Caribbean Studies ISSN: Instituto de Estudios del Caribe Puerto Rico

Transcription

1 ISSN: Instituto de Estudios del Caribe Puerto Rico Velyvis, Kristen; Thompson-Colón, Theresa; Winsborough, Halliman PUBLIC USE SAMPLES OF 1910 AND 1920 PUERTO RICO CENSUSES, vol. 35, núm. 2, julio-diciembre, 2007, pp Instituto de Estudios del Caribe San Juan, Puerto Rico Available in: How to cite Complete issue More information about this article Journal's homepage in redalyc.org Scientific Information System Network of Scientific Journals from Latin America, the Caribbean, Spain and Portugal Non-profit academic project, developed under the open access initiative

2 3 Public Use s of 1910 and 1920 Puerto Rico Censuses Kristen Velyvis Theresa Thompson-Colón Halliman Winsborough Ab s t r a c t This article describes new public use data samples made from the first and second ordinary U.S. Censuses of Puerto Rico and how and why these samples were created. The digital files were made by sampling and transcribing information from archival microfilms of the original census schedules. Textual information was transcribed verbatim. Procedures and coding conform to the Integrated Public Use Microdata Series (IPUMS) standard for census samples. Versions of the files are currently available from IPUMS ( edu/usa/) and the Inter-University Consortium for Political and Social Research (ICPSR) Data Archive ( Keywords: Puerto Rico, U.S. census data, public data files, historical, IPUMS, population characteristics Re s u m e n Esta monografía describe datos nuevos para uso público de muestras extraídas de los primeros dos censos ordinarios de los Estados Unidos en Puerto Rico. Además, detalla cómo se crearon estas muestras y la razón de su creación. Los archivos digitales se hicieron por medio de muestreo y trascripción de la información en microfichas de los formularios originales del censo. La información textual fue transcrita directamente. Los procedimientos y la codificación de datos son conformes a los estándares de muestras de censos de los Servicios Integrados de Microdatos para Uso Público (IPUMS, por sus siglas en inglés) ( Versiones de los archivos están disponibles en IPUMS y en la base de archivos del Consorcio Inter-Universitario para la Investigación de Ciencias Políticas y Sociales (ICPSR, por sus siglas en inglés) ( icpsr.umich.edu). Palabras clave: Puerto Rico, datos del censo de los Estados Unidos, archivos de datos públicos, históricos, IPUMS, características de población

3 4 Ve ly v i s, Th o m p s o n-co l ó n, Wi n s b o r o u g h Ré s u m é Cette étude présente une description des échantillons de données disponibles au grand public émanant des premier et deuxième recensements ordinaires des États-Unis à Porto Rico d une part; d autre part, il examine comment et pourquoi ces échantillons ont été crées. Les archives digitales ont été preparées selon les données statistiques et les transcriptions des informations microfilmées provenant des documents originaux du recensement. Les informations textuelles sont des transciptions directes. Les procédures et la codification sont conformes aux normes des Services d Integrations de Microinformations d utilisation publique, connu par son sigle en anglais IPUMS (Integrated Public Use Microdata Series). Des exemplaires des documents sont disponibles sur le site de IPUMS ( et aux archives du Consortium Interuniversitaire pour les Recherches en Sciences Politiques et Sociales (ICSPR) ( Mots-clés: Porto Rico, données du recensement des États- Unis, archives des informations publiques, historiques, IPUMS, caractéristiques de la population Received: 19 April 2007 Revision received: 17 October 2007 Accepted: 19 October 2007 Introduction The 1910 and 1920 Puerto Rican Census Project consisted of a largescale data entry activity undertaken at the University of Wisconsin- Madison with the collaboration of the Integrated Public Use Microdata Series (IPUMS) at the University of Minnesota. The project produced two sets of public use sample files of individual and household records drawn from the Puerto Rican censuses of 1910 and This article describes how these public use files were produced, why this project was undertaken, and the final data that are now available. Project Overview In brief, this project produced two sets of data by sampling and transcribing information from the microfilms of the original schedules of the first and second Censuses of Puerto Rico taken by the U.S. Bureau of the Census. The first public use files are held by the Inter-University Consortium for Political and Social Research (ICPSR) Data Archive, and reflect the data exactly as they were transcribed from the original

