Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 Investigaciones de Historia Económica - Economic History Research www.elsevier.es/ihe Article A new statistical methodology for evaluating the diffusion of milk in the Spanish Population: Consumer groups and milk consumption, 1865–1981 Ismael Hernández Adell a,b , Francisco Muñoz Pradas c,∗ , Josep Pujol-Andreu d a Escola Superior CCSocials i de l’Empresa, Tecnocampus Mataró, Universitat Pompeu Fabra, C/ Ernest LLuch 32, 08302 Mataró, Barcelona, Spain b Departament Història i Institucions Econòmiques, Fac. Economia i Empresa, Universitat de Barcelona, Avda. Diagonal 690-696, 08034 Barcelona, Spain c Departament de Geografia, Campus de Bellaterra, Universitat Autònoma de Barcelona, 08193 Bellaterra, Cerdanyola del Vallès, Spain d Departament d’Economia i Història Econòmica, Campus de Bellaterra, Universitat Autònoma de Barcelona, 08193 Bellaterra, Cerdanyola del Vallès, Spain a r t i c l e i n f o a b s t r a c t Article history: This article analyses the diffusion of milk consumption in Spain between 1865 and 1980, and uses a new Received 9 September 2016 statistical approach to estimate consumer groups and their milk consumption. This new methodology Accepted 21 March 2017 shows that these variables increased at different speeds chronologically and geographically. Two main Available online 30 June 2017 phases can be distinguished in terms of the diffusion of milk consumption. The first phase, between the mid-19th century and the 1950s, was characterised by concentrated consumption in producer regions and JEL classification: big cities and the persistence of stark differences between regions. The second, between the 1950s and the N34 1980s, was chiefly characterised by an increase in the total number of consumers and the disappearance N54 N94 of these regional differences. © 2017 Asociación Española de Historia Económica. Published by Elsevier España, S.L.U. All rights Keywords: reserved. Milk consumption Nutritional transition Spain Regional differences Nueva estrategia estadística para evaluar la difusión de la leche en la población española: poblaciones consumidoras y consumo de leche, 1865-1981 r e s u m e n Códigos JEL: En este artículo, analizamos la difusión del consumo de leche en España entre (1865-1980), tomando N34 en consideración una nueva aproximación estadística para estimar las poblaciones y el consumo de N54 leche en estos colectivos. Con la nueva metodología, mostramos que estas variables aumentaron con N94 diferente intensidad, a lo largo del tiempo y a escala regional. Esta circunstancia nos permite distinguir Palabras clave: dos grandes etapas en la difusión de la leche en la dieta. En la primera, hasta los años 1950, destacaría, sobre Consumo de leche todo, el aumento del consumo en las poblaciones consumidoras y las grandes ciudades, y la pervivencia Transición nutricional de elevados niveles de desigualdad a escala regional. En la segunda, hasta los años 1980, destacaría el España aumento del número de consumidores, y la desaparición de aquellas desigualdades. Diferencias regionales © 2017 Asociación Española de Historia Económica. Publicado por Elsevier España, S.L.U. Todos los derechos reservados. ∗ Corresponding author. E-mail address: Francesc.Munoz@uab.es (F. Muñoz Pradas). https://doi.org/10.1016/j.ihe.2017.03.008 1698-6989/© 2017 Asociación Española de Historia Económica. Published by Elsevier España, S.L.U. All rights reserved. 24 I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 1. Introduction the analysis; and (b) a brief explanation of the methodological strat- egy followed for estimating consumer groups and the consumption The European nutritional transition in the 19th and 20th cen- of milk among them. The second section presents new indicators turies was not only the consequence of an increase in income for the three key variables in Spain, and proposes a new sequence levels caused by economic growth. It was also caused, perhaps of the diffusion of milk consumption in this country between the to a greater extent, by significant changes in consumption pre- mid-19th century and 1981. The third sections analyses provincial ferences, as clearly demonstrated by milk consumption.1 As late and regional trends and their main differences. The conclusions as the second half of the 19th century, medical specialists and compare our results to those obtained by other studies and pro- the population in general regarded foodstuffs on the basis of their pose future avenues of research. In order to make the text more calorie, protein and fat content, and liquid milk was, therefore, readable, the statistical references are presented in an Appendix. not highly esteemed. By the 1930s, however, milk was consid- ered a staple food throughout Europe. Three factors contributed to increase consumer preference for milk.2 These factors are: (a) 2. Sources, estimations and methodology scientific advances that identified new nutritional qualities in milk (calcium, vitamins, good-quality proteins) and its bacteriological In our study, we shall take into consideration the consump- composition; (b) the development of new methods to slow down tion of all forms of liquid milk: that is, cow and goat milk (which its deterioration (pasteurisation, refrigeration, sterilisation); and could reach the consumer either without having undergone any (c) the implementation of numerous initiatives by public bodies, preliminary treatment or after pasteurisation, sterilisation and/or the sanitary sector and the dairy companies to boost consumption.3 bottling), as well as evaporated and powdered milk. Our estimates As a result of these changes, milk went from playing a secondary for these variables may be found in Appendix 1. In the following role in the diet of Europeans, especially in the Mediterranean, paragraphs, we shall make reference only to their more significant where climatic conditions were unfavourable, to become a prod- features. Until the mid-1950s, the Spanish population consumed uct which was consumed in high quantities. However, while milk almost exclusively cow and goat milk, but statistics offer evidence consumption in numerous regions of Central and Northern Europe concerning only its production and commercialisation. Like other rose above 100 litres/person/year (l/p/y) before World War II, this studies, for the period 1865–1955 only apparent consumption level was not reached in Spain until the 1980s. Moreover, during can be estimated and do not take into consideration product loss this process national consumption figures tended to converge, and throughout the distribution chain. These estimates will be based on eventually levelled out around 100/p/y during the second half of the livestock censuses undertaken by different agencies in 1865, 1925 20th century.4 This trend is also attested in Spain on a regional scale. and 1933, and annual production statistics issued by the Ministry In Spain, regional differences in consumption were still substantial of Agriculture between 1954 and 1957.7 Furthermore, these statis- in the 1930s, but they tend to disappear after that time.5 tics offer uneven information on the number of animals and their Previous studies on these changes in milk consumption have, productivity, the quantity of milk consumed by calves and kids and however, a significant flaw: they only take into consideration the the destination of the net milk produced (direct consumption or overall average consumption. This involves assuming that changes industrial uses); for this reason, some preliminary estimates have in per capita consumption and in the number of consumers are had to be calculated. These estimations affect especially the years commensurate: that is, that these variables had increased simul- 1865 and 1933. taneously and at the same pace. The analysis of the diffusion of The livestock census of 1865 provides information on the total new products and services, however, demonstrates that such an number of bovine and caprine units and their distribution by sex, assumption is not necessarily justified: the proportion of con- age and productive use, but not on the production of milk or its final sumers and their average consumption can change independently destination. In addition, the available information does not include of each other.6 This article will examine this circumstance from a the total number of adult cows and goats. This lack of information new methodological perspective that may be of application to other is particularly problematic with regard to cows, which were, by far, diffusion processes. With this purpose we shall propose new ways the main source of milk in Spain (Appendix 1.1). In order to esti- to estimate per capita consumption, consumer populations and mate apparent consumption, therefore, we have had to estimate milk consumption among these between 1865 and 1981, and we the number of adult animals, their productivity and the propor- shall re-examine the dissemination of milk consumption in Spain tion of the production used in calf-rearing. For this purpose, we during this period. Ours is, doubtless, a risky approach, and it will assume: that 50% of calves born were female,8 that 66% of female have to be fine-tuned in future studies. In our opinion, however, goats were older than two years, and that the data provided by the new methodology that we propose can provide further evi- the Junta Consultiva Agronómica in 1891 on yields and the con- dence for the diffusion of milk consumption, and of the different sumption of rearing animals can be extrapolated to 1865. As is well converging and diverging trends at a regional scale. We also expect known, in the late 19th century the Spanish agricultural sector spe- our results to trigger new research hypotheses both in Spain and in cialised in the production of cereal and other vegetal foodstuffs, and other European countries. the stock-breeding sector was largely focused on the production of In order to achieve these targets, the text is divided into three sections. The first is methodological in nature, and is divided into two parts: (a) a critical examination of the statistical sources used in 7 These censuses were undertaken on behalf of the Junta General de Estadística (1865), the Asociación General de Ganaderos del Reino (AGGR) (1925) and the Ministerio de Agricultura (1934). Annual statistics will not be used owing to their 1 Atkins (1992, 2005), Popkin (1993), Grigg (1995), Kjærnes (1995), Murcott lack of reliability (Grupo de Estudios de Historia Rural, 1991). The FBS of 1958 was (1999), Moreno, L.A., Sarría, A. and Popkin, B.M. (2002). also rejected, because it was poorly executed (see report by the Instituto Nacional de 2 Nicolau, R., Pujol-Andreu, J. and Hernández Adell, I. (2010). Estadística (INE) at http://www.ine.es/metodologia/t25/t2530p458.pdf). The publi- 3 Nicolau and Pujol-Andreu (2008), Hernández Adell and Pujol-Andreu (2016), cation of the AGGR has traditionally been dated to 1923, but according to the Revista Orland (2005, 2007), Pujol-Andreu and Cussó (2014). de Higiene y Sanidad Pecuarias (Livestock Hygiene and Healthcare Journal) (February 4 International Dairy Federation (1982). 