- 1 . Popkin, B. M., 1993 ; Crigg, D., 1995
1One feature of modern nutrition transition is the growing consumption of animal proteins. Of particular note among these is the contribution of foodstuffs such as meat and milk. In Western Europe this process went with changes in the organization of agricultural and livestock farming, as well as in consumption preferences, under the momentum of industrialization and urbanization1. Quantitative study of consumption patterns provides a basic approach to illustrating changes in nutritional status.
2The study of a nutritional change of this nature, however, comes up against a significant limitation. An accurate description of this process requires knowledge not only of average consumption, but also of the number of consumers. If a dietary change is considered in terms of a diffusion process, it is therefore essential to take both parameters into account. This second parameter – number of consumers – is only rarely estimated despite its obvious relevance. Historical statistics permit the calculation of a mean consumption but not of the number of consumers.
- 2 . Steckel, R. & Floud, R., 1997, p. 1-16 review all these effects.
3One of the major consequences is the lack of information about distribution patterns of food consumption and, therefore, of the population’s nutritional status. The hypothesis and the associated evidence that in the initial stages of industrialization and urbanization nutritional status was unequally distributed have been subject to intensive research. One method of approaching these difficulties has been anthropometry: the pattern of height distribution might reflect the underlying effects of food consumption levels and other environmental factors2.
- 3 . Fogel, R, 1992.
- 4 . Here population denotes the total number of inhabitants living in an area. Population numbers in (...)
- 5 . A historical perspective on these changes in the demand of goods and the behavior of consumers ca (...)
4The need of distributional assumptions to study the nutritional of a population is well shown3. When individual data is not available – the most common in historical studies –, then the basic indicators stem from aggregate statistics and are used to compute average levels of consumption. In these cases major differences in indicators are a consequence of the geographical information provided by published sources. Thus, those levels of consumption can reflect averages from local to national level. Beyond all these spatial scales, however, the averages suffer the same flags: if a particular foodstuff does not have a pattern of universal consumption, the average obtained from the tabulated data cannot be used to accurately estimate the ‘true’ level of consumption. This is because the total average (x̄) is calculated as the total output consumed on the total population (O/P).when the number of consumers is lower than the total population it can be seen that statistical x̄ < x̄- ‘true’ value4. A simple and hypothetical example may illustrate this problem. There are two countries, A and B, with the same population, one million inhabitants, but different consumptions levels. In country A, consumption stands at around 70kg per person per year and in country B, 130kg. If the proportion of consumers is 50 per cent of the total population in country A and 100 per cent in B, i.e. universal consumption, it is clear that the effective consumption level would be greater in country A (around 140kg) than in country B (130kg). As the historical process shows, changes in consumption involve not only an increase in the consumption per capita of certain products, but also a growing number of new consumers5. Because of these two dimensions researchers must be very cautious when comparing patterns of food consumption between different countries. In fact, without first applying a « standardization » (or statistical adjustment) in some basic indicators such as the mean, any conclusion about differences in consumption levels will be biased by differences in the number of consumers.
5The relationship between the mean and its underlying distribution is a common topic in textbooks on statistics. Measures of central tendency and dispersion are related but not in a univocal way. It is well known that the same mean can be the result of different distributions, and testing differences between means involves certain assumptions about distributional patterns.
- 6 . This is nothing new in historical research, where exploring relationships between mean and distri (...)
6This article deals with this problem6. Despite the simple way the problem can be set out, its solution is far from easy. This article does not provide a theoretical solution to this issue; only an empirical one: a new methodological approach is developed under some assumptions to estimate the consumers’ population. Clearly, the consumers’ population differ according to the analysed foodstuff. In this paper, the estimation methodology will be tested on the specific case of milk consumption in Sp s a form of animal protein that can be consumed directly through various products containing fresh milk or can be transformed into cheese, butter, etc. b) As has been noted previously, changes in milk consumption belong to the central decades of the Spanish nutritional transition. c) Basic data are provided by consumption statistics from official sources published in relation to agriculture and livestock activities; this is not always the case with other foods because the available data were previously published only as production statistics.
7As it has been indicated in the previous section, the problem and proposed solution will be illustrated by the evolution of milk consumption in Spain between 1925 and 1981. The choice of this food can be justified by three factors: a) Milk is one of the foodstuffs acting as « indicators » of nutritional transition because it is a form of animal protein that can be consumed directly through various products containing fresh milk or can be transformed into cheese, butter, etc. b) As has been noted previously, changes in milk consumption belong to the central decades of the Spanish nutritional transition. c) Basic data are provided by consumption statistics from official sources published in relation to agriculture and livestock activities; this is not always the case with other foods because the available data were previously published only as production statistics.
- 7 . The Statistical Yearbooks, published regularly in the twentieth century since 1912, also record d (...)
- 8 . Aggr, 1925. This involves a publication that is normally dated to around 1923. However, according (...)
- 9 . Ministerio de Agricultura, 1934.
- 10 . Ministerio de Agricultura, 1954, 1955.
- 11 . Ine 1969, 1983.
- 12 . Provincial data on provincial production and consumption of milk were published in the first year (...)