4 Public use samples of 1910 and microfilms. Textual information was transcribed verbatim. These data can be used by historians and others as reflections of the complex specificity of the time and place from which they were collected. The second public use files were created using coding procedures that conform to the IPUMS standard for census samples. They are analogous and strictly comparable to the U.S. public use samples created as part of the IPUMS project (Ruggles and Sobek 1995). These files are most useful for making comparisons across time and place. This project was undertaken by Alberto Palloni, Professor of Sociology; Halliman Winsborough, Emeritus Professor of Sociology; and Francisco Scarano, Professor of History, all faculty members at the University of Wisconsin-Madison. Each principal investigator (PI) was responsible for different facets of the project. Professor Winsborough oversaw the data entry activities and acted as liaison between the PIs and other collaborators on the project. Later, he primarily saw that the documentation was properly prepared and dealt with errors in the data found in the intensive analysis phase. Professor Scarano provided deep historical knowledge of Puerto Rico. He insured the proper interpretation of words and circumstances throughout the process and added substantive weight to analysis. Professor Palloni had primary oversight of the project, but his major contribution was to the analytic part of the research using the data. The PIs contracted with the University of Wisconsin Survey Center (UWSC) to establish a data entry and processing operation under the guidance of the Historical Census Project (HCP) staff at the Minnesota Population Center, University of Minnesota. The HCP staff were also sub-contracted to undertake the allocation of missing data so that the files would be comparable to mainland U.S. IPUMS files for the same dates. They also undertook the standardization necessary for inclusion of the samples into the IPUMS database. Dr. Ronald Goeken, Research Associate, was the main contact on this project at the Minnesota Population Center. How the Files Were Produced The project took about four and a half years to complete starting in January of There were four main project stages: Pre-data entry, data entry, post-data entry, and analysis. The first six months of the project were spent in the pre-data entry stage, which included acquiring the microfilm of the censuses, buying microfilm readers, setting up software and Spanish-speaking operators. It also included training for

5 6 Ve ly v i s, Th o m p s o n-co l ó n, Wi n s b o r o u g h the data entry operators in the entry programs and in the interpretation of the manuscript materials. 1 The data entry stage ensued. Four data entry operators worked full time for 15 months to enter all of the data. Six months into the data-entry stage, the post data-entry processing began. This included data checking and verification, data coding, and format standardization to get the data prepared for IPUMS specifications. The HCP staff also conducted data allocation for the IPUMS data files. The allocation process assigned values for missing, illegible or inconsistent data points in three main ways: Logical edits, hot deck allocation and cold deck allocation. Logical edits included edits such as assigning sex as female for the wife of a male householder. If logical edits could not be made, hot deck allocation was used to replace missing data with data from a randomly chosen similar case. If a similar case could not be found, cold deck allocation was used to assign a value from a pre-determined distribution or to assign a modal value (Integrated Public Use Micro-data Series USA 2007). Flags were created for allocated data points. Finally, UWSC delivered preliminary datasets in May of The analysis of the data continued under the grant until June How These Data Can Be Useful These data samples are valuable because of their uniqueness, the compelling time period in Puerto Rico they cover, their comparability with data from other places, and their completeness. Data from Puerto Rico in the early 1900 s are rare and valuable since the time period is of great historical interest. The period covered by these data samples includes the origins of one of the most sudden and rapid demographic changes to take place among developing countries in general and Latin American countries in particular. The period is also one of deep transformations in Puerto Rico s economy. As the PI s point out, The case of Puerto Rico is unique. It is the result of a historical experiment that brought together a blend of contrasting economic, social and cultural influences. Up until 1898 the history of Puerto Rico was intimately connected to Spanish colonial domination, and its most important historical features replicated quite closely characteristics encountered in other parts of the Latin American and Caribbean regions under Spanish colonial rule. Starting in 1898, however, Puerto Rico became a territory legally and administratively dependent on the US. The

6 Public use samples of 1910 and economic infrastructure was completely revamped, with formidable effects on the social class composition and the regional distribution of the population. Furthermore, the ideological and cultural suprastructure experienced an overhaul that led to a contrast between the cultural and social experience of the island and that of the rest of Latin America. More than figuratively, the political incorporation of Puerto Rico into the US at the turn of the century constituted an extraordinary historical experiment with a number of consequences that historians, sociologists, economists and demographers cannot afford to ignore. The changes experienced and their impact on a number of conditions led also to a social, economic and demographic regime that is in stark contrast to those created in the rest of Latin America. Yet, simultaneously, Puerto Rico was the site of a historical experience quite dissimilar from the one in the US (Palloni, Scarano and Winsborough 2000:3). The data that have been compiled offer detailed individual information and associated community and geographic location identifiers and characteristics from Puerto Rico during this time; this is powerful information for researchers. The data can be used to study social and economic conditions in specific regions of the island and the social and economic processes that took place during this historical moment. They can also help us understand how these conditions affected the structure of the labor force, household organization, population distribution and differential growth of cities (Palloni et al. 2000). The two samples provide rich data for many types of researchers, from demographers interested in population dynamics to social and economic historians who study the economic and social conditions of the population in developing areas in the aftermath of colonial domination (Palloni et al. 2000:1). The following two sections of this paper lay out in more detail how and why these data are especially valuable and useful to demographers and socioeconomic historians. Of Interest to Demographers The richness of these data and the importance of the time period they cover create a boon for demographers. These data are a rich resource for the study of demographic processes that took place at the turn of the century, at the onset of an irreversible demographic transition. They allow for the testing of theories of fertility decline, the role of