1926, p. 138) the real publication year was 1925. For further information, see Barciela 5 Collantes (2015), Domínguez Martín (2003), Muñoz Pradas (2011) and Nicolau, et al. (2005) and Hernández Adell and Pujol-Andreu (2016). 8 Pujol-Andreu and Hernández Adell (2010). Powell, Norman and Dickinson (1975: 1723–1726). Goat estimates according to 6 For these diffusion processes, see Duval and Biere (2002). age data, from Junta General de Estadística (1868). I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 25 working animals and meat.9 Meat usually from adult animals. Dif- sumption. Between 1965 and 1981, in contrast, the data reflect real ferent studies have also shown that the number of bovine animals consumption. The dearth of information for the periods 1865–1925 and the breeds used did not change significantly in 19th century.10 and 1933–1955 is, in any case, of little relevance for our purposes. Concerning the estimates for 1865, finally, we also assumed that As shown by previous studies, until the early 20th century milk only a small proportion of the net yield was used for the produc- consumption was very low in Spain. Afterwards, milk consump- tion of cheese and butter, and is not considered here. As illustrated tion increased continuously until the 1930s, decreased during the in Appendix 1.1: (a) milk production was very low in most Span- Civil War and post-war years, and increased again from the 1950s ish regions; and (b) the amount of milk used for calf-rearing was onwards.17 Similarly, the use of apparent consumption data before very high. In the 1920s and 1930s, in addition, although net milk 1955 and of the FBS from 1965 onwards does not cause difficulty. production was much higher, the production of dairy products was The differences between apparent consumption and that reflected still very low. in the FBS for the late 1960s are not so wide.18 Furthermore, it is no It is also important to note that stock-keeping statistics did longer justified to continue using apparent consumption estimates not always include urban milk production, probably because those for 1965 and 1981. We must not forget that the interprovincial in charge of data-compilation (the engineers of the Junta Con- milk trade increased continuously from the 1950s onwards, and sultiva Agronómica) regarded the keeping of urban dairies as an that this is only reflected in the FBS. Until the 1930s, this trade was industrial, rather than an agrarian, activity. This is clearly the case close to inexistent, and could have decreased even further during with the 1933 figures for Madrid and Barcelona.11 By this time, early Francoism.19 milk consumption was already quite high in these cities, and the Finally, we must point out that some of the changes detected on urban dairies were a major source of milk. These points were less milk consumption at the provincial level (Appendix 1), will need relevant in other provinces. In the central and southern regions, to be confirmed by future research, but that this has little effect on urban milk consumption was still very low in 1933. In the north- subsequent estimates concerning consumer populations and milk ern stock-keeping regions, the relative importance of urban dairies consumption among these. For instance, for the period 1925–1933, was always much smaller than in Barcelona and Madrid.12 This is our data suggest that milk consumption decreased in Oviedo and probably also applicable to Valencia and other eastern cities, which Santander, but that it always was above 200 l/p/y. Similar trends, bordered on extensive huertas (irrigated agricultural landscape). however, can also be detected in Switzerland in this same period, Most of the milk consumed in Valencia came from that area and and after World War II in other countries where consumption was included in the statistics.13 For all these reasons, only the infor- was high.20 Conversely, the reduction in consumption attested in mation concerning Barcelona and Madrid has been adjusted, with Huesca and León is of little relevance because milk consumption the addition of the production of urban dairies to net provincial in these provinces was always very low. Concerning the increase production: 29 and 20 million litres respectively.14 in milk consumption in Gerona, La Coruña, Salamanca or Málaga, Finally, it is likely that our estimate of apparent milk con- there is not much more evidence available. In Gerona, however, sumption in 1955 (1954–1957) is too low, and thus we should we know that milk consumption in the capital was around 25 l/p/y be cautious when dealing with data between 1955 and 1965. It shortly before World War I, and that in the 1930s consumption is possible that the consumption of powdered and condensed milk rocketed to over 80 l/p/y. The consumption in hospital institutions already increased as early as the 1950s, but reliable estimates are also increased significantly in the province. In the hospital of Sant impossible.15 Likewise, the data available does not allow us to Jaume, in Olot, milk consumption was less than 5 l/patient/year determine to what extent was urban milk consumption estimated in 1885, and shortly before the Civil War was 180 l.21 Concerning in the new statistics elaborated by the ministry.16 the period 1955–1965, it should not be surprising that milk con- From 1965 onwards, the data available are more reliable because sumption increased quickly in central and southern provinces. On Encuestas de Presupuestos Familiares (FBS; Family Budget Surveys) the one hand, the interprovincial milk trade was increasing during provide direct milk consumption data. The 1965 FBS, however, did that period, and local production in many of these provinces was not take into consideration powdered and condensed milk, and also on the rise. On the other hand, migration from these regions we have had to estimate it. For this purpose, we have used the intensified in the 1950s and 1960s, as did the urbanisation process. information collected by the Instituto Nacional de Estadística in As a consequence of these phenomena, the population increased 1969. In conclusion, stock-breeding statistics only allow for a rea- but only a little in Andalusia, Murcia and Aragón, and decreased in sonably accurate estimation of milk consumption for 1865, 1925, Extremadura and both Castiles. At the same time, the proportion 1933 and, less accurately, for consumption in 1955. IN addition, for this period, statistics only allow for the estimation of apparent con- 17 Between the end of the Civil War and 1954 (year of first publication of Resumen 9 Junta Consultiva Agronómica (1892). See also Domínguez Martín (2003), estadístico de la producción, destino y valor de la leche (Statistical summary of pro- Hernández Adell and Pujol-Andreu (2016), Langreo (1995). duction, destination and value of milk), milk production statistics were published 10 Grupo de Estudios de Historia Rural (1985), Hernández Adell and Pujol-Andreu only in 1941 and 1943. According to these estimates, since the 1930s consumption (2016). had dropped by nearly 50%, and the number of cows and goats by 20% (Anuario 11 Madrid dairies produced 45.6 million litres, and the production in the province Estadístico de España (1943: 500–505; 1944–45: 472–475). See also, Hernández was, according to stock-breeding statistics, 40 million (Doaso (1931: 26–28)). In Adell and Pujol-Andreu (2016). 18 1933, the city of Barcelona consumed 72 million litres of milk, and the production of The apparent consumption of cow and goat liquid milk in 1964–1969 was the province was estimated at 71.6 million (Mas (1933: 20–21)). See also Ministerio 70 l/p/y. According to the FBS for 1965, liquid milk consumption was 78.7 l/p/y. 19 de Agricultura (1934) and Hernández Adell and Pujol-Andreu (2016). Milk brought from other provinces to Madrid and Barcelona in the 1930s 12 Puente (1992) and Domínguez and Puente (1996: 128–132). amounted to only 17% and 13% respectively of all commercialised milk (Doaso (1931: 13 In 1895, 81% of dairies, and 71.6% of dairy cows, were in the huerta, which 26–28), Generalitat de Catalunya (1937: 5, 57–65), Mas (1933: 20–21)). According to produced 75.6% of all commercialised milk (Calatayud and Medina, 2017). the Ministerio de Agricultura (1954), the interprovincial milk trade barely amounted 14 García (1927: 959), Ayuntamiento de Madrid (1929: 248), Mas (1935: 25). to 5% of total production. 15 20 The consumption of these forms of milk was very low in the 1930, and had The Annuaire Statistique de la Suisse (1930: 153) indicates that consumption increased considerably by 1965. decreased from 264 to 235 l/p/y between 1910 and 1940. A similar trend can be 16 In order to estimate per capita consumption before 1955, the provincial popula- attested in France, Holland, Denmark, Great Britain and Sweden from 1950 onwards tion figures are estimated by linear interpolation with reference to the two censuses (International Dairy Federation (1982)). 21 undertaken at the two closet dates, before and after. FBS already include population Archivo Comarcal de la Garrotxa, Fondo Hospital Olot (c-19 a 31 and 230; and data. Ajuntament de Girona (1934). 