8During the studied period here an average of around 80 per cent of the total amount of milk produced in Spain was consumed by the population as fresh milk, and 9 per cent was transformed into cheese. The main type of fresh milk consumed was cow milk. Accordingly, the evolution of the percentage of this type of milk can be estimated at between 79 and 91 per cent of the total amount consumed between 1925 and 1981. The remainder consisted of goat milk. Table 1 shows the mean consumption levels per year, and the associated standard deviation, calculated using the provincial data and the mean consumption levels calculated for the entire Spanish population between 1925 and 1981. These data are derived from two statistical sources: livestock statistics and household budget surveys7. The first kind of sources provides data on the number and type of livestock as well as the total production and milk consumption levels for each province, while the second offers data on the consumed quantities and family expenses. In short, the first group of sources pertain, in chronological order, to the statistics published by the Asociación General de Ganaderos del Reino (General Association of Livestock Owners of the Kingdom) in 19258, the Censo de la ganadería en España (Livestock Census in Spain) with data for 19339 and the new series on direct human consumption of animal milk published by the Ministry of Agriculture10 in the Resúmenes estadísticos de la producción, destino y valor de la leche (Statistical summary of the production, destination and value of milk) from 1954 and 1955. About the second group, the data from 1965 and 1981 on the consumption per person per year at provincial level correspond to those published in the « Household Budget Surveys »11. There is clearly a significant lack of information between 1933 and 1965. This can be explained by the Spanish Civil War (1936-39) and the post-war period that lasted until the mid 1950s. The rationing imposed on the population during this period quite logically made statistics on consumption unnecessary12. With the exception of the estimates for provincial consumption per capita provided directly by the results from the Household Budget Surveys (hbs), the remaining values have been calculated using the consumption data provided by the sources and the provincial population corresponding to that year. The provincial populations are obtained by linear interpolation between the two chronologically closer censuses. In all cases data have been used at a provincial scale because this is the way they were collected.
9Table 1 distinguishes between two features of the long term evolution of milk consumption; the increase in average levels of consumption, and the reduction in standard deviation values. The provincial differences become more noticeable in what seems to be a strong period of growth between 1925 and 1933 with an increase in the standard deviation from 38.80 to 70.80. Following this it progressively decreases towards what seem to be improvements in the mean consumption levels, together with a major universalization of these.
Table 1. Milk consumption in Spain (1925-1981) (Liters per person per year)
(1) Arithmetic mean estimated from per capita consumption for each Spanish province.
Source. agr, 1925; Ministerio de Agricultura, 1934; Ministerio de Agricultura, 1934; Ministerio de Agricultura, 1954 and 1955,
ine (Instituto Nacional de Estadística), 1969; ine, 1981.
- 13 . Williamson, J. G, 1965.
10Given that the basic data are provincial, Table 2 offers a panoramic view of the evolution of the distribution of consumption at this scale between 1925 and 1981, in accordance with some intervals of consumption, and highlights the values of the maximum and minimum consumption levels at each point and the name of the provinces to which they correspond. In some Spanish provinces the minimum consumption levels were always below 20 liters between 1925 and 1965, and in the years 1925 to 1933 they were always below 10 liters, in the same way that between these two dates a considerable proportion of the provinces had constant levels of consumption that were under 25 to 50 liters per person per year. The data reflect how the distance between minimum and maximum consumption increased between 1925 and 1933 and then went on to decrease from 1955 onwards. This distance reveals differences in the consumption magnitudes of this foodstuff, but also in habits of consumption. Moreover, from a territorial point of view, which cannot be studied in detail here, these differences between maximum and minimum consumption levels are very stable. As we can see in Table 2 some provinces reappear throughout these years at both extremes of the distribution. All of this could be interpreted as evidence for a process of diffusion of consumption in which an initial stage of low and homogenous average consumption levels is followed by a stage of expansion, with a growing territorial disparity, studied the process ends with one of the highest average levels of consumption and again homogenous. This represents a typical inverted U shaped trajectory, highlighted by a stage of expansion, accompanied by a growing territorial disparity culminating in one of the highest average levels of consumption and again homogenous13.
Table 2. Distribution of Spanish provinces according to consumption intervals (Liters per person per year)
Source. Statistics cited in Table 1.
11All these data serve to illustrate some of the limitations in average estimates. In some Spanish provinces in the 1930s the mean consumption was around 5 liters per person (Table 2); quite a low value. Obviously this reflects low consumption standards but it is most probably a consequence of there being a high proportion of population who were not consuming this food. In fact, a contemporary observer wrote: « Thousands of Spanish households don’t usually have milk, with the exception of sick people »14. Thus, the comparison of averages of milk consumption between Spanish provinces, or within the same province but in two different periods of time, may lead to erroneous conclusions if the consumer population is not taken into account.
12Other aspects of average estimates should also be considered. On this point it might be useful to distinguish between two parameters: the quantities consumed, and their frequency. In terms of the first parameter the most reasonable thing would be to think of some magnitudes that could be expressed according to the fractions of a basic and viable unit of consumption. With regard to the second parameter, it would be fair to accept that in the case of this foodstuff consumption could be regular, daily, occasional or lacking entirely. The evidence from the data suggests that these elements should be taken into account in any analysis. For example, if the mean annual consumption per inhabitant in a province in 1925 is 4 liters (the province of Cuenca in Table 2) this would imply 11 ml a day, which is an unrealistic quantity in practical terms. It seems unlikely that the consumption of a product, in a time when prolonged conservation was not possible, would have occurred using utensils other than cups or glasses, the normal size of which, due to their own design, exceeds such quantities. These modalities of consumption should be expressed in equivalences, or fractions, of a bowl of milk (250 ml); for example, a half or a quarter of its size. This way, this level of provincial consumption would only make sense if it meant expressing a non-daily consumption frequency; if, for example, it were once a month, the quantity would be closer to 250 ml. Alternatively, as reasoned in the previous argument, such a figure could indicate that the proportion of the non-consumer population was very high. It is obvious that, at the other extreme of consumption levels, the provinces with over 150 liters per person per year would involve a daily consumption close to half a liter of milk. Behind this value is likely to be an almost universal pattern of consumption in the entire provincial population.