7 8 Ve ly v i s, Th o m p s o n-co l ó n, Wi n s b o r o u g h public health interventions, and the factors that generate changes in residential preferences (Palloni et al. 2000). They also allow scholars to describe levels, patterns and trends of mortality, fertility and nuptiality at the start of the century, and compare these with those experienced in the U.S. and in the rest of the Latin American and Caribbean region (Palloni et al. 2000). In addition, the 1910 Puerto Rican census, unlike the majority of censuses ever taken, contains information on duration of marriage. This allows analyses of fertility that are unique and of unusual robustness (Palloni et al. 2000). The uniqueness of the Puerto Rican situation and data can also be exploited to look at other sociodemographic issues such as race, which during this period was considered differently in Puerto Rico than in the U.S. These data for this time of rapid transition also allow scholars to reconsider Modernization Theory. Modernization Theory posits that as a consequence of transformations in Puerto Rico s economy, labor markets and social relations, and the weakening of social and cultural order, household and family composition must change as well, passing from an extended form to a nuclear form. Researchers can now determine whether such a transition ever took place at all, whether the household composition at the outset of the process was what would be expected, and if a change did occur, whether it is possible to distinguish the influence of changing demographic constraints from the influence of changing values (Palloni et al. 2000). Of Interest to Social and Economic Historians There are three potential uses of these data that could be of particular interest to social and economic historians. The first use will, for the first time, allow scholars of twentieth-century Puerto Rico to answer questions that have remained bothersome for decades. That is, beyond the general implications of migrant displacements, how specifically did the economy of Puerto Rico under U.S. administration, with its rapid shift from coffee to sugar and tobacco planting, change the residential and family organization of working groups? And how did families deal with a new labor market, and in particular, with the entry of substantial numbers of women into the waged labor force (Acosta- Belén 1986)? The second use will allow scholars to exploit complementary landholding and farm tax data to look at who among smallholders of the coastal plains of Puerto Rico took advantage of the American tariff

8 Public use samples of 1910 and subsidy to become satellite cane farmers in the plantation economy (Scarano 1990). Scholars can also look at the effects this had on farmers demographic profiles. These are two specific issues that occurred within the broad historical emergence of newer working classes. They focus on the processes of the proletarianization of the peasantry who were previously involved in coffee production and on their associated transformation into small land holders or laborers (Palloni et al. 2000). The third use will allow scholars to explore what realities of social interaction and family formation underlie the marked increase in the percentage of whites and the related decline of blacks in the early twentieth century. While some scholars easily dismiss certain census categories as social constructions, these data allows research into what lies beneath these changes (Palloni et al. 2000). These questions and more could influence new research on Caribbean and Latin American plantation spaces, which were quite comparable to Puerto Rico. Economic and social historians of those areas that Wagley calls plantation America can benefit substantially from the availability of the Puerto Rican data. It has been a precept of comparative historians for some time that in this part of the American tropics the limits of social and economic life were primarily determined by the human ecology associated with one or another of the after-dinner crops, especially those that prevailed in Puerto Rico: Sugar, coffee and tobacco (Wagley 1957; Steward 1956). For these scholars, these data will allow the examination of as yet untested hypotheses about the comparability of plantation economies and of the working populations shaped by them (Palloni et al. 2000). The Data in More Detail The samples that were created are available for public use in two forms. The 1910 and 1920 files that reflect the data as they were transcribed verbatim are held in the ICPSR Data Archive and are available at These files include all variables unedited and unallocated for missing values. These versions of the files contain all original text variables as well. 2 The second set of 1910 and 1920 files was created in accordance with IPUMS data content and format specifications. The HCP staff reviewed the unedited data and carried out data edits and allocation procedures so that the data conform to their standard for census samples. The IPUMS versions of the samples will be released as part of

9 10 Ve ly v i s, Th o m p s o n-co l ó n, Wi n s b o r o u g h their public archived collection available at edu/usa/ (IPUMS is responsible for preserving these versions of the data files and for providing support to data users.) Both the ICPSR and the IPUMS versions of the data contain individual level and household level data. The population covered by the samples includes the entire population of the island of Puerto Rico in 1910 and The data files consist of 1-in-10 samples for most of the island. The exceptions are the coffee-producing regions and the municipio, or county, of Loíza (a traditionally African-descent municipio), which were over-sampled at a rate of 1-in-5 from each census year. These very high sampling rates resulted in a total of 298,374 cases being entered for the two censuses combined 207,806 in the main sample; 84,796 in the coffee-producing regions; and 5,772 in the county of Loíza. Table 1 presents the distribution between the two censuses, the main sample and the over-sample. Table 1. Distribution of the 1910 and 1920 Puerto Rican s Census Year Main Coffee Region Over- Loíza Over- Total ,124 40,092 2, , ,682 44,704 3, ,483 Total 207,806 84,796 5, ,374 An interesting feature of the data is the inclusion in the samples of dwelling units clustered in particular areas that are identified as solares in the 1910 census and posteriores in the 1920 census. These dwelling arrangements existed in Puerto Rico during this time, but did not exist in the mainland U.S. In 1910, a solar was defined as a clustering of dwellings in an urban area where a land-tract had been subdivided and had one dwelling built behind another. In 1920, the Census Bureau was more specific in its directions and indicated that if a house was situated behind a main house that was facing the street, this house should be noted as a posterior. The sampling of solares and posteriores allows researchers to examine clustering of family relationships within this type of geographic grouping. Another interesting aspect of these data is the complete enumeration of the military and naval institutions in the area of San Juan in A total of 1,981 individuals who made up the military and naval population in the 1920 census were included in a separate spreadsheet available with the data held by ICPSR.