26 I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 of the population living in cities with over 100,000 inhabitants Concerning the format of milk consumption, we shall assume increased from 45% to 60% between 1940 and 1970. that consumer groups drank milk from containers of around 250 ml In order to carry out our calculations of the two new variables or fractions thereof.22 This is a reasonable assumption because that we propose, we shall follow the methodology proposed by milk was generally drunk from mugs, glasses or other similar con- Muñoz Pradas (2011) in Histoire & Measure. The most technically tainers of this approximate capacity (or half). We shall also assume relevant aspects of this proposal are synthesised in Appendix 3. The that milk was drunk on a daily basis, or at least on alternate days; following pages will merely examine the methodology from a con- a more sporadic consumption pattern would require conservation ceptual perspective. We hope that, by choosing this way to present techniques that would not become available until well into the 20th our study, we shall help historians to appreciate the basic outlines of century. The evidence provided by local doctors (medical-sanitary our proposal while giving those with statistical knowledge enough topographies) between the mid-19th century and the 1930s, and in material to revise our figures. a wide variety of studies and surveys, supports this assumption.23 The estimation of total average consumption, per person per These texts, also, distinguish between three types of consumer: year (or any other time unit), it is a useful variable, but limited in (a) those who never drank milk; (b) those who drank milk only scope. For instance, let us assume two provinces A and B, with the when they were proscribed liquid diets (for example, as a conse- same population (100,000), but with different consumption lev- quence of an illness); and (c) those who drank milk frequently.24 els. A consumes 7 million litres, and the average consumption is, Based on this, we may assume that, in the second half of the 19th therefore, 70 l/p/y. B consumes 130 million litres, and the aver- century, there were already social groups which consumed milk age consumption is 130 l/p/y. The automatic conclusion is that per habitually and which drank it from 250 or 125 ml containers. Thus, capita consumption in B is higher than in A. However, if the pro- we understand that an observed average consumption of, say, portion of consumers in A is 50% of the total population and 100% 13 l/p/y (35 ml/p/day), in a given province and year, is more reflec- in B, that means the average consumption among consumers in A tive of the existence of an elevated percentage of non-consumers is higher (140 l/p/y) than in B (130 l/p/y). This is due, obviously, to than of the occasional consumption of most of the population. Sim- the fact that the average consumption, as reflected in the statistics ilarly, if observed consumption is 130 l/p/y (356 ml/p/day), we shall (X Cob ), is the total milk consumption (TMC) divided by total pop- assume that milk consumption was habitual among the majority ulation (TP), and that, when the consumer population (CP) is less of the population. than the TP, the resulting statistical average is lower than if only Concerning milk consumption diffusion patterns, we presume the CP was taken into consideration. This average will be referred that they will follow Rogers’ (2003) ‘diffusion of innovation’ trends. to as adjusted consumption (X CAj ). In formal terms: In the most simplified version of these statistical models, those If observed consumption is processes are divided into three sequential stages – initial, inter- mediate and final – among which the population is distributed TMC according to a Standard Normal Distribution. Following Rogers, we X COb = TP shall begin by assuming that the proportion of consumer popula- and adjusted consumption is tion was initially between 2.5 and 13.5% of the population (‘early adopters’), that in the intermediate phase this proportion would TMC reach 50% and that the process would be completed during the final X CAdj = CP stage. It has to be taken into account that, in the long term, diffusion processes are irreversible.25 Some historical evidence supports our the relationship between these coefficients is assumption. According to recent research, in the late 19th and the 1 TMC 1 early 20th centuries milk consumption was habitual among high- X CAdj = X Cob × = × (1) income groups or in stockbreeding areas in the north of Spain,26 fc TP fc but negligible among the rest of the population. Later, consumption where fc = CP(Consumer Population)/TP(Total population) is the increased, but not evenly. This was partially due to the incorpora- ‘correction factor’, or the proportion of consumers in the total pop- tion of new groups into the category of consumers: children and ulation. If fc = 1, then the two estimations of consumption per capita in the expression (1) are identical. If fc is less than 1, then observed consumption is lower than adjusted consumption (fc cannot be 22 This excludes milk used in cooking (for the preparation of sauces and similar higher than 1). uses). The distinction between observed and adjusted consumption is 23 Different studies show that workers in Madrid and Barcelona were already in the relevant not only in statistical terms but also in analytical terms habit of drinking coffee with milk with their breakfast in the early 20th century (AEB, 1902: 526; Luís and Yagüe, 1903: 9–10; Gresa, 1929: 104), Other studies also indicate because it reflects those factors which affected the evolution of that, around 1910, students in the schools of El Bosch (Barcelona) and San Ildefonso per capita consumption (observed consumption) most directly: in (Madrid), and also the interns at the Hospicio de la Santa Casa de la Misericordia expression (1), the ‘correction factor’ (proportion of consumers) (Bilbao) were served milk with coffee or chocolate milk in the mornings (Carro, and ‘adjusted consumption’ (average consumption among the con- 1917: 14–15; Pérez Castroviejo and Martínez Castroviejo, 1996: 66 and 79; Cussó sumer population). In order for expression (1) to be practicable, and Garrabou, 2004). 24 A list of topographies is in Vallribera (2000) and section 14.2. of the database in however, the consumer population must be estimated beforehand, www.proyectonisal.org. See also Nicolau, Pujol-Andreu and Hernández Adell (2010) or the expression would include two unknown variables. In this and Pujol-Andreu and Cussó (2014). regard, our methodology relies on a first indirect estimation of 25 These assumptions will have to be corrected later. The model proposes a generic adjusted consumption which is based on three assumptions that description of the diffusion of innovations, but does not claim that this kind of pro- cess systematically repeats the same trend every time. Also, the model does not we regard as reasonable. These assumptions refer to the format take into account the pace of the process, for example if the consolidation of an and frequency of liquid milk consumption, the diffusion process early group of innovators takes longer than that of the early adopters, or vice versa. of the product and the statistical model used in order to repre- Rogers (2003) reviews several examples that fit with the model, but the model does sent the distribution of consumption among the total population. not explain all processes (Kauffman and Techatanasoontorn (2009)). 26 These variables will be used to estimate the adjusted consumption According to the data collected by the Comisión de Reforma del Impuesto de Consumos (Commission for the Reform of Consumption Tax) in 1908, concerning and, with the aid of the available data on observed consumption the consumption habits of 434 individuals distributed across 79 families, milk con- and total population, also to estimate the consumer populations sumption among those in higher income brackets was 266 ml/p/day. A more detailed solving expression (1). discussion of these issues is in Muñoz Pradas (2011: 144). I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 27 Table 1 it is worth stressing that log-normal distribution allows for the Observed consumption (1) and population that could consume between 125 and representation of asymmetrical distributions, which is particularly 250 ml per day (2): Spain, 1865–1981. useful in the analysis of consumption habits. In these processes, Year (1) (l/p/y) (2) (%) averages and modes can change drastically and even follow diverg- 1865 20.19 8 6 ing trajectories. 