13The previous discussion and the problem to be resolved can be expressed formally:
14The problem is set out in order to adjust specific average levels of observed consumption, computed from published statistics (x̄Effective-Consumption), which could even be rather improbable, to other more realistic ones; the effective consumption levels (x̄Effective-Consumption). In short, this would mean transforming the likely difference between these two means into an equality as a consequence of the introduction of a correction factor (fc) into the average consumption level obtained from the statistics, where such a factor would be equivalent to the proportion of the consumer population15. Logically, if the total population is also the consumer population, this factor would be equivalent to the unit. This relationship between consumption levels and proportion of the existing consumer population for the total Spanish population would be applied in the same way to each « j » province, because at this scale the disparity x̄C-Observed(j) < x̄C-Effective(j) would also be met and the « correction » or adjustment of this could be obtained by estimating the proportion of the consumer population for each of the provinces.
- 16 . From expression (1) if the total output consumed and the effective consumption per capita are kno (...)
15This equation also illustrates the main difficulty to be dealt with the presence of two unknowns. If an estimate of the effective consumption were available from equation (1), calculating the consumer population would be straightforward16. The adjustment of observed consumption to effective consumption seems to be an arbitrary solution if it is taken without reference to a particular foodstuff. It requires some assumptions, and data on patterns of distribution and diffusion of consumption in the population. In this case the evolution of milk consumption in Spain from 1925 to 1981 will be used to illustrate this approach and methodology.
- 17 . A revision of various applications in economic studies in R. Lawrence, 1988. In the specific case (...)
16In the first place, it seems necessary to have a statistical model for the distribution of consumption that would permit us to infer what proportion of the population would consume particular quantities. On this point, the use of log-normal distribution would be the most appropriate because it has been used for modelling various economic activities, including the consumption of different types of products or services17. The distribution of the consumer population of this foodstuff could thus be estimated using the following density function applied to the Spanish provincial statistics data:
- 18 . In log-normal distribution the median is included between the mode and the mean, closer to the fi (...)
17Where f(xj) corresponds to the total population consuming a given quantity of milk on a daily basis in province « j », xj corresponds to the average of consumption of each « j » province and the two parameters « m » and « σ » make reference to the averages and standard deviation of the distribution obtained from all the provincial values. The f(xj) values describe a theoretical distribution of consumers for the total Spanish population according to the mean and variance provincial values. One of the characteristics shape of this distribution is its asymmetric nature, deviated towards the left: as a consequence the mean, mode and median do not present identical values18. This property would be of greater interest if a study of the temporal evaluation of consumption habits were to be considered, because its leaves open to consideration the fact that the variations in mean or mode do not have to be of the same magnitude and therefore follow similar trajectories.
- 19 . Remember that a random variable follows the log-normal distribution if its logarithm follows the (...)
- 20 . D(n) in the Kolmogorov-Smirnov-Lilliefor and W in the Shapiro-Wilk tests measure the discrepancy (...)
18It would of course be better to verify that the statistical model adopted for reconstructing the distribution of the consumption of this foodstuff were the most appropriate option. Due to the lack of individual data the only confirmation can be provided by the provincial values. Table 3 displays the results of the two statistics that test the hypothesis where the distribution of the logarithms of the aforementioned provincial consumption averages adjusts itself to the normal distribution19. In any event, as the significance levels show, the null hypothesis cannot be rejected20. It thus seems reasonable to suppose that the consumption of milk follows this model of distribution.
Table 3. Distribution of average provincial consumption of milk, in Spain (1925-1981). Normality Tests
Source. Statistics cited in Table 1.
- 21 . Rogers, E. M., 2003, p. 279-283.
19Moreover, knowledge about patterns of consumption cannot be approached without being considered, as it has been advised, as a process of diffusion. From a theoretical perspective, the most basic representation of a process of diffusion is that in which the people adopting it follow a normal distribution pattern where the different categories of these people are located21. Therefore, 2.5 per cent of the innovators and 13.5 per cent of the early adopters out of the population would have become integrated at an early stage. In this phase the rest of the population would either have practiced sporadic consumption or they would not have consumed milk at all. In the following stage 34 per cent of the denominated early majority and 34 per cent of the late majority of adopters would have joined. Consumption as well as its frequency would have become generalized, and the distance between those sectors consuming more and those consuming less would have increased. In the final stage, 16 per cent of late adopters would have joined, at which point the generalization of consumption would have been accompanied by higher average levels and a greater uniformity in consumption habits.
- 22 . Nicolau-Nos, R., Pujol-Andreu, J. & Hernández, I., 2010.
- 23 . An overview of the regional patterns of Spanish food consumption in the first third of the twenti (...)
20The adoption of milk as a foodstuff for daily use in Spain, according to the quantitative and qualitative information available, should be understood as a process with these characteristics. This is a foodstuff that for reasons of technology, health and food culture at the end of the nineteenth and beginning of the twentieth century was an object of restricted consumption, often related to therapeutic purposes; it went on gradually to form part of the diet of particular social groups and sections of the population, for example, children and young people22. But milk was also used in the preparation of certain dishes (such as desserts and pastry) in pastoral regions with high levels of milk production, mainly in the North of Spain23.
- 24 . These surveys, known as medical topographies or medical geographies and submitted to the Royal Ac (...)
- 25 . For example, of 45 surveys conducted in Catalonia from 1798 to 1907, only 7 gave quantitative inf (...)
- 26 . References to different amounts of milk consumed in Spanish health centers and hospitals prior to (...)
- 27 . Documentos y trabajos de la comisión consultiva para la transformación del impuesto de consumos, (...)