10 Public use samples of 1910 and Sampling Issues The 1910 and 1920 samples from the Puerto Rican Censuses are probability samples created using both clustering and stratification methods. How to deal with the analytic complications that such methods present depends in great measure on what one wants to learn from the data. This section will discuss the issues, their impact on estimates derived from the data, and ways to cope with the sampling structure. The files are cluster samples in two ways. They were constructed by randomly selecting a sample line from every fifty-line enumeration form on the microfilm. If a household head appeared on that line or on any of the four succeeding lines, that person and all members of his/ her household were taken into the sample. More than one household could be drawn from the five-line cluster. Thus, lines are clustered on the page and persons are clustered within households. The files are stratified samples in three ways. First, households in coffee-growing barrios, or neighborhoods, and those in a traditionally African-descent municipio were sampled at twice the rate of other households in order to yield samples of these relatively rare people that are of sufficient size to permit robust analysis. Thus, in the affected barrios, nine succeeding lines rather than four are added to the cluster of lines. Second, choosing to draw sample lines from each page of the enumeration forms effectively stratifies the sample by geography since pages cover contiguous households or at least neighborhoods that are proximate one to another. Third, in preparing to draw the sample the research team discovered that these files include instances of solares and posteriores. These residences resemble urban compounds since they include a number of more or less separate households at a single address. The Census Bureau established rules to distinguish the separate households, but during initial inspections by the research team it was unclear whether the rules had been effective in the field. Early in the project, then, the decision was made to include everyone in a selected solar or posterior in the sample, regardless of whether they were in the initially selected household or in another unit in the compound and without regard to whether that additional unit would fall within the sample or not. This decision ensured there would be data to deal with enumerators problems with the rules if necessary (it turned out not to be much of a problem) and to permit study of this interesting and rather unique residence type. In most cases this last stratification is easy to manage: Omit the extra households, and thus return to the initial sampling design. There

11 12 Ve ly v i s, Th o m p s o n-co l ó n, Wi n s b o r o u g h is a variable in each record that distinguishes the added households from those that conform to the initial sampling design, and researchers can drop the extra households at the solar or posterior address based on this variable. One way to accomplish this exclusion is to drop all cases in which that variable, SOLRULE, is equal to 0. Dealing with the over sampled coffee-growing and African-decent barrios is more difficult to manage. The variable DATANUM indicates the rate at which the household was sampled. A value of 1 indicates membership in the basic 10% sample group, a value of 2 indicates membership in the 20% sample of the African-descent region, and a value of 3 indicates membership in the 20% sample of the coffeegrowing region. Many people deal with stratified samples by constructing weights and using the weights in their data analysis. Here, one might assign a weight of 1 to cases coded 1 on DATANUM and a weight of one-half to cases coded 2 or 3 on that variable. For many purposes a sample so weighted can be regarded as a 10% sample of the population. Thus, the total of the weighted sample size multiplied by 10 yields an estimate of the population total. Using this method, the sample estimates the 1910 population of Puerto Rico as 1,112,360 persons. The Censuses reports the figure of 1,118,012. The estimate is within 0.5% of the Census count. Is that departure within a reasonable confidence interval? Answering that question requires computing a variance as well as an estimate. The issue of estimating variances is addressed below. Clustering is the most difficult to manage when making probability statements about the accuracy of estimates calculated from these data. Cluster samples can be either more or less powerful than simple random samples depending on the variable considered. If the withincluster variance of a variable is greater than it would be for randomly drawn clusters from the same population, the sample will be more powerful. If the within-cluster variance is smaller than for randomly drawn clusters, it will be less powerful. Households typically contain variables of both kinds. Age and sex within households is usually more heterogeneous than within randomly drawn clusters while measure of wealth, education, race, and other social status variable are usually more homogeneous within households than within randomly selected clusters. Figuring out the variance of the parameters of a multivariate model estimated from such data is very difficult; sometimes impossible to do analytically.

12 Public use samples of 1910 and One way to deal with this problem is by providing the user with the possibility of producing an empirical estimate of the variance of any calculation s/he wishes to make. To do this, one can take advantage of the fact that the samples for these census records are drawn independently from each page of the enumeration forms, thus one can allocate each page randomly to a select number of samples and in this way have several independent samples of the data. If the calculation of interest is made on each sample separately, a variance for the quantity can be estimated empirically and the sampling distribution of the quantity investigated. Then the resulting variance can perhaps be used for statistical tests and confidence intervals. The variable PAGENUM provides the relevant page number. As an example of this procedure, here is a comparison of the simple random sample confidence interval for the population total and that derived from an empirical estimate of its variance. To accomplish this empirical estimate of the variance, each enumeration page in the 1910 sample was randomly assigned to one of ten separate groups. Each group, then, is a one-in-a-hundred sample of the population of Puerto Rico in The population size for each group is calculated, and its mean is 11, The estimated variance of the population size is 211, and the standard deviation is If this had been a simple random sample with p=0.1, the variance would be Np(1-p) or 100,621 and the standard deviation would be Thus, the sampling design increases the standard deviation of the population count by about 50%. The sampling procedures used seem to have produced samples that reflect the census populations fairly well. The following tables show how the sample data compare to the census data in a number of ways. Using the procedures described above, weights have been used to deal with over-sampling and extra cases entered from solares and posteriores have been dropped, that is, extra cases that would not have fallen within the sample under the original sampling plan. The results show how the samples reflect the census populations. The first table shows a comparison on the sample data and census data by race for 1910 (Table 2). Table 3 shows the same comparison for The next table shows comparisons of the sample data and census data by sex for each year (Table 4). Tables 5a, 5b, 6a and 6b show comparisons of the sample data and census data by age and sex. Finally, Tables 7 and 8 show a comparison between the sample data and the census data by population per municipio. These tables show