1925 44.75 17 10 The only way to test the reliability of the third assumption, 1933 56.28 34 18 given the lack of individual data, is through the observed average 1955 56.32 37 17 consumption on the provincial level. As previously noted, these 1965 87.97 89 38 1981 141.96 100 100 averages are calculated for the total population of each province and do not take into consideration, therefore, the specific consumer Source: Authors’ own after Appendix 1 and the indications contained in the text. groups to which the assumption applies. In our opinion, however, we can assume that the range of variation of average consumption young people, chiefly, at an early stage, and healthy adults later among consumer populations will be lower than among the whole (Nicolau and Pujol-Andreu, 2008). population, at least up to consumption levels which are in the range Also, a new piece of evidence concerning of this diffusion pat- of 250 ml/day (91 l/p/y). In order to explore this, we extend the tern is the proportion of the population that could have consumed initial normality contrast to provincial distributions in which the between 125 and 250 ml of milk daily between 1865 and 1981, overall average consumption increases but is always below 91 l/p/y, assuming that nobody else consumed milk. Table 1 shows that the and assuming consumer populations that represent 80%, 60% and proportion of the population that could have consumed those quan- 40% of the total population. This exercise is presented in Table 2.30 tities of milk closely matches the proposed trend for innovation The result is significant, with the exception of Shapiro–Wilk test for diffusion processes. Around 1865, the proportion of the popula- 1925 (0.045) and 1965 (0.017, assuming a consumer population of tion that could drink milk habitually would be below 10% (initial 40%). In both cases, the level of significance is above 1% but below phase). Afterwards, consumer populations could have increased to 5%, and the null hypothesis must be accepted.31 In the rest of cases, between 20% and 30%, shortly before the Civil War, and to 40 y 60% the null hypothesis of normality is not rejected at the usual lev- in the 1960s (intermediate phase), covering the practical totality of els of significance. It is reasonable, therefore, to assume that milk the population in the 1980s (final phase).27 Concerning the initial consumption followed that distribution model. diffusion of milk consumption among children, the references to Once our assumption is accepted, the next step will be to use the founding of Milk Depots (Gota de Leche) and other promotion the log-normal distribution in order to estimate the adjusted con- campaigns in schools and charitable institutions in the first third of sumption among consumer populations. In this way, we shall the 20th century are abundant.28 be able to estimate the proportion of consumer populations Concerning our third assumption (the statistical model), we which correspond to given levels of adjusted consumption, always suggest that milk consumption patterns respond to a log-normal assuming that consumption was, as a minimum, in the range of distribution. This distribution is often used for the modelling of 125–250 ml/p/day (our first assumption). Fig. 1 presents the dif- economic activities and the consumption of products and services ferent levels of daily consumption among consumer populations, (Lawrence, 1988), and it is also frequently used in historical nutri- with an average level of adjusted total consumption of 25, 45 and tion analysis (Fogel, 1992; Floud et al., 2011). Following this line 130 l/p/y, and a standard deviation of 20, 40 and 60 l. We suggest of thought, we estimate the distribution of consumer population that these magnitudes accurately reflect most possible variations in using the density function: adjusted consumption on the provincial level in the 20th century.32 1  1  The horizontal lines indicate adjusted consumption levels between 2 f (xj ) = √ exp − 2 (ln xj − m) (2) 125 and 250 ml/p/day, and those above 250 ml. xj  2x 2 The results of these simulations indicate that an adjusted total This expression, f(xj ), represents the total population that con- consumption level of 25 l/p/y means that barely 10% of the popula- sumes daily a xj quantity of milk; and ‘m’ and ‘’ refer to the average tion consumed 125 ml of milk or more daily. If we consider 45 l/p/y, and the standard deviation of consumption distribution among the at least a quarter of the population (two deciles and a fraction of total population.29 This modelling of milk consumption patterns the third), could consume the equivalent of half a mug of milk daily, will be used for two different purposes. Firstly, in this section, in or a mug every two days. That is, the former figures equate to the order to simulate consumer distributions on the basis of hypo- ‘early adopters’ stage while the second to the intermediate phase.33 thetical provincial adjusted consumption averages and standard According to the simulation, an average of 130 l/p/y indicates that deviations. These simulations will allow us to obtain the average the adjusted total consumption would approach 250 ml, and that adjusted consumption levels at the beginning of the diffusion pro- some groups (the top three deciles) could even reach half a litre. cess. Secondly, in the next section, we shall apply the model in These magnitudes indicate generalised milk consumption. order to reconstruct the distribution of milk consumption among On the basis of the simulations and our assumptions concern- consumer groups, after estimating adjusted consumption. Finally, ing the diffusion of milk consumption, we shall adopt therefore as 27 For a detailed review of other data on the diffusion of milk consumption among 30 It must be kept in mind that random variables follow log-normal distributions different social groups, see Muñoz Pradas (2011). 28 if their logarithm follows the normal distribution. Muñoz Pradas (2016), Rodríguez, Ortiz and García (1985), Nicolau, Pujol-Andreu 31 The null hypothesis to be tested is that observed distribution is equivalent to and Hernández Adell (2010). A study of milk consumption in Germany in the 1930s the normal distribution. estimated that consumption among children aged under 5 was 222 l/y, and 197 l/y 32 These magnitudes and, especially, the variance are based on observed consump- for children aged 5–10. For older age brackets, consumption was lower: between tion data and variances presented in Table 6. 100–120 l/y (Casado de la Fuente, 1931: 50–51). 33 29 For a more detailed analysis of these simulations, see Muñoz Pradas (2011). Average (m) and variance ( 2 ) of the distribution have been calculated on the It is to be noted that the proportions of consumer groups are estimated on the  average (x) and variance (V) according to the following equations: basis of provincial basis of the total population, while the simulation refers to consumer groups, based 2 2 m = ln(x2 / x + V ); and  2 = ln(1 + V/x ). See Calot (1988: 182–197). For a bet- on the log-normal distribution. This means we are assuming that the distribution of ter understanding of this distribution in the assessment of the nutritional state of the consumer population among the different consumption ranges is equivalent to historical populations, see Floud et al. (2011: 49–57). that of the total population. This statistical premise does not affect the estimates. 28 I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 Table 2 Spain (1865–1981): distribution of provincial observed milk consumption. Normality tests for observed consumption and different adjustments made in order to reflect the proportion of consumer groups. Year Observed consumption Observed consumption: consumer population 80% Test Test K-S-L Shapiro–Wilk K-S-L Shapiro–Wilk D (n) Sign. lev W Sign. lev D (n) Sign. lev W Sign. lev 1865 0.092 0.200 0.977 0.451 0.086 0.200 0.977 0.465 1925 0.103 0.200 0.952 0.045 0.088 0.200 0.965 0.147 1933 0.109 0.200 0.967 0.177 0.087 0.200 0.979 0.540 1955 0.098 0.200 0.978 0.499 0.076 0.200 0.987 0.873 1965 0.069 0.200 0.985 0.765 0.088 0.200 0.966 0.173 1981 0.075 0.200 0.963 0.125 0.075 0.200 0.963 0.125 Year Observed consumption: consumer population 60% Observed consumption: consumer population 40% Test Test K-S-L Shapiro–Wilk K-S-L Shapiro–Wilk D (n) Sign. lev W Sign. lev D (n) Sign. lev W Sign. lev 1865 0.083 0.200 0.977 0.452 0.084 0.200 0.975 0.370 1925 0.082 0.200 0.970 0.234 0.077 0.200 0.975 0.362 1933 0.069 0.200 0.987 0.853 0.065 0.200 0.988 0.897 1955 0.069 0.200 0.989 0.933 0.069 0.200 0.989 0.926 1965 0.101 0.200 0.962 0.110 0.109 0.200 0.942 0.017 1981 0.075 0.200 0.963 0.125 0.075 0.200 0.963 0.125 K-S-L: Test Kolgomorov–Smirnov–Lilliefors. Source: Authors’ own after indications contained in the text. characteristic consumption levels among consumer groups 25 l/p/y Table 3 Observed milk consumption: Spain, 1865–1981 (l/p/y). for the mid-19th century, and 45 l/p/y for the early decades of the 20th century. These two magnitudes will be used as initial levels Years Consumption Years Consumption of adjusted consumption, in two lineal scales progressing towards 1865 20 1960–64 64 130 l/p/y. These scales cover, according to the available information, 1890 20 1964–65 (FBS) 88 the potential consumption patterns in the 19th and 20th centuries. 1925 45 1965–69 70 The values corresponding to these scales are tabulated in Appendix 1933 56 1970 81 1943-49a 46 1975 95 2 and represented in Fig. 2. 1950-54a 52 1980 104 This figure also illustrates the adjustment applied for a given 1954–55 56 1981(FBS) 147 amount of milk consumed, based on the statistics (horizontal axis), 1956–60 65 to the amount of milk supposedly consumed by the consumer Source: Authors’ own after Appendix 1, Anuario Estadístico de España (1943–54), population (vertical axis), according to our second assumption Contreras Hernández (2002), García Dory and Martínez Vicente (1988) and regarding the diffusion process. For example, an average observed Ministerio de Agricultura (1956–60). a consumption of 50 l/p/y would correspond to an average consump- In these cases, the source presents a single figure, which includes the quantity of milk earmarked for human consumption and the quantity consumed by rearing tion among the consumer population (adjusted consumption) of animals; we have assumed that the former amounted to 65% of the total. See Anuario nearly 80 l/p/y. These adjustments will be applied to all estimations Estadístico de España (1943–54). of average observed consumption in the provincial level, and will be the starting point for the methodology developed in Section 2. Table 4 As an illustration, Appendix 3 explains our estimates of provincial Basic statistical data concerning milk consumption: Spain, 1865–1981. consumer populations and their consumption levels in 1925. 