- 28 . Jiménez, F. & Jiménez, M., 1934.
21It is beyond the scope of this article to introduce a detailed account of this process of diffusion. Unfortunately, sources of data are scarce. For example, medical surveys24 conducted by doctors in many Spanish localities in the nineteenth and first half of the twentieth centuries give more qualitative than quantitative information about nutritional habits and conditions of the inhabitants25. More quantitative sources are available in hospitals, but this kind of data has not always been published26, or can be found only in some special surveys. Two documents can be mentioned here. One is the report written by The Commission on the Reform of Consumption Tax in 190827 (Table 4) and the other is a piece of research on the nutritional status of the population living in rural areas of the province of Jaen (in the region of Andalusia) in the first years of the 1930s, carried out by the physician F. Jiménez and the pharmacist M. Jiménez28.
- 29 . Cussó estimates a consumption of 36 liters per person per year in 1865 (Cussó, X., 2001, p. 322 ; (...)
22The report from the 1908 Commission surveyed only 79 families (434 individuals) and this document clearly cannot be considered as representative of the entire Spanish population of the time; however it does offer two types of results that are of interest for our topic. On the one hand, an average level of consumption per annual capita of 42 liters would not be too far above the averages known for the beginning of the twentieth century, estimated at around 35 or 45 liters (Table 1)29. On the other hand, this permits an estimate of the social differences in consumption, depending on the various levels of income of the breadwinners and, as a result, it illustrates the different modalities of consumption. An aspect of particular interest is the use disparity between the annual (and daily) milk consumption averages for social groups of higher and lower income. In the case of the first group this would mean taking around the equivalent of a quarter of a liter a day, while for the rest of the groups these daily quantities would be so meager that they could be interpreted as evidence of irregular or occasional consumption. Additional information in the report shows that the high income group was composed of entrepreneurs, landowners and stockholders, whereas civil servants, service workers, liberal professions and other middle class occupations were surveyed as a middle income group.
23The research on the nutritional status of the population of the rural area of the province of Jaen carried out in the first years of the 1930s surveyed 3,592 persons older than 8 years. Data were collected from 70 out of 98 villages of the province. Detailed information on daily consumption of every foodstuff was even collected at household level but unfortunately the estimates were not published. However, the abstract with the main results show the proportion of daily consumers of certain foods such as meat, eggs and milk. Around 97 per cent of the families from the high social class were consuming milk daily whereas this proportion was 63 per cent for the middle class families and 9 per cent for the working class. According to the aforementioned around 1908 consumption data and those available for the rest of the first third of the twentieth century, there would exist social groups with daily consumption of that foodstuff.
Table 4. Distribution of milk consumption according to the breadwinner’s income level circa 1908
Notes. Mean Income level per family (pesetas 1908):
Low: 1,838; Intermediate: 6004; High: 25,781.
Source. « Documentos y trabajos de la comisión consultiva para la transformación del impuesto de consumos » (Documents and studies from the consultative commission for the transformation of consumption tax), tome iv, Madrid, 1910, p. 198.
- 30 . Ministerio de Trabajo Comercio e Industria, 1922, vol 4, p. 421-424.
24The idea that, within the entire Spanish population, these groups of new consumers could have represented about 16 per cent of the population – the proportion of adopters in the basic diffusion model – seems plausible in the mid 1920s in the light of the 1920 census data30. With some 21,338,381 inhabitants, this would mean accepting that close to three and a half million people would be daily consumers. An indirect clue for calibrating this is the classification of the population by profession. Despite the corresponding problems of interpretation, the total amount of registered employers in agriculture, industry and trade, plus those linked to the administration, liberal professions and rentiers amounts to almost two million individuals (1,921,917 persons). These groups have a social profile close to that of the high income group surveyed in the Commission report of 1908. Around 70 per cent were enumerated as married. If each one is the breadwinner of a household, when they are multiplied by the number of members, for example 3 or 3.5, the percentage of potential daily consumers fluctuates between 18 and 22 per cent.
- 31 . These statistics do not include consumption of other related foods such as cheese, butter, etc., (...)
25The first Household Budget Survey for the period 1964-1965 collected the most complete evidence for consumption patterns of the Spanish population in the twentieth century. Here it is collated detailed information about the quantities of milk consumed in various groups from within the population31. The data tabulated per socioeconomic category of the main breadwinner of the household allows for estimations (considering averages only; see Table 5) a rang between 120 ml for farm workers and 365 ml for white collar workers; this is between the equivalent of half a cup, and a cup and a half of milk a day. These values could be considered as evidence of quite a complete process of diffusion of this foodstuff because 8 out of 11 social groups were having 250 ml or more per person per day. Only one group still remained at the low consumption levels of 120 ml per day, the result of either irregular consumption or a remarkable proportion of non-consumers.
Table 5. Household Budget Surveys in Spain, 1964-1965. Levels of milk consumption per socioeconomic category of the main breadwinner
Source. Household Buget Survey 1964-1965; ine, 1969.
- 32 . Data from fao statistics published in 1953, cited in C. Agenjo, 1957, p. 315 indicate that in the (...)