13 14 Ve ly v i s, Th o m p s o n-co l ó n, Wi n s b o r o u g h the number of families and the number of individuals per municipio found in the census data and the sample data. Italicized cases indicate municipios where the comparison between the sample and the census data results in Chi-square values higher than 10 with one degree of freedom. Table 2. Comparison of 1910 Census and Weighted Distributions by Race Race Census Population Percent of Population Weighted Percent of White 732, % 72, % Black 50, % 5, % Mulatto 335, % 34, % Other % % Total 1,118, % 111, % Notes: Census (U.S. Bureau of the Census 1913); ICPSR sample data (ICPSR04343-v1). data do not include solares. Table 3. Comparison of 1920 Census and Weighted Distributions by Race Race Census Population Percent of Population Weighted Percent of White 948, % 94, % Black 49, % 4, % Mulatto 301, % 28, % Other % % Total 1,299, % 128, % Notes: Census (U.S. Bureau of the Census 1913); ICPSR sample data (ICPSR04344-v1). data do not include posteriores.

14 Public use samples of 1910 and Table 4. Comparisons of 1910 and 1920 Censuses and Weighted Distributions by Sex Year Sex Percent of Population Percent of 1910 Men 49.85% 49.93% Women 50.15% 50.07% 1920 Men 49.84% 49.93% Women 50.16% 50.07% % % Notes: Census (U.S. Bureau of the Census 1913 & 1922); ICPSR sample data (ICPSR04343-v1 and ICPSR04344-v1). data do not include solares or posteriores. Quality of the Data Careful procedures were followed to assure high-quality data. The data-entry procedure for the 1910 and 1920 samples emphasized accuracy and efficiency, rather than speed. Data checking and data verification were also scrupulously undertaken. The research team first checked the data entry by running the raw entered data through a logical check program, which highlighted potential data entry errors based on the relationship between different fields. For example, if the variable sex was male but the variable relationship corresponded to a female, the team compared the flagged data to the original microfilm data and corrected the information. After the checking process was completed, the cleaned data were run through the verifying process. For quality control, this entailed double-entering one-tenth of the cases, comparing the cleaned data to the re-entered file (entered by a second data entry operator) and again reconciling the cases. Reconciling entailed comparing cases that did not match to the original microfilm and choosing the correct value. At this point, the data checker assigned errors. Error reports listed the number of differences or discrepancies found in the cleaned data. Overall, the data entry error rates were 0.12% for both the 1910 data and the 1920 data, however, if one calculates the error rates including the null cells, which a data entry operator must recognize, only

15 16 Ve ly v i s, Th o m p s o n-co l ó n, Wi n s b o r o u g h 0.002% of the actual entries in the 1910 sample were erroneous and only 0.007% of the entries in the 1920 sample were erroneous. The quality of the data was improved during the process of allocating missing variables. Further logical checks were done during this process, and the data were looked at more closely at the case level to determine whether there were errors. These processes indicated that the final data available at ICPSR and IPUMS are of high-quality. Table 5a. Comparison of 1910 Census and Weighted Distributions by Age for Men Age Groups Census Population Men Percent of Population Weighted Percent of Weighted Under 5 94, % 9, % 5 to 9 76, % 7, % 10 to 14 74, % 7, % 15 to 19 53, % 5, % 20 to 24 53, % 5, % 25 to 29 45, % 4, % 30 to 34 35, % 3, % 35 to 39 32, % 3, % 40 to 44 26, % 2, % 45 to 49 18, % 1, % 50 to 54 16, % 1, % 55 to 59 9, % % 60 to 64 8, % % 65 to 69 4, % % 70 to 74 3, % % 75 to 79 1, % % 80 to 84 1, % % % % Age Unknown % % Total 557, % 55, % Notes: Census (U.S. Bureau of the Census 1913); ICPSR sample data (ICPSR04343-v1). data do not include solares.

