estimates of relative productivity in
the two countries to be acceptable.
It is interesting that Toutain's mean-2,290 kcal per consuming unit-also implies that the energy available for
work was excessive, when it is assumed that French and
English agriculture had the same outpuUinput ratios of dietary
energy. In this case, however, the available energy
input becomes consistent with the estimated energy output
when it is assumed that the French were about half as productive
as the English, a figure that is in accord with current
estimates of the relative levels of productivity in the agricultural
sectors of the two countries (Marineau 1971;
O'Brien and Keyder 1978, 1979; Crouzet 1985; Wrigley
1987b; Hoffman 1988, 1991; Allen 1988, 1993; Grantham
1989, 1990, 1991; cf. Cameron and Freedman 1983;
Allen and O'Grada 1988).
Why is it that Toutain 's estimate of mean consumption
circa 1785 fits with the productivity constraint so well, despite
his acknowledgment that wines and cider omitted
from his estimate may have accounted for 10 percent of
consumption? Part of the answer is offsetting errors. His
figure refers not to calories ingested but to calories produced
for human consumption. It does not contain adequate
allowances for losses in inventory, in processing, during
distribution, or for plate waste, all of which may have run
in the neighborhood of 15 percent of output ( cf. McCloskey
and Nash 1984). Consequently, if 2,694 kcal per consuming
unit were available for consumption at the point of production
(and immediately after the harvest) an average of
2,290 would have been consumed. The difference between
these numbers is 404 kcals per consuming unit, enough to
cover omitted wine and still leave over 140 kcal for wild
foods. In other words, even though it underestimates production
by 15 percent, 2,290 kcal could be an appropriate
estimate of calories consumed.
The preceding discussion of Toutain's estimate is meant
to illustrate the procedures and constraints that can be
brought into play when evaluating contending estimates of
mean caloric consumption and is not meant to prematurely
close off debate about these estimates. I have tried to stress
the need to drop the often implicit assumption that all individuals
consume the mean level of caloric intake, and to
consider the implications of the tails of caloric distributions
when assessing proposed means. In this connection special
attention should be paid to the implications for mortality
rates as well as the consistency between the calories available
for work and implied level of labor productivity. Such
implications need to be tested against independent estimates
of mortality rates and of labor productivity, especially
in agriculture."
New Sources and New Techniques for the Study
of Secular Trends in Nutritional Status, Health,
Mortality, and the Process of Aging
Robert William Fogel
To cite this article: Robert William Fogel (1993) New Sources and New Techniques for the
Study of Secular Trends in Nutritional Status, Health, Mortality, and the Process of Aging,
Historical Methods: A Journal of Quantitative and Interdisciplinary History, 26:1, 5-43, DOI:
10.1080/01615440.1993.10594215
To link to this article: http://dx.doi.org/10.1080/01615440.1993.10594215
Articles
Aucun commentaire:
Enregistrer un commentaire