26Despite the scarcity of data available from the beginning of the twentieth century by comparison with the first Household Budget Survey from 1964-1965, the best evidence available is that the diffusion of milk consumption in the Spanish population was based, on the one hand, on the existence of consolidated groups of regular consumers and, on the other hand, on the increase in the quantities consumed and the number of consumers. It is important to note that initial high consumption standards seem to converge to similar levels of between 100 and 130 liters per person per year. In this case, these values fit with the historical statistics of high consuming countries or, even for the most recent periods, with the information collected from surveys on consumption. Here all the data available seem to suggest that with consumption above 350 ml per day per person (130 liters per person per year), over 90 per cent of the population should have been consuming (always in terms of mean values)32. At the other extreme, data on minimum consumption do not converge to similar levels. This result could be expected because, as has been previously observed, these means compute not only irregular – non-daily – consumption but also total lack of consumption. This last observation suggests that the search for a reasonable level of minimum consumption, under a general pattern of diffusion would be more the result of a simulation than of data collection. A simulation based on modelling consumption patterns through a log-normal distribution could therefore be put forward. The results should justify the final choice of a minimum average level of consumption related to a critical mass of daily consumers.
27Figure 1 presents the different levels of daily consumption that correspond to the first, second (median) and third quartiles of consumption distributions. These have been calculated using some levels that would situate the average consumption per person per year within 25, 35, 45, and 55 liters, and magnitudes of standard deviation of 20, 40 and 60 liters, which cover a large part of the provincial variations in the observable consumption patterns of the twentieth century (see Table 1). Table 6a and Table 6b show estimates of the proportions of consumer populations having more than or equal to 250 ml and 125 ml a day according to the same values of average consumption and standard deviation as Figure 1.
Figure 1. Proportion of the consumer population in quartiles according to average dispersion levels of consumption of milk per person per year
Source. Own elaboration.
28At the lowest levels of consumption, 25 and 35 liters per person per year (Figure 1), distributions would not have reached average daily quantities which can be expressed as fractions of 250 ml, the equivalent of one cup a day (marking the limits with some discontinuity between 125 and 250 ml). These distributions suggest, at the very best sporadic consumption. According to the levels that correspond to the higher quartiles, we could talk about a weekly amount of something over three quarters of a liter, which would not actually add up to more than three cups. Under these conditions there would be no regular or stable consumer population. Table 6a and Table 6b confirm these appreciations. Estimates of the proportion of consumers with 25 liters on average show how, at that low level of consumption, a critical mass of daily consumers might hardly exist: fewer than 4 per cent (see mean values) would be consuming 250 ml or more a day and 13 per cent 125 ml or more a day. With 35 liters, patterns of consumption are improving but people consuming more than 125 ml a day still represent a mean lower than 25 per cent. The coexistence between occasional and other consumption patterns might be better reflected between 45 and 55 liters of average consumption per person per year. In both cases, the higher quartiles would have included the consumer populations of 125 ml and above.
29Tables 6 show how, with 45 liters and 55 liters per person per year, between one third and half of the consumer population would have 125 ml or more a day and between 10 and 16% 250 ml or more. These values are close to the theoretical percentage of innovators and early adopters mentioned before. This group of results illustrates how the gradual increase in consumption levels per person per year would have led to distributions in which the upper quartiles, given the daily levels that they would have reached, would have facilitated daily habits of milk consumption.
30The confirmation of this trend can now be verified by the distributions associated with levels of consumption higher than those used: for example, from 65 to 75 liters per person per year (Figure 2). In both cases, at least 50 per cent of the population could have been drinking 125 ml or more of milk a day, as is confirmed by the estimates in Tables 6. In the case of 75 liters this number of consumers almost achieves 80%.
Table 6a. Proportion of consumers of 250 ml or more a day according to average and dispersion levels of consumption of milk per person per year
Proportion of consumers of 125 ml or more a day according to average and dispersion levels of consumption of milk per person per year
Source. Own elaboration.
Figure 2. Proportion of the consumer population in quartiles according to average and dispersion levels of consumption of milk per person per year
Source. Own elaboration.
31In terms of maximum levels of consumption, a guarantee of the generalized diffusion of consumption, the simulation has been run with mean values of 130 liters per person per year (approximately the average consumption of the Spanish population in 1981). The results displayed in Figure 2 show how, under the same magnitudes of standard deviation as in previous cases, practically the entire population would have consumed a quarter of a liter of milk per day of this product. The estimates in Table 6a and Table 6b illustrate this observation with values close to 90 per cent.
32The results of this simulation exercise suggest that, assuming the existence of a small group of early and innovator consumers in the Spanish population, the scale of minimum to maximum consumption might span 45 to 130 liters per person per year. 45 liters has been chosen as the minimum value because, according to the previous results and the scarce historical information available, it represents a sort of threshold between a first stage of reduced number of daily consumers and a second one of a greater proportion which seems to fit better the likely consumption patterns of milk in Spain during the first decades of the twentieth century.
33In view of all the results and observations put forward so far, the methodological strategy to be followed would be the one represented in Figure 3. This methodological strategy is based on either displacing the initial levels of consumption upwards or applying a direct correction to the observed data. More specifically, the basic characteristics of this methodology are the following.
34a) The scale of consumption variation is delimited within that of the proportion of the associated consumer population, between a minimum of 125 ml and a maximum of 350 ml a day or their equivalents in terms of average levels of 45 or 130 liters per person per year (Appendix 1). Adopting this minimum value means accepting, first, that consumption of this foodstuff was in discrete units equivalent to a cup (250 ml) or half a cup (125 ml); second, there was a group of stable, daily, consumers of this foodstuff in all the Spanish territory and so in each of its provinces. In terms of the maximum value, it is understood that there would not be as much of an increase in the number of consumers as in the quantities consumed when the consumer population was close to 100 per cent.
- 33 . Data on the population that does not consume certain foodstuffs, in the specific case of milk con (...)
35b) The populations do not reach universal consumption of this foodstuff. For reasons of consumer preference, taste, or digestive diseases a percentage of the population would have been impeded from drinking milk on a regular basis (here this would be equal or less than 5 per cent)33.