16 Public use samples of 1910 and Table 5b. Comparison of 1910 Census and Weighted Distributions by Age for Women Age Groups Census Population Women Percent of Population Weighted Percent of Weighted Under 5 90, % 9, % 5 to 9 74, % 7, % 10 to 14 69, % 6, % 15 to 19 60, % 6, % 20 to 24 55, % 5, % 25 to 29 48, % 4, % 30 to 34 35, % 3, % 35 to 39 32, % 3, % 40 to 44 25, % 2, % 45 to 49 18, % 1, % 50 to 54 16, % 1, % 55 to 59 9, % % 60 to 64 10, % 1, % 65 to 69 5, % % 70 to 74 4, % % 75 to 79 1, % % 80 to 84 1, % % , % % Age Unknown % % Total 560, % 55, % Notes: Census (U.S. Bureau of the Census 1913); ICPSR sample data (ICPSR04343-v1). data do not include solares. Some Comments on the Taking of a U.S. Census in Puerto Rico in 1910 The U.S. Census has always been designed to collect and provide data that many kinds of contemporary users will find meaningful and useful. By 1910, experience, thought, and trial and error had produced a method of census taking that seemed to lead to that end for users in the United States. Of course these methods were culture-laden depending, as they did, on shared cultural practices and shared meanings. To make this assertion is not to say that the Census was narrow in its acceptance of the diverse practices and meaning that it encountered. It was fairly well adapted to such diverse groups as immigrant communities in New

17 18 Ve ly v i s, Th o m p s o n-co l ó n, Wi n s b o r o u g h Table 6a. Comparison of 1920 Census and Weighted Distributions by Age for Men Age Groups Census Population Men Percent of Population Weighted Percent of Weighted Under 5 100, % 9, % 5 to 9 99, % 9, % 10 to 14 85, % 8, % 15 to 19 60, % 6, % 20 to 24 62, % 6, % 25 to 29 44, % 4, % 30 to 34 37, % 3, % 35 to 39 38, % 3, % 40 to 44 31, % 3, % 45 to 49 28, % 2, % 50 to 54 21, % 2, % 55 to 59 11, % 1, % 60 to 64 12, % 1, % 65 to 69 5, % % 70 to 74 4, % % 75 to 79 1, % % 80 to 84 1, % % , % % Age Unknown % % Total 647, % 63, % Notes: Census (U.S. Bureau of the Census 1922); ICPSR sample data (ICPSR04344-v1). data do not include posteriores. York, cotton plantations in Louisiana, dairy farms in the Midwest, large ranches in Montana, Indian reservations, somewhat nomadic groups in Alaska and multiracial individuals in Hawaii. Alaska and Hawaii were the only non-contiguous territories enumerated prior to Nonetheless, assumptions about culture abounded. The U.S. Census counts the population and assays its characteristics by first figuring out where people live. People are assumed to live at a single location at a certain point in time, one that can be reliably situated on a map and placed within the politically and socially relevant communities for which the census will publish data. Further, peo-

18 Public use samples of 1910 and Table 6b. Comparison of 1920 Census and Weighted Distributions by Age for Women Age Groups Census Population Women Percent of Population Weighted Percent of Weighted Under 5 99, % 9, % 5 to 9 95, % 9, % 10 to 14 82, % 7, % 15 to 19 66, % 6, % 20 to 24 66, % 6, % 25 to 29 51, % 5, % 30 to 34 40, % 3, % 35 to 39 38, % 3, % 40 to 44 30, % 2, % 45 to 49 21, % 2, % 50 to 54 19, % 1, % 55 to 59 9, % % 60 to 64 12, % 1, % 65 to 69 5, % % 70 to 74 5, % % 75 to 79 2, % % 80 to 84 2, % % , % % Age Unknown % % Total 651, % 64, % Notes: Census (U.S. Bureau of the Census 1922); ICPSR sample data (ICPSR04344-v1). data do not include posteriores. ple are expected to live in recognizable dwelling units and be grouped therein by households made up of mostly related individuals. One person in the household in 1910, usually male, is designated head of the household and an individual s family membership is determined in relationship to one s kinship with the head. Thus, membership in a family is dependent on co-residence and, hence, on the definition and boundaries of a household. Sticking points in Puerto Rico in 1910 also included that some relationships are recognized as kin and some are not. For example, a step-child is kin but a god-child is not. When the Census Bureau recognizes that a group of people do not fit well within its expectations, definitions, and rules, it typically