1865 1925 1933 1955 1965 1981 3. Milk consumption in Spain: evolution and distribution Consumer population (%) 34 43 52 52 80 95 Non-consumer population (%) 66 57 48 48 20 5 Observed average consumption (l/p/y) 20 45 56 56 88 147 Table 3 summarises the main trends in observed milk con- Adjusted average consumption (l/p/y) 59 105 108 108 110 154 sumption in Spain since 1865 (without considering consumer Source: Authors’ own after indications contained in the text. populations). As can be appreciated, consumption increased chiefly during two periods, between the late 19th century and the Sec- ond Spanish Republic and in the second half of the 20th century. If we apply our methodology here, we appreciate that the Observed consumption remained low throughout the 19th century, significance of the two variables that we have estimated – con- and never surpassed 20 l/p/y. During the 20th century, consump- sumer population and their average milk consumption (adjusted tion grew progressively and reached 56 l in the 1930s; it decreased consumption) – is very different.35 According to our estimates in the following decade and did not reach pre-Civil War levels until (Table 4), in the late 19th century most of the Spanish population the mid-1950s. After that, consumption grew again in a sustained did not consume milk on a daily basis, and although this situa- manner until it broke the 100 l/p/y barrier.34 tion improved over the three following decades, this improvement 34 35 Domínguez Martín (2003), Hernández Adell and Pujol-Andreu (2016) and These calculations modify those published in Muñoz Pradas (2011), as a result Langreo (1995). of the new estimations in Appendix 1. I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 29 Average adjusted consumption 25 l/p/y 500 450 400 Daily consumption 350 300 60 250 40 200 20 150 100 50 0 t d d h h h h h h th 1s 2n 3r 4t 5t 6t 7t 8t 9t 10 Deciles Average adjusted consumption 45 l/p/y 500 450 Daily consumption (ml) 400 350 300 60 250 40 200 20 150 100 50 0 t d d h h h h h h th 1s 2n 3r 4t 5t 6t 7t 8t 9t 10 Deciles Average adjusted consumption 130 l/p/y 1100 1000 900 Daily consumption (ml) 800 700 60 600 40 500 400 20 300 200 100 0 t d d h h h h h h th 1s 2n 3r 4t 5t 6t 7t 8t 9t 10 Deciles Fig. 1. Distribution of consumption levels by deciles, according to three levels of standard deviation for this variable. Source: Authors’ own after the indications contained in the text. affected only some social groups. Around 1865, the percentage of 1950s onwards, consumption increased again, but owing to two consumer population was still below 35%,36 increasing later until different circumstances. Until the mid-1960s, the increase in con- it reached between 40% and 45% in the mid-1920s, and nearly sumption was largely due to the growth of consumer populations. 50% during the Second Spanish Republic. Consumption among After this, consumption chiefly grew due to the increase in con- consumer populations (adjusted consumption), conversely, must sumption within consumer groups. Between 1955 and 1965, the have grown considerably. Around 1865, the average consump- adjusted average consumption remained almost the same, while tion among these groups in the whole of Spain would be close to the proportion of daily consumers grew from 50% to nearly 80% 60/l/p/y, and around 110 l/p/y in 1925 and 1933. Finally, from the of the total population. Between 1965 and 1981, the proportion of consumer populations reached its biological maximum,37 and the 36 A detailed examination of the results for this year reveals that only 7% of the population consumed over 90 l/p/y (250 ml/p/day). In the terminology of the dif- 37 fusion model, these consumers were the ‘early adopters’. As we shall see later, in Based on FENIL (2008), and the microdata of the 1980 Family Budget Survey we the comment to Figure 3, the modal value of consumption in 1865 (78 ml/p/day) place this limit at 95% of the population. In this FBS, 4425 interviewees out of a total suggests, furthermore, very sporadic consumption. of 88,540 did not drink milk at all. 30 I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 Table 5 Distribution of population according to adjusted consumption levels, 1865–1981 (%). Milk (ml) 1865 1925 1933 1955 1965 1981 <125 63.11 35.51 23.34 8.63 0.59 125–174 20.01 18.83 19.67 17.27 7.24 175–225 9.17 13.67 16.31 19.70 19.78 0.33(a) 225–274 4.09 9.47 12.07 16.85 24.85 3.10 275–324 1.86 6.49 8.53 12.52 20.38 10.68 325–374 0.87 4.48 5.93 8.64 13.06 19.02 375–424 0.42 3.12 4.10 5.74 7.22 21.74 425–474 0.21 2.20 2.85 3.74 3.65 18.33 ≥475 0.25 5.77 6.84 6.80 3.22 26.79 (a) <225 ml 50% Consumers 53.52 54.34 51.70 52.77 56.46 50.37 <110 ml <175 ml <200 ml 150–299 ml 200–324 ml 350–474 ml Source: Authors’ own based on the indications contained in the text. 140 Our indicators also allow for a clearer appreciation of the sin- gularities of each stage in the diffusion of milk consumption. In 120 1865, nearly 60% of the consumer populations consumed less than 125 ml per day. If we also take into consideration that the mode was 78 ml (see table associated with Fig. 3), we shall also conclude that Adjusted consumption l/y 100 most of the consumption must have been sporadic, as is confirmed 80 by the historical record.39 In 1925, the consumer populations and, mostly, consumption among this group (adjusted consumption) were higher, but a more detailed examination reveals that nearly 60 75% of the consumer populations drank less than 225 ml (one mug) a day, and 34% less than half that Moreover, in 1925, the relative dif- 40 ference between daily consumption mode (99) and average (207) reached its historical maximum (109%). According to our calcula- 20 tions, therefore, this was the period during which inequality in milk consumption in Spain reached its peak. If, finally, we compare the 0 number of consumers who drank over 275 ml per day in 1925 and 5 12.5 20 27.5 35 42.5 50 57.5 65 72.5 80 87.5 95 102.5 110 117.5 125 1933, it is also reasonable to conclude that a selective improve- ment in consumption must have taken place in the 1930s. In 1933 Observed consumption l/p/y almost 30% of the population consumed over 275 ml of milk per day (22% in 1925), and nearly 7% of the population drank over 475 ml Scale 45 l/p/y Scale 25 l/p/y (175 l/p/y). It is also significant that the differences between mode and average tended to decrease from the mid-1920s onwards. Fig. 2. Initial adjustment of observed consumption. The trend towards higher levels of consumption among the Source: Authors’ own after Appendix 2. consumer populations can also be attested for 1955, when the dif- ference between mode and average also shrank substantially. It is, therefore, quite likely that the stagnation of milk consumption in average adjusted consumption grew by nearly 30% (from 110 to the 1940s and the early 1950s was linked to a significant change in 143 l/p/y). the distribution pattern. Between 1933 and 1955, the percentage Table 5 and Fig. 3 show the frequency distribution of con- of the population that consumed 275 ml of milk or more per day sumers between different consumption levels, after the application increased from nearly 28% to 37%. of expression (2) (log-normal model) to the averages and standard The frequency distributions for 1965 and 1981, finally, clearly deviations of adjusted consumption, previously estimated province indicate the increase and generalisation of milk consumption and by province.38 According to these new indicators, and as could be the progressive reduction of inequality (convergence of mode and expected in a context of growing and increasingly generalised con- average). The data for 1965 correspond to an advanced stage of sumption: (a) frequency distributions tended to move towards the the nutritional transition process: 25% of habitual consumers con- right, (b) their profile became more symmetrical, and (c) the mode sumed between 225 and 274 ml/p/day. Around 1980, most people and the average converged (see table associated with Fig. 3). The consumed more than 325 ml of milk per day, and those who drank bottom of Table 5 – the consumption levels of 50% of the consumer less than 275 ml made up below 5% of the total population. All of population – presents a numerical representation of this trend. this suggests, in agreement with Collantes (2015), that the 1980s These levels increased from under 110 ml/p/day around 1865 to witnessed the beginning of a new trend in milk consumption. almost 200 ml in 1933, and more than 350 ml in 1981. 4. Territorial diffusion patterns in milk consumption in Spain in the 20th century 38 For the calculation of the average and standard deviation, see Footnote 29. It must be taken into consideration that consumption estimates (based on provincial The diffusion of milk consumption proceeded at a very dif- averages and standard deviations) provide more compressed distributions around ferent pace in different regions. Evidence for this is provided in certain intervals than those that would result from data concerning household consumption. The degree of dispersion of provincial averages compared to total consumption averages could be expected to be less pronounced than the disper- 39 sion of consumption averages for each household compared to general household As previously noted, these references indicate that milk consumption was often average consumption. related to illness and the prescription of liquid diets by doctors. I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 31 12 10 Consumer population, % 8 1865 6 1925 1933 4 1955 1965 2 1981 0 0 100 200 300 400 500 600 700 800 900 1000 Adjusted daily consumption (ml) Statistics of daily milk consumption (ml) 1865 1925 1933 1955 1965 1981 Mode 78 99 134 190 246 393 Average 121 207 235 265 283 424 (1) 55 109 75 39 15 8 Fig. 3. Distribution of the consumer population according to adjusted consumption intervals: Spain, 1865–1981. Note: (1) Difference between average and mode in %. Source: Authors’ own based on the indications contained in the text. Table 6 1981 a similar proportion of provinces presented a consumption Observed milk consumption (l/p/y) and standard deviation: Spain, 1865–1900. of 125 l or more. Similarly, as previously noted, the differences Year Consumption average provinces Standard deviation between provinces with the highest and the lowest consumption 1865 21.56 29.5 levels were very great in 1925 and 1933, and tended to decrease 1925 42.91 64.03 thereafter. These differences, on the other hand, were very stable. 1933 54.65 63.76 The provinces of Santander and Cuenca, for instance, remained in 1955 61.24 56.32 the same position most years. 