36c) As a consequence of the same process of diffusion, the relationship between average levels of consumption and proportion of the consumer population varies over time. Therefore, as shown in Figure 3, with the increase in average levels of consumption of the total number of inhabitants not only would the distance between the minimum and maximum levels decrease, but its slope would change, tending towards a more horizontal position as an effect of the growing homogeneity of consumption habits.
Figure 3. Relationship between daily consumption and the percentage of the consumer population
37This approach allows for a transformation of the observed data into the effective levels of consumption to be made for each province and from here a process follows that ends in the final estimate of the magnitude of consumer populations, at provincial as well as a national scale. The stages required by the entire procedure will be illustrated step by step with the reconstruction of the consumer population in 1925 (Tables Appendix 2a, 2b and 2c); other tables will present summary statistics and indicators for all years.
- 34 . The averages of consumption for each province are set in the new scale through the use of linear (...)
38Stage 1. The observed consumption average is identified with one that falls within the scale previously mentioned, between 125 and 350 ml (see Appendix 1 with the table of equivalences)34. For example: as shown in Appendix 2.a, provinces with under 5 liters per person per year in column (2) were replaced by 45 liters in column (3), which is the minimum consumption adopted according to the discussion on simulations in the previous section. At the opposite extreme of the consumption scale, those provinces with computed milk consumption in excess of 130 liters per person per year remain practically unchanged; they see only a slight increase in order to introduce the assumption that there is no universal consumption (100 per cent of the provincial population consuming this product).
39Of course, the average level of consumption and its standard deviation from the basic data are altered to new statistics in the new scale. Table 7 presents the differences between the averages and the standard deviation of the provincial consumption levels observed from published statistics and those proposed in the table of equivalences; these values in the year 1925 can also be observed at the end of columns (1) and (2) in Appendix 2.a. As expected, these differences between observed and adjusted values decrease as the general levels of consumption rise.
Table 7. Average milk consumption and standard deviation in Spanish provinces (1925-1981) (Observed and adjusted values at provincial level, liters/person/year)
Source. Statistics cited in Table 1.
- 35 . Mean (m) and variance (σ2) of the distribution have been calculated from the empirical – provinci (...)
40Stage 2. Based on the new statistics, the distribution of the consumer population can be generated through the log-normal distribution (according to the formulation (2))35. This theoretical consumer population is presented using a cumulative function in column (6), Appendix 2b. The results of this operation for each province in the year 1925 are shown in ascending order in both columns in this table. The pattern of variation in the corrected consumption levels by comparison with the increasing proportion of consumer population allows for the estimation of the functional relationship between these two variables. Figure 4 shows the adjusted function, which is exponential and not linear, as it was assumed in applying the initial scale in Appendix 1. The results for the remaining years are displayed in Table 8. They make evident how, in the first place, basic levels of consumption have been increasing over the decades, as the progressive increase in the constant value of the equation reveals. At the same time, new consumers are being encompassed, as indicated by the decreasing tendency of the exponent which, therefore, causes the relationship between levels and proportions of consumers to approach a horizontal line.
Table 8. Daily consumption and proportion of the consumer population (Adjusted using log-normal distribution)
Notes. y(j) = Corrected medians for provincial consumption
x(j) = Proportion of consumer population according to lognormal distribution.
Source. Own elaboration.
Figure 4. Relantionship between daily consumption and acumulated Spanish consumer population in 1925
41Stage 3. The parameters of the functions adjusted in Table 8 will be a starting point for the aforementioned readjustment of the initial scale. Given that the basic data are provincial averages and that the proportion of the consumer population is unknown, such parameters can only be applied to the total of provincial populations, to be organized hierarchically (from the smallest to the largest) for each year studied according to the following formula:
- 36 . This estimate is where 0.0925 and 0.0117 are the parameters of the adjusted equation in Table 8 a (...)
Where « α » and « β » correspond to the two parameters estimated in Table 8 and the component of the exponent represent the percentage of accumulated population from each province « j » according to an established hierarchical order of consumption « k ». In Appendix 2c column (9) this procedure applies to the 1925 data. In this case, expression (3) applied to column (8) gives a result of 101 ml, as for example in the case of the province of Almeria (37 liters per person per year)36. The result obtained is the final adjusted provincial consumption average per person per day (column 9) and per year (column 10).
42Stage 4. Once these final corrected levels of consumption have been established, the rest of the procedure is immediate. Therefore, in the first place, the consumer population can be calculated for each province according to the expression:
Where the numerator corresponds to the Total Consumption (TC) of the province « j » according to the published statistics and the denominator to the effective or adjusted average level of consumption of the same province, as obtained in (3). So, finally, it will be possible to obtain the final percentage of the total consumer population from the sum of the estimated consumers in each province, according to expression (4), divided by the total of the Spanish population, for each of the years under study. In the case of Spanish consumption for the year 1925, this procedure is applied in columns (12) and (13) in Appendix 2c. The final estimate of consumer population was 43 per cent of the total Spanish population and the adjusted level of consumption per person per year was 84.53 liters.
43The final results of this procedure are presented in Table 11. Finally, the log-normal distribution permits calculation of the milk distribution percentage for the consumer population during the various intervals of consumption for each year, presented in Table 12 and Figure 5. Before the analysis of the results it could be useful to evaluate, through a sensitivity analysis, some of the basic assumptions made in this reconstruction procedure.