19 20 Ve ly v i s, Th o m p s o n-co l ó n, Wi n s b o r o u g h Table 7. Comparison of 1910 Census and Weighted Distributions by Municipios Number of Families per Municipio Number of Persons per Municipio Municipio Census Popltn % of Popltn Wt ed % of Census Popltn % of Popltn Wt ed % of Adjuntas 3, % % 16, % 1, % Aguada 2, % % 11, % 1, % Aguadilla 4, % % 21, % 2, % Aguas Buenas 1, % % 8, % % Aibonito 2, % % 10, % % Añasco 2, % % 14, % 1, % Arecibo 8, % % 42, % 4, % Arroyo 1, % % 6, % % Barceloneta 2, % % 11, % 1, % Barranquitas 1, % % 10, % 1, % Barros 2, % % 15, % 1, % Bayamón 6, % % 29, % 2, % Cabo Rojo 3, % % 19, % 1, % Caguas 5, % % 27, % 2, % Camuy 2, % % 11, % 1, % Carolina 2, % % 15, % 1, % Cayey 3, % % 17, % 1, % Ciales 3, % % 18, % 1, % Cidra 1, % % 10, % 1, % Coamo 3, % % 17, % 1, % Comerio 1, % % 11, % 1, % Corozal 2, % % 12, % 1, % Culebra % % 1, % % Dorado % % 4, % % Fajardo 4, % % 21, % 2, % Guayama 3, % % 17, % 1, % Guayanilla 2, % % 10, % % Gurabo 1, % % 11, % 1, % Hatillo 1, % % 10, % 1, % Humacao 5, % % 26, % 2, % Isabela 3, % % 16, % 1, % Juana Díaz 5, % % 29, % 2, % Juncos 2, % % 11, % 1, % Lajas 2, % % 11, % 1, % Lares 4, % % 22, % 2, % Las Marías 2, % % 10, % % Loíza 2, % % 13, % 1, % Manatí 3, % % 17, % 1, % Maricao 1, % % 7, % % Maunabo 1, % % 7, % % Mayagüez 9, % % 42, % 4, % Moca 2, % % 13, % 1, % Morovis 2, % % 12, % 1, % Naguabo 2, % % 14, % 1, % Naranjito 1, % % 8, % % Patillas 2, % % 14, % 1, % Peñuelas 2, % % 11, % 1, % Ponce 14, % 1, % 63, % 6, % Quebradillas 1, % % 8, % % Rincón 1, % % 7, % % Río Grande 2, % % 13, % 1, % Río Piedras 3, % % 18, % 1, % Caribbean Sabana Grande Studies 2, % % Vol. 35, No. 11,523 2 (July 1.03% - December 1, ), % Salinas 2, % % 11, % % San German 4, % % 22, % 2, % San Juan 10, % 1, % 48, % 4, %

20 Public use samples of 1910 and Number of Families per Municipio Number of Persons per Municipio Municipio Census Popltn % of Popltn Wt ed % of Census Popltn % of Popltn Wt ed % of Río Adjuntas Grande 2,497 3, % 1.48% % 1.45% 13,948 16, % 1.52% 1, , % 1.52% Río Aguada Piedras 3,488 2, % 1.06% % 1.10% 18,880 11, % 1.04% 1, , % 1.06% Sabana Aguadilla Grande 2,157 4, % 1.92% % 1.92% 11,523 21, % 1.92% 1, , % 1.90% Salinas Aguas Buenas 2,375 1, % 0.67% % 0.63% 11,403 8, % 0.74% % 0.75% San Aibonito German 4,516 2, % 0.94% % 0.86% 22,143 10, % 0.97% 2, % 0.89% San Añasco Juan 10,293 2, % 1.33% 1, % 1.28% 48,716 14, % 1.29% 4, , % 1.30% San Arecibo Lorenzo 2,634 8, % 3.85% % 3.85% 14,278 42, % 3.80% 1, , % 3.71% San Arroyo Sebastian 3,638 1, % 0.72% % 0.67% 18,904 6, % 0.62% 1, % 0.60% Santa Barceloneta Isabel 1,508 2, % 1.02% % 0.95% 11,644 6, % 1.04% 1, % 0.96% Toa Barranquitas Alta 1,585 1, % 0.83% % 0.87% 10,503 9, % 0.94% 1, % 0.96% Toa Barros Baja 1,167 2, % 1.17% % 1.22% 15,028 6, % 1.34% 1, % 1.49% Trujillo Bayamón Alto 1,056 6, % 2.76% % 2.50% 29,986 6, % 2.68% 2, % 2.63% Utuado Cabo Rojo 7,631 3, % 1.69% % 1.69% 41,054 19, % 1.75% 4, , % 1.77% Vega Caguas Alta 1,547 5, % 2.40% % 2.29% 27,160 8, % 2.43% 2, % 2.39% Vega Camuy Baja 2,305 2, % 0.97% % 0.98% 12,831 11, % 1.01% 1, , % 1.03% Vieques Carolina 2,539 2, % 1.30% % 1.30% 10,425 15, % 1.37% 1, % 1.46% Yabucoa Cayey 3,456 3, % 1.62% % 17,338 17, % 1.58% 1, , % 1.48% Yauco Ciales 6,515 3, % 1.47% % 1.45% 31,504 18, % 1.65% 3, , % 1.66% Total Cidra 219,827 1, % 0.91% 22, % 0.93% 1,118,012 10, % 0.95% 111, , % 1.02% Coamo 3, % % 17, % 1, % Notes: Census (U.S. Bureau of the Census 1913); ICPSR sample data Comerio 1, % % 11, % 1, % Corozal (ICPSR04343-v1). 2, % data do not 0.97% include 12,978 solares. 1.16% Italicized 1,272.0 cases indicate where Chi-square % values are 41.0 higher 0.18% than 101,315 with 1 degree 0.12% of freedom. 0.12% 1.14% Culebra Dorado % % 4, % % Fajardo 4, % % 21, % 2, % creates Guayama a special 3,874 procedure 1.76% or rule to 1.74% deal 17,379 with them. 1.55% Census 1,731.0 day 1.56% is set Guayanilla for April 1, 2,096 a time 0.95% when a minimum 0.95% number 10,354 of 0.93% people are not 0.89% in Gurabo 1, % % 11, % 1, % their Hatillo usual location. 1,918 People 0.87% who are 0.92% away 10,630 from home 0.95% on 1,163.0 the census 1.05% day Humacao are enumerated 5,167 on 2.35% a special transit 2.44% night. 26,678 College 2.39% students 2,658.0 have 2.39% Isabela 3, % % 16, % 1, % sometimes been reallocated back to their place of usual residence. A Juana Díaz 5, % % 29, % 2, % considerable Juncos effort 2,225 is 1.01% made to search 0.95% out hobos 11,692 and 1.05% homeless 1,188.0 people 1.07% Lajas 2, % % 11, % 1, % on transit night. Over time elaborate rules for defining the extent and Lares 4, % % 22, % 2, % boundaries Las Marías of 2,132 a household 0.97% have come 0.98% into 10,046 being, 0.90% but in they 0.89% were Loíza still comparatively 2, % primitive % 13, % 1, % Manatí 3, % % 17, % 1, % Maricao The Census 1,604 Bureau 0.73% conducted % its first 7,158 enumeration 0.64% in Puerto 0.65% Rico Maunabo in Since 1,455 Puerto 0.66% Rico had 0.68% not previously 7, % contributed to 0.68% the Mayagüez 9, % % 42, % 4, % development of Census methodology there was a certain culture clash Moca 2, % % 13, % 1, % in Morovis this first encounter 2, % that is as revealing 0.90% about 12,446 the 1.11% United 1,157.0 States as 1.04% it Naguabo 2, % % 14, % 1, % is about Puerto Rico. Many of the definitions, rules, and procedures Naranjito 1, % % 8, % % imbedded Patillas in a 2,890 census 1.31% seem transparent 1.30% to 14,448 contemporaries 1.29% 1,475.0 because 1.33% they Peñuelas are so well 2,355 articulated 1.07% with the culture 1.02% of 11,991 the time 1.07% and 1,209.0 place. Only 1.09% Ponce 14, % 1, % 63, % 6, % a change makes these definitions, rules, or procedures more visible. Quebradillas 1, % % 8, % % Applying Rincón the well 1,346 established 0.61% procedure 0.60% of the 7,275 U.S. 0.65% Census to Puerto 0.67% Río Grande 2, % % 13, % 1, % Rico in 1910 made some interesting issues visible. The research team Río Piedras 3, % % 18, % 1, % was Sabana prepared Grande to 2,157 find some 0.98% of them, but 0.95% others 11,523 were 1.03% discovered 1,158.0 seren- 1.04% Salinas 2, % % 11, % % San German 4, % % 22, % 2, % San Juan 10, % 1, % 48, % 4, % San Lorenzo 2, % % 14, % 1, % San Sebastian 3, % % 18, % 1, % Santa Isabel 1, % % 6, % %