1965 85.31 39.04 Map 1 illustrates the territorial trends in more detail. Three main 1981 147 33.61 trends may be observed between 1865 and 1965. Firstly, that of the Source: Authors’ own after Appendix1. north of the Peninsula (Cantabrian and Atlantic provinces), where consumption was always higher than elsewhere. Secondly, that of Table 6, which shows the evolution of provincial aggregate aver- the northern half of the peninsula, with the exception of the north- age consumption and the standard deviation. Two issues must ernmost fringe, where consumption increased until the 1930s. be stressed. Firstly, the average consumption increased and the Thirdly, that of the provinces where consumption was always low- provincial differences shrank. Secondly, this was not a linear pro- est and which joined the diffusion process at a later date (region of cess, but follows the characteristic sequence of diffusion processes. Valencia and provinces of Teruel and Cuenca). Between 1965 and An early stage characterised by low and homogeneous average 1981, average consumption levels grew and reached 100 l/p/y in all consumption levels was followed by a stage of expansion, in regions. which standard deviation increased and the differences between Map 2 presents adjusted consumption. It repeats some features provinces widened, and by a last stage with higher and converging already seen in Map 1, but it also illustrates other issues. In some consumption levels and a decrease in standard deviation: a typi- cases, territorial patterns vary, especially between 1925 and 1933. cal inverted ‘U’ pattern (Williamson, 1965). As can be appreciated, During this period, a number of provinces in Old Castile, the south differences became more accentuated until they reached their his- of the Peninsula, and Barcelona and Girona show an increase in torical maximum between 1925 and 1933 (divergence stage), and consumption, compared to Map 1. decreased later in a sustained manner until the 1980s (convergence Table 8 provides additional information about regional patterns stage). in the diffusion of milk consumption. The table shows the adjusted Table 7 presents the sequence more accurately. In 1865, consumption per person per year according to historical regions, observed consumption in 36 of the 49 provinces was below 25 l/p/y. and the relative position of each region. In some regions, consump- In 1925, however, this occurred in only 28 provinces, and in only tion was always above the national average: Asturias, the Basque one in 1965. Until 1955, on the other hand, observed consump- Country and Navarre, from the beginning, and later also Galicia tion in more than half the provinces was under 50 l/p/y, while in and Old Castile. The Canaries join this group in 1981, owing to the 32 I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 Table 7 Distribution of Spanish provinces into intervals of observed consumption, 1865–1981 (l/p/y). Intervals 1865 1925 1933 1955 1965 1981 Number of provinces <25 36 28 17 9 1 0 25–49 8 13 20 20 5 0 50–74 1 2 3 10 16 0 75–99 2 1 2 3 13 0 100–124 1 0 0 0 6 13 125–149 1 1 2 2 5 13 150> 0 4 5 5 3 23 49 49 49 49 49 49 Maximum consumption 138 325 278 270 206 277 Santander Santander Guipúzcoa Lugo Santander Canaries Minimum consumption 1 3 5 8 19 101 Barcelona Cuenca Cuenca Cuenca Cuenca Balearic I. Differences cons Max–Min 137 322 273 262 187 176 Source: Authors’ own after Appendix 1. The Canaries have been regarded throughout as a single province. Table 8 Adjusted milk consumption by historical region: Spain, 1865–1981 (l/p/y). Regions 1865 Ra 1925 R 1933 R 1955 R 1965 R 1981 R Andalusia 36.22 8 50.45 12 72.13 8 78.09 8 81.52 12 139.57 9 Aragón 33.66 9 72.68 6 58.00 12 70.21 10 89.70 11 138.96 10 Asturias 129.00 1 235.00 1 220.00 1 261.00 1 180.00 1 196.00 2 Baleares 22.96 14 58.26 10 61.30 11 117.88 5 92.29 10 110.72 14 Canaries 39.79 7 72.53 7 103.85 6 97.06 7 156.00 2 277.30 1 New Castile 30.81 10 64.12 9 72.45 7 63.68 12 117.98 6 165.36 7 Old Castile 78.03 3 161.65 2 121.49 4 126.88 4 130.50 4 186.62 3 Catalonia 29.27 11 88.05 5 104.64 5 65.37 11 101.99 7 122.83 12 Extremadura 47.89 6 34.12 14 49.50 13 75.82 9 95.39 9 154.11 8 Galicia 58.03 4 136.04 4 151.48 3 158.18 2 117.86 5 165.95 6 León 48.94 5 66.71 8 68.03 9 114.29 6 97.28 8 174.51 5 Murcia 27.45 12 56.19 11 44.87 14 58.14 13 71.40 14 125.15 11 Basque Country-Navarra 78.74 2 152.18 3 176.61 2 138.70 3 153.93 3 180.60 4 Valencia 23.41 13 34.76 13 64.69 10 55.29 14 75.11 13 114.01 13 Spain RegionsAverage 48.87 91.62 97.79 105.76 111.50 160.83 Source: Authors own after the indications contained in the text and Appendix3. The Appendix 5 shows these same indicators divided by Regions. a R: Ranking. high consumption of powdered milk in this region.40 Previously, most obvious among these similarities is the persistence of a high the Canaries held a position around the middle of the table, with percentage of consumers – over 70% – in the northernmost regions values not far from the national average. (Cantabrian and Atlantic provinces). There are, however, specific Secondly, Table 8 also shows that the bottom positions in the trends in this regard. In the second half of the 19th century, when ranking were always occupied by the same regions (Valencia, Mur- the percentages of consumer population were generally low – cia and Balearic Islands). Other regions with consumption levels approximately one-third of the population – the contrast is largely that were below the national average but which did not always between the east (less consumers) and the west (more consumers). occupy the bottom tiers are Andalusia, Aragon and Extremadura. In the first third of the 20th century, when consumer population The overall stability of the rankings, however, must not obscure two reached 50% of the total population, the contrast takes a north- other issues. Firstly, adjusted consumption always grew in most south character, with a number of provinces in the interior and regions: between 1865 and 1925, it grew in 13 of the 14 historical the Mediterranean coast in which the proportion of consumers regions; between 1925 and 1933, it grew in 10; between 1933 and remained at 19th-century levels. In the second half of the 20th 1955, it grew in 9; between 1955 and 1965, it again grew in 10; and century, the highest difference is between the Valencian provinces between 1965 and 1981, it grew in all of them. Secondly, there are and their neighbours (Cuenca, Teruel and Albacete), which still pre- significant shifts in the ranking over time. On the one hand, Cat- sented a low percentage of consumers, and the rest of the country, alonia entered the group of regions with the highest consumption with the exception of Huelva and Cádiz. among consumer populations in the first third of the 20th century, Territorial trends are, again, more easily appreciated if historical but then dropped below the national average. New Castile followed regions are taken into consideration. Table 9 shows the percentage an opposing trajectory, staying below the national average until of consumers by region, and, again, the position of each region in 1955 and surpassing it between 1965 and 1981. a ranking. Only the period 1865–1965 will be taken into consider- Let us now analyse consumer populations. Map 3 follows closely ation. By 1981, milk consumption was generalised and the regional the patterns set by the map depicting adjusted consumption. The differences were, therefore, negligible. Between 1865 and 1933, the consumer populations in Asturias, the Basque Country and Navarre had already reached 80% of the total population. In Galicia, consumer population was approxi- 40 The FBS indicates that a good deal of consumption in the Canaries ‘corresponds mately 60% in 1865, but later this percentage grew until it almost to powdered milk (62 g), in amounts which are richer in nutrients than the amount of reached the levels of Asturias, the Basque Country and Navarre (76% fresh milk consumed in other regions’. The FBS also indicates that ‘turning powdered milk into fresh milk at a rate of 1/10, this region becomes the top milk consumer in 1933). These regions, in addition, always occupied the top four (752 g/p/day)’ (INE (1981, vol. 5: 47–48)). positions in the ranking until 1965. At the opposite end of the scale I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 33 1865 1925 <25 <25 25-49 25-49 50-74 50-74 75-99 75-99 100> 100> 1933 1955 <25 <25 25-49 25-49 50-74 50-74 75-99 75-99 100> 100> 1965 1981 <25 100-124 25-49 125-149 50-74 150-174 75-99 175> 100> Map 1. Average observed consumption by province: Spain, 1865–1981 (l/p/y). Source: Authors’ own after Appendix 1. were Andalusia, Aragon and, especially, Murcia and Valencia. In diffusion process was coming to an end. Using our approach, three these four regions, the percentage of consumers was always below long-term trajectories can be appreciated. Firstly, that trajectory the national average, and Murcia and Valencia were always at the of those regions in which consumption levels and percentage bottom of the ranking. Yet again, the overall regularity of the rank- of consumers were already high by the mid-19th century and ing must not draw attention away from some interesting issues: the opening decades of the 20th century (Galicia, Asturias, the (a) the percentage of consumers increased very slowly between Basque Country and Navarre). Secondly, that of another group 1933 and 1955, when the national average barely grew from 53% of regions in which the percentage of consumers and adjusted to 58%; (b) each region followed a distinct trajectory. In Catalonia, consumption levels did not increase to high figures until the for instance, the percentage of consumers increased significantly 1960s (Old Castile and New Castile, Leon and Extremadura, two until 1933. In New Castile, on the other hand, it did so only in the Mediterranean regions – Balearic Islands and Catalonia – and the second half of the 20th century. Canaries). Thirdly, that of those regions which joined the dif- Table 10, finally, presents a coherent chronology of the dif- fusion process latest (Valencia, Murcia, Andalusia and Aragon). fusion of milk consumption in Spain between 1865 and the These trajectories place each region in one of the three stages of 1930s, if we accept that, with adjusted consumption levels above the diffusion process: pioneering or early, intermediate and late 100 l/p/y and percentages of consumer populations around 75%, the adopters. 