44As it has been stated, a central step in this approach is the adjustment of the statistical means with the scale presented in Appendix 1. In spite of the final adjustment in the third stage of the procedure it is quite obvious that this initial scale might have strongly conditioned the results. Tables 9 and 10 display the results of a sensitivity analysis for the years 1925 to 1965, before the proportion of consumer populations achieved 95 per cent. This analysis evaluates the effects caused by a change in the initial scale, particularly in minimum levels of consumption, on the final estimates of consumers and their consumption statistics (mode, median and mean). Two different assumptions about the range of variation in the scale of adjustment have been adopted: (A) Between 35 and 130 liters per person per year and (B) Between 55 and 130 liters per person per year. These values 35 and 55 are 10 liters above and below the 45 liter minimum consumption level in Appendix 1.
Table 9. Sensitivity analysis in the estimates of consumer populations, in 1925, 1933, 1955 and 1965
Source. Own elaboration.
45As it might be expected, if minimum consumption had been higher (55 liters), the proportion of consumers would have been overestimated (Table 9). For example in the year 1933 it would have been 51 rather than 55 per cent. If minimum consumption had been lower (35 liters), these effects would have run in the opposite direction. However, the differences between estimates shown in Table 9 suggest that changes in the initial scale of adjustment do not involve strong variations in the estimates of population proportions. Most differences are of less than 5 per cent and, in general, intervals do not alter appreciations about consumption patterns for each year. For example, an interval of consumers estimated for 1925 would be between 39 and 49 per cent, always showing that less than half of the Spanish population consumed milk. At the end of this period, 1965, a range of variation in consumer population between 72 and 80 per cent does not modify any basic conclusion about the fact that three quarters of the population were drinking milk daily.
Table 10. Sensitivity analysis in the estimates of statistics of consumption, in 1925, 1933, 1955 and 1965 (daily consumption per person ml/gr)
Source. Own elaboration.
46As far as the estimates of basic consumption statistics are concerned, Table 10 shows how they change – according to the levels of minimum consumption – in the expected way. The magnitude of the statistics is greater with 55 liters rather than 35 liters. In any case, these sets of estimates seem to converge during the years when consumption was increasing. Differences between those minimum levels and the initial scale (45 liters) are greater at the lower consumption levels than at the higher ones, for example if 1925 and 1933 are compared with 1955 and 1965. Between the three statistics, the range of variation in the estimates of the mean consumption is lower than the other two. In absolute values, the error would be of around 15% in the worst estimate. Mode reflects the statistics with a less accurate estimate, especially in 1925 when the results varied between 99 to 161 ml per person per day. Table 10 shows that the most difficult consumption patterns to reconstruct are the lower ones, but it is important to note that in any case, the obtained rank of variations does not change the conclusions about the probable scale of consumption. For example, in 1925, all estimates were far from the daily cup of milk.
47Two general conclusions can be drawn from these results. First, data correction for low levels of consumption seems more difficult and is subject to greater uncertainty. Second, this range of variation of approximately +10/-10 liters around the minimum scale of 45 liters per person per year does not seem to modify significantly the initial conclusions.
48Unfortunately the lack of estimates for proportions of consumers requires indirect testing of the validity of results. The basic criteria will be to place these results in connection with other published data on consumption of milk in the Spanish population.
- 37 . Nicolau-Nos, R., Pujol-Andreu, J. & Hernández, I., 2010.
- 38 . Simpson, J., 1995, p. 257-262 ; Domínguez Martín, R., 2003.
- 39 . From data collected by Fao and published by C. Agenjo, 1956, p. 315. Liters of milk consumption i (...)
49From a more historical perspective it is worthy to remember that animal milk – cow or goat – does not seem to have been particularly important in the traditional Spanish Mediterranean diet. But, as it has been mentioned previously, this foodstuff was incorporated into the diet of the Spanish population, particularly during the first third of the twentieth century37. The results obtained from this study affirm that this was a relatively slow and uneven process of diffusion, in terms of its evolution in the average consumption levels and the number of consumers. These results fit well with general appreciations about the difficulties faced by Spanish agriculture in the first half of the twentieth century in reallocating resources to livestock production and in specializing its production to develop a national milk market38. From an international perspective these new results do not change the fact that in terms of real consumption Spain, like other Mediterranean countries, remains at a low level by comparison with western and northern European nations39.
- 40 . See J. Bernabeu-Mestre, J. X. Esplugues-Pellicer, & E. Galiana-Sánchez, 2007. Another element tha (...)
- 41 . Carrasco, E., 1934.
- 42 . Simpson, J,, 1995, p. 371-379.
50Prior to the Spanish Civil War, in the mid 1930s, almost 45 per cent of the population did not consume milk (Table 11). Observations on both the lack of consumption and the lack of certain nutrients the milk offers, such as calcium, carried out using modern food science techniques since the 1930s,40 demonstrate this fact. In any case, the quantities obtained confirm the opinion held by contemporary researchers in connection with the fact that a substantial part of the Spanish population would not have consumed this foodstuff at the beginning of the 1930s41. From a broader perspective, these estimates around 1925 give empirical support to general assumptions about the distribution of the calories in the Spanish population around 1900. According to these, about 50 per cent of the population obtained less than the minimum requirement.42 According to the distribution frequency of consumers in 1925, practically three quarters of the consumer population would have consumed less than 225 ml per day (the equivalent to approximately one cup a day), with 27 per cent consuming less than half this amount (Table 12).
Table 11. Reconstruction of basic statistics indicators of milk consumption of the population, in Spain (1925-1981)
Source. Statistics cited in Table 1 and own elaboration.
Table 12. Distribution of Daily Milk Consumption Spanish Population (1925-1981), (Percentage of Consumers)
Source. Statistics cited in Table 1 and own elaboration.
- 43 . Amounts that are, for example, very close to the Danish consumption averages of the 1930s. See fo (...)