34 I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 1865 1925 <50 <50 50-74 50-74 75-99 75-99 100> 100> 1933 1955 <50 <50 50-74 50-74 75-99 75-99 100> 100> 1965 1981 <50 100-124 50-74 125-149 75-99 150-174 100> 175> Map 2. Average adjusted consumption by province: Spain, 1865–1981 (l/p/y). Source: Authors’ own after the indications contained in the text and Appendix 4. 5. Conclusion discussion and interpretation of results The present article has analysed this process once more, tak- ing into consideration the usual analytical methods that centred The incorporation of milk into diet was, probably, the most char- on average consumption per capita per year (observed consump- acteristic feature of the European nutritional transition. During this tion), but also two other variables for which data are generally not process, consumption preferences changed substantially and milk available: consumer populations and their average consumption consumption increased significantly. In many regions, milk con- (adjusted consumption). Our approach is supported by three rea- sumption increased more than that of other foodstuffs, including sonable assumptions: (a) consumption took the format of 250-ml animal products. This was also the case in Spain. Milk consumption portions or viable fractions thereof; (b) diffusion patterns of milk was negligible in this country in the 19th century, and reached lev- consumption were similar to those followed by new products and els that could compare with those in other European countries in services; and (c) distribution of consumption among the popu- the second half of the 20th century. lation followed a log-normal distribution. Following these three I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 35 1865 1925 <30 <30 30-49 30-49 50-69 50-69 70-89 70-89 90> 90> 1933 1955 <30 <30 30-49 30-49 50-69 50-69 70-89 70-89 90> 90> 1965 <30 30-49 50-69 70-89 90> Map 3. Proportion of consumers by province: Spain, 1865–1981 (%). Source: Authors’ own after the indications contained in the text and Appendix 4. assumptions, we have analysed again the evolution of milk con- on milk prices at the provincial level and on milk consumption in sumption in Spain between the mid-19th century and the 1980s, cities. and we have reached some conclusions which are consistent with Concerning our case study, at any rate, our main result appears historical records and previous studies. Without a doubt, our solid: the evolution of consumer populations and the consumption methodological approach is open to improvement, but we must levels within these groups changed from region to region and from not forget that the re-evaluation of relevant statistical variables is period to period. That is, although the increase in total average a common practice in economic history (for example in the calcula- consumption was in general closely related to the increase in tion of GDPs and HDIs) and a very useful one for the formulation of the proportion of consumers, the historical sequence of these new hypotheses and the outlining of new research avenues. In order two variables went through two key stages. The first, between to improve our results, it would be useful to compile more informa- the mid-19th century and the 1950s, was characterised by: (a) tion about the observations made by sanitary specialists as well as an increase of milk consumption among the consumer groups; new data on milk consumption in hospitals, army schools and char- (b) the concentration of this process in the stockbreeding northern ity institutions. It would also be useful to collect further information regions and some big cities; and (c) the existence of substantial 36 I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 Table 9 Proportion of consumers by historical region: Spain, 1865–1981 (%). Regions 1865 Ra 1925 R 1933 R 1955 R 1965 R 1981 R Andalusia 26.00 8 32.39 10 45.36 9 43.66 9 72.44 12 95.38 8 Aragón 23.41 9 38.33 8 44.47 10 43.73 12 75.79 11 95.03 12 Asturias 95.68 1 95.52 1 95.52 1 95.04 1 96.60 1 95.33 9 Baleares 12.04 13 33.62 9 45.68 7 74.46 3 84.30 6 91.64 14 Canaries 30.88 7 41.01 6 46.18 6 56.39 7 95.80 2 98.00 2 New Castile 22.02 10 33.34 14 38.31 12 39.84 13 82.46 8 95.19 11 Old Castile 37.11 6 48.92 5 54.94 4 61.70 6 88.57 5 95.40 7 Catalonia 12.61 12 39.47 7 52.73 5 39.61 14 78.78 9 97.06 3 Extremadura 42.00 5 30.10 12 31.12 14 43.10 8 76.66 10 95.46 5 Galicia 60.60 3 65.98 4 76.24 3 75.65 4 90.09 4 95.27 10 León 48.44 4 36.29 3 42.90 11 69.15 5 84.28 7 95.65 4 Murcia 20.09 11 28.80 11 35.99 13 35.56 11 62.65 14 98.09 1 Basque Country-Navarra 74.74 2 82.50 2 85.76 2 85.84 2 95.43 3 95.42 6 Valencia 11.32 14 28.02 13 45.61 8 42.19 10 63.28 13 93.27 13 Spain Regions Average 36.93 45.31 52.92 57.60 81.94 95.44 Spain Communities Average. Source: Authors’ own after the indications contained in the text and Appendix3. Appendix 5 shows these same indicators divided by Autonomous Regions. a R: Ranking. Table 10 some big cities (Langreo, 1995; Domínguez Martín, 1996, 2001; Chronology of the diffusion of milk consumption by historical region: Spain, Hernández and Pujol-Andreu, 2016), and that milk relative prices 1865–1981. in the north were lower (Nicolau and Pujol-Andreu, 2006). Con- Regions Average adjusted Percentage of sumer groups, therefore, could only grow in those areas and cities, consumption consumers and it was the growing consumption among these groups which ≥100 l/p/y ≥75% made the total consumption levels go up. In consequence, inequal- Andalusia 1981 1965 ity widened. Therefore, as late as the 1930s, sanitary specialists still Aragon 1981 1965 observed that large groups of the population did not consume milk Asturias 1865 1865 Balearic Islands 1955 1955 at all (Jiménez and Jiménez, 1934). Canaries 1933 1955 Regarding the second phase of diffusion of milk consumption New Castile 1965 1965 in Spain, previous studies and historical records also support our Old Castile 1925 1965 results. In the 1950s, milk was already accepted as a basic food- Catalonia 1965 1965 Extremadura 1965 1965 stuff, and the number of potential consumers was thus high. It Galicia 1925 1933 must, therefore, not come as a surprise that the proportion of con- Leon 1955 1965 sumers and the average consumption among this group increased Murcia 1981 1981 in the following years, when milk production and interregional Basque Country-Navarre 1925 1865 trade increased, and household income also increased. Several of Valencia 1981 1981 Spanish average 1955 1965 previous studies have already shown that this increase was affected by a number of different innovations: the introduction of new ani- Source: Authors’ own after own estimations. mal breeds, the distribution of milk in cistern trucks, the creation of new milk processing plants and the emergence of new products regional differences. The second stage, between the 1950s and the (pasteurised, sterilised, bottled and powdered milk) (Domínguez 1980s, was characterised by: (a) a generalisation of consumption Martín, 2003; Collantes, 2015; Hernández and Pujol-Andreu, 2016). as a result of the sharp growth in the number of consumers first, As previously noted, different socioeconomic factors also promoted and of consumption among these later; and (b) the progressive milk consumption: for instance, the accentuation of population decrease in regional differences. Probably, differences according movements towards wealthier regions and especially cities. Thus, to income also decreased over time, but this aspect has not been it can reasonably be inferred that: (a) milk consumption increased considered in the present study. especially in those regions in which, in the early 1960s, as much Concerning the first stage, some historical records provide evi- as 50% of the population still consumed no milk at all; (b) that dence for a small proportion of consumers and a low average regional differences tended to disappear; and (c) that the geograph- consumption (adjusted average) in the second half of the 19th ical connection between consumption and production disappeared. century, and also for the important role played by this variable This is also attested by the correlation coefficients that reflect the in the increase of observed consumption up to the 1930s. Those association between number of cows per 1000 inhabitants and records also provide evidence for wide regional and socioeco- the percentage of consumer population. Taking all provinces into nomic differences. The information provided by late 19th- and account, in 1865 this coefficient was 0.91 and in 1925 it was 0.82; early 20th-century medical-sanitary surveys is illustrative con- later, it decreased to 0.66 in the mid-1950s and to 0.34 in 1981. cerning the small number of consumers, as are the reports on the stock-breeding sector in 1891, and the more disperse data on family budgets (Ballesteros, 1997; Pujol-Andreu and Cussó, Acknowledgments 2014). Previous research also shows that milk consumption became widespread first among children and young people (Cussó and This work was carried out within the framework of the project Garrabou, 2004; Nicolau et al., 2010; Muñoz Pradas, 2016; Nicolau HAR 2016-76814-C2-1-P(AEI/FEDER,UE). We wish to thank James et al., 2010), and that differences according to income in this group Simpson, Josep Lluis Raymond and David Govantes (translation) were substantial (Carrasco Cadenas, 1934). Finally, other stud- for their comments. Remarks made by the referees have been very ies have shown that the increase in the availability of milk in helpful for improving the manuscript. Any mistakes or misunder- those years chiefly affected the stockbreeding northern areas and standings in these pages are our own. I. Hernández Adell et al. / Investigaciones de Historia Económica - Economic History Research 15 (2019) 23–37 37 Appendix A. Supplementary data Hernández, I., Pujol-Andreu, J., 2016. Economic growth and biological innovation: the development of the European dairy sector, 1865–1940. Rural History: Econ- omy, Society, Culture 27 (2), 187–212. Supplementary data