51Between 1925 and 1933 a considerable increase in consumption would have occurred. Average consumption – corrected mean – increased from 84.5 to 115 liters (Table 9). The change inferred by the 1933 distribution lies in the extension of consumption to higher daily quantities. Figure 5 shows this change in the form of a distribution chart. The comparison of the figures for consumption between 1925 and 1933 suggests that a double process of generalization and polarization of consumption was consolidated in the 1930s. Differences between mode and mean achieved the highest level in the century. Although approximately 25 per cent continued to consume less than the equivalent of one cup a day, the remainder now acceded to higher quantities; so 8 per cent of the population would have consumed over 475 ml a day, the equivalent to almost 175 liters per person per year43 (Table 12).
Figure 5. Distribution of consumer populations according to daily milk consumption intervals in Spain (1925-1981)
52Source. Statistics cited in Table 1 and own elaboration.
53The return to levels of consumption previous to the Spanish Civil War took time to occur, as confirmed by the slight increase in the proportion of the consumer population (around 3 per cent) and the stability of averages of the effective estimated consumption from 1933 to 1955 (situated between 115 and 112liters per person per year). This trend contrasts with the changes observed in the 1955 distribution: 52 per cent would have consumed between 150 and 299 ml and almost 45 per cent of the consumers would have drunk over 275ml a day.
- 44 . Agenjo, C., 1956, p. 315
54From the second half of the 1950s and the first half of the following decade, the average effective consumption did not show relevant changes, but the proportion of consumers entered a stage of expansion that reached, according to estimates made from the hbs, almost 80 per cent of the population. This led, therefore, to a phase of generalized consumption in the 1980s. It is important to observe that between 1965 and 1981, in contrast to what had occurred in previous stages, a double process of increase in consumption and increase in the number of consumers took place. At this point, however, it should be remembered that the average consumption estimated for 1981, close to 136 liters per person per year, was equivalent to that reached by the German population in the 1930s44. The Hbs of 1965 and 1981 infer consumption distributions of very different tendencies to the years 1925 and 1933. The distribution of 1955 could be described as closer to the patterns of the following, than the past years. Figure 5 shows a displacement to the right and a remarkable increase in the mode. The distribution of 1965 could contribute evidence of a step forward in the nutritional transition in Spain. Indeed, in those years a substantial proportion of those who consumed milk did so in quantities that were close to the equivalent of one cup a day; 50 per cent of the population drank between 175 and 275 ml a day, whereas in 1955 it was only 37 per cent. In fact, in 1965 around 63 per cent were consuming over 225 ml a day (Table 12). Distributions estimated from the years 1965 and 1981 demonstrate both the growth and generalization of consumption: the progressive approximation of the mode and average consumption reveal this fact (Figure 5). From the 1960s, together with the regular improvement of income in Spain, this foodstuff, just like all other animal based foodstuffs, would begin a trajectory of growing averages of consumption per capita as well as the universalization of milk consumption.
- 45 . Rodríguez, F. et al. 1996.
- 46 . Varela, G., 2000 ; Moreno, L. A. et al., 2002.
55This last characteristic, according to the results obtained, should have been reached in the 1970s. Milk and milk-derived products constituted, between 1969 and 1984, a regular proportion – around 55 per cent – of the total animal-based foodstuff consumed per person per year, which in terms of its calcium attributes represented 7 and 9 per cent of the total diet45. Towards 1981, the rise in consumption was consolidated in levels above 351 ml. So it was necessary to wait until the last decades of the twentieth century to witness the growing importance of dairy products in the national diet46.
56Nutritional transitions involve not only changes in levels of consumption, but also in numbers of consumers. The reconstruction of consumer populations is not an easy task because most of the historical statistics do not publish this kind of data. In countries with statistical yearbooks, the data available only allow the computation of statistical means at some spatial scale, according to the administrative units defined in each country. In order to overcome this limitation a methodological approach has been presented here. This methodology has been illustrated with the case of milk consumption, because an increase in the use of milk and its derivatives, as well as in the use of animal proteins in general, constitutes one of the main components of the modern nutritional transition, as it was the case in Spanish too. A basic step in the approach implemented in these pages has been the initial adjustment applied to the observed consumption levels, for them to become effective consumption levels, where the latter were the main unknown factor. In order to avoid an arbitrary solution this paper has presented an approach based on the combination of certain hypotheses about the diffusion of this foodstuff in the Spanish population with some results obtained from simulations of log-normal distributions. Such a combination has given a reasoned solution, but not a definitive one. This would explain why these results should be viewed as reconstructive. Unfortunately these estimates can only be tested indirectly, but they seemed to fit very well with the available information and evidence on the evolution of Spanish milk consumption in the twentieth century. In synthesis they suggest that, from the point of view of the number of consumers, there was a slow and late process of diffusion. Because this is a reconstruction the assumptions could be a source of error; this has been evaluated through a sensitivity analysis. The effects of a 10 liters variation in minimum consumption levels around the adopted scale of adjustment, between 45 and 130 liters, have shown that the conclusions previously obtained should not be significantly altered.
57This kind of reconstruction has some drawbacks. On the one hand, without further historical information on the population’s consumer habits it is difficult to evaluate effective consumption levels; on the other hand, the published statistics present two types of important limitations. First, the data were published according to certain administrative units, for example provinces. In the best cases, consumption means can be estimated, but not variances, so any calculation requires accepting variance at national scale as representative of that at a provincial level. Second, a lack of information on internal trade can either underestimate or overestimate levels of consumption. If all the above mentioned aspects are taken into account this approach could be extended to other foodstuffs involved in the same process of nutritional change. The possibility of estimating levels and trends in the evolution of consumer populations may shed new light on the study of the nutritional transition.