associated with this article can be found, in INE – Instituto Nacional de Estadística, 1969. Encuesta de Presupuestos Familiares the online version, at doi:10.1016/j.ihe.2017.03.008. (Marzo 1964-Marzo 1965). INE, Madrid. INE – Instituto Nacional de Estadística, 1983. Encuesta de Presupuestos Familiares 1980-81, 5 vols. INE, Madrid. References International Dairy Federation, 1982. Consumption Statistics for Milk and Milk Prod- ucts 1980 (Including Summary 1966–1980), Bulletin de la Fed. Internationale de Ajuntament de Girona, 1934. El proveïment de llets a Girona: l’Alcalde de Girona a Laiterie-International Dairy Fed., pp. 144. l’opinió pública. Gráficas Darío Rahola, Girona. Jiménez, F., Jiménez, M., 1934. La alimentación en la provincia de Jaén. Deducciones Annuaire Statistique de la Suisse, 1930. Ofice Fédéral de la Statistique, Bern. sanitarias. Archivos de Medicina, Cirugía y Especialidades XXXVII (19), 503–506. AEB – Anuario Estadístico de Barcelona, 1902. Ayuntamiento de Barcelona, Junta Consultiva Agronómica, 1892. La Ganadería en España: avance sobre la riqueza Barcelona. pecuaria en 1891. Tip. de L. Péant é Hijos, Madrid. Anuario Estadístico de España, 1943-54. Madrid. Junta General de Estadística, 1868. Censo de la ganadería en España según el recuento Asociación General de Ganaderos del Reino (AGGR), 1925. Leche, queso y manteca. verificado en 24 de septiembre de 1865. Madrid. Estadística de la producción en España, Madrid. Kauffman, R.J., Techatanasoontorn, A., 2009. Understanding early diffusion of digital Atkins, P., 1992. White poison? The social consequences of milk consumption, wireless phones. Telecommunications Policy 33, 432–450. 1850–1930. Social History of Medicine 5, 207–227. Kjærnes, U., 1995. Milk: nutritional science and agricultural development in Atkins, P., 2005. Fattening children or fattening farmers? School milk in Britain, Norway, 1890–1990. In: den Hartog, P. (Ed.), Food Technology, Science and 1921–1941. Economic History Review, LVIII 1, 57–78. Marketing: European Diet in the Twentieth Century. Tuckwell Press, Forlag, Ayuntamiento de Madrid, 1897–1939. Boletín del Ayuntamiento de Madrid. Ayun- pp. 103–115. tamiento, Madrid. Langreo, A., 1995. Historia de la industria láctea española: una aplicación a Asturias. Ballesteros, E., 1997. Niveles de vida en España, siglos XIX y XX. Unpublished doctoral Ministerio de Agricultura, Pesca y Alimentación, Madrid. thesis, Universidad Complutense de Madrid, Madrid. Lawrence, R., 1988. Applications in economics and business. In: Crow, E., Shimizu, Barciela, C., Giradlez, J., Grupo de Estudios de Historia Rural, López, I., 2005. Sector K. (Eds.), Lognormal Distributions. Theory and Applications. Marcel Bekker Inc., agrario y pesca. In: Carreras, A., Tafunell, X., Coord. Estadísticas Históricas de New York, pp. 230–264. España, siglos XIX y XX. Fundación BBVA, Bilbao, pp. 245–356. Luís y Yagüe, R., 1903. Bromatología popular urbana. Idamor Moreno, Madrid. Calatayud, S., Medina, F.J., 2017. Leche sin prados: los factores ambientales e institu- Mas Alemany, J., 1933. Memoria corresponent a la labor dels Serveis de Sanitat Vet- cionales en el consumo lácteo (Valencia, 1870-1936). Ayer-Revista de Historia erinària practicada l’any 1932. Ajuntament de Barcelona, Perfectus, Barcelona. Contemporánea 105, 157–185. Mas Alemany, J., 1935. Els serveis del cos de veterinària a l’any 1934. Ajuntament de Calot, G., 1988. Curso de Estadística Descriptiva. Paraninfo, Madrid. Barcelona, Barcelona. Carrasco Cadenas, E., 1934. Lo que se come en España. Interés sanitario de este Ministerio de Agricultura, 1934. Censo de Ganadería en España en 1933. Madrid. problema. Archivos de Medicina, Cirugía y Especialidades XXXVII (núm. 25, 23), Ministerio de Agricultura, 1954–1960. Resumen estadístico de la producción, des- 669–675. tino y valor de la leche. Dirección General de Ganadería, Madrid. Carro García, S., 1917. Alimentación en la edad escolar: Cantinas escolares. Enrique Moreno, L.A., Sarría, A., Popkin, B.M., 2002. The nutrition transition in Spain: a Teodoro, Madrid. European Mediterranean country. European Journal of Clinical Nutrition 56, Casado de la Fuente, C., 1931. Abastecimientos de leche a las grandes poblaciones. 992–1003. Giménez Moreno, Madrid. Muñoz Pradas, F., 2011. Consumer populations and nutritional transition in Spain Collantes, F., 2015. Más allá de los promedios: patrones de segmentación del con- in the 20th century: a methodology for their reconstruction. Histoire & Mesure sumo de productos lácteos en España, 1964-2006. Investigaciones de Historia XXVI (2), 131–173. Económica 11, 103–115. Muñoz Pradas, F., 2016. La implantación de las Gotas de Leche en España (1902- Contreras Hernández, J., 2002. Alimentación y sociedad. Sociología del consumo 1935): un estudio a partir de la prensa histórica. Asclepio 68 (1 (Enero-Junio)), alimentario en España. In: Gómez Benito, C., Jesús González, J. (Eds.), Agricultura 131–176. y Sociedad. En el cambio de siglo. MacGraw Hill, Madrid, pp. 297–332. Murcott, A., 1999. Scarcity in abundance: food and non-food. Social Research 66 (1), Cussó, X., Garrabou, R., 2004. L’Escola del Bosc. Un referent pioner a la transició 305–339. nutricional moderna a Catalunya. Estudis d’Història Agrària 17, 497–512. Nicolau, R., Pujol-Andreu, J., 2006. Variaciones regionales de los precios de consumo Doaso Olasagasti, M., 1931. Abastecimiento de leche en las grandes poblaciones. y de las dietas en España, en los inicios de la transición demográfica. Revista Asociación General de Ganaderos de España, Madrid. de Historia Económica, Journal of Iberian and Latin American Economic History Domínguez Martín, R. (Ed.), 1996. La vocación ganadera del norte de España. Del XXIV (3), 521–554. modelo tradicional a los desafíos del mercado tradicional. MAPA, Madrid. Nicolau, R., Pujol-Andreu, J., 2008. Los factores condicionantes de la transición nutri- Domínguez Martín, R., 2001. Las transformaciones del sector ganadero en España cional en la Europa occidental: Barcelona, 1890-1936. Scripta Nova XII (265). (1940-1985). Ager 1, 47–83. Nicolau, R., Pujol-Andreu, J., Hernández, I., 2010. Milk, social acceptance of a new Domínguez Martín, R., 2003. La Industria Láctea en España, 1830-1985. In: Barciela, food in Europe: Catalonia, 19–20th centuries. Dynamis: Acta Hispanica ad C., Di Vittorio (Eds.), Las industrias agroalimentarias en España e Italia durante Medicinae Scientiarumque Historiam Illustrandam 30, 119–140. los siglos XIX y XX. Pub. Universidad de Alicante, Alicante, pp. 457–495. Orland, B., 2005. Milky ways. Dairy, landscape and nation building until 1930. In: Duval, Y., Biere, A., 2002. Product diffusion and the demand for new food products. Sarasúa, C., Scholliers, P., Molle, L.V. (Eds.), Land, shops and kitchens. technology Agribusiness 18 (1), 23–36. and the food chain in twentieth-century Europe. CORN Publication Series, vol. FENIL, 2008. Plan de Nutrición, Salud y Comunicación de Productos 7. Brepols Publishers, Turnhout, Belgium, pp. 212–254. Lácteos 2007-2010, PDF document available at: www.revistafrisona. Orland, B., 2007. Bad habits and liquid pleasures milk and the alcohol abstinence com/temas/leche/Plan Fenil07-10.pdf movement in late 19th century Germany. Food and History 5 (2), 153–169. Floud, R., Fogel, R.W., Harris, B., Hong, Ch., 2011. The Changing Body. Health, Nutri- Pérez Castroviejo, P.M., Martínez Castroviejo, I., 1996. La alimentación de los pobres. tion and Human Development in the Western World Sice 1700. Cambridge Estrategias de gasto alimentario y la dieta en la Santa Casa de Misericordia de University Press, Cambridge. Bilbao, 1840-1940. Ayuntamiento, Área de Cultura y Turismo, Bilbao. Fogel, R.W., 1992. Second thoughts on the European escape from hunger: camines, Popkin, B.M., 1993. Nutritional patterns and transitions. Population and Develop- chronic malnutrition, and mortality rates. In: Osmani, S.R. (Ed.), Nutririon and ment Review 19 (1), 138–157. Poverty. Clarendon Press, Oxford, pp. 243–286. Powell, L.R., Norman, H.D., Dickinson, F.N., 1975. Sire differences in sex ratio of García Dory, M., Martínez Vicente, S., 1988. La ganadería en España. Alianza Editorial, progeny. Journal of Dairy Science 58 (11), 1723–1726. Madrid. Puente Fernández, L., 1992. Transformaciones agrarias en Cantabria: el proceso de García Izcara, D., 1927. El problema de la leche en Madrid. In: Revista de Higiene y especialización del ganado vacuno, 1860-1930. Servicio de Publicaciones de la Sanidad Pecuarias, Madrid, Diciembre de 1927, pp. 959–961. Universidad de Cantabria, Santander. Generalitat de Catalunya, 1937. El proveïment de Barcelona de llet. Archivo Nacional Pujol-Andreu, J., Cussó, X., 2014. La transición nutricional en Europa occidental, de Cataluña, Fondo Generalitat de Catalunya (II República). Cataloging unit: 1865-2000: una nueva aproximación. Historia Social 80, 133–150. 7589. Rodríguez Ocaña, E., Ortiz Gómez, T., García Duarte Ros, O., 1985. Los consultorios Gresa de Mirambell, A., 1929. Topografía Médica de Navarcles. Manuscript. de lactantes y gotas de leche en España. Jano - Medicina y Humanidades, 24 de Grigg, D., 1995. The nutritional transition in Western Europe. Journal of Historical octubre–4 de noviembre de 1985 663 (H), 1066–1072. GeoFigurey 22 (1), 247–261. Rogers, E.M., 2003. Diffusion of Innovations, fifth edition. The Free Press, New York. Grupo de Estudios de Historia Rural, 1985. Contribución al análisis histórico de la Vallribera Puig, P., 2000. Les topografies mèdiques de la Reial Acadèmia de Medicina ganadería española 1865-1929. In: Garrabou, R., Sanz, J. (Eds.), Historia Agraria de Catalunya. Publicacions del Seminari Pere Mata de la Universitat de Barcelona, de la España Contemporánea. Expansión y crisis (1850-1900), vol. 2. Editorial Barcelona. Crítica, Barcelona, pp. 229–278. Williamson, J.G., 1965. Regional inequality and the process of national development: Grupo de Estudios de Historia Rural, 1991. Estadísticas históricas de la producción a description of patterns. Economic Development and Cultural Change 13 (4), agraria española. 1859-1935. Ministerio de Agricultura, Pesca y Alimentación, 3–52. Madrid.