21 22 Ve ly v i s, Th o m p s o n-co l ó n, Wi n s b o r o u g h Table 8. Comparison of 1920 Census and Weighted Distributions by Municipios Number of Families per Municipio Number of Persons per Municipio Census % of wt ed % of Census % of wt ed % of Municipio Popltn Popltn Popltn Popltn Adjuntas 3, % % 17, % 1, % Aguada 2, % % 12, % 1, % Aguadilla 4, % % 24, % 2, % Aguas Buenas 1, % % 10, % 1, % Aibonito 2, % % 13, % 1, % Añasco 2, % % 13, % 1, % Arecibo 9, % % 46, % 4, % Arroyo 1, % % 7, % % Barceloneta 2, % % 13, % 1, % Barranquitas 1, % % 11, % 1, % Barros 2, % % 15, % 1, % Bayamon 6, % % 30, % 3, % Cabo Rojo 4, % % 22, % 2, % Caguas 7, % % 35, % 3, % Camuy 2, % % 14, % 1, % Carolina 2, % % 15, % 1, % Cayey 4, % % 23, % 2, % Ceiba 1, % % 5, % % Ciales 3, % % 20, % 2, % Cidra 2, % % 14, % 1, % Coamo 3, % % 17, % 1, % Comerio 2, % % 14, % 1, % Corozal 2, % % 14, % 1, % Culebra % % % % Dorado 1, % % 5, % % Fajardo 2, % % 14, % 1, % Guanica 2, % % 9, % % Guayama 4, % % 19, % 1, % Guayanilla 2, % % 12, % 1, % Guaynabo 1, % % 10, % 1, % Gurabo 2, % % 12, % 1, % Hatillo 2, % % 13, % 1, % Hormigueros % % 4, % % Humacao 3, % % 20, % 2, % Isabela 3, % % 19, % 1, % Jayuya 2, % % 12, % 1, % Juana Diaz 3, % % 18, % 1, % Juncos 2, % % 13, % 1, % Lajas 2, % % 11, % 1, % Lares 4, % % 25, % 2, % Las Marias 2, % % 10, % 1, % Las Piedras 1, % % 10, % % Loiza 2, % % 15, % 1, % Luquillo 1, % % 6, % % Manati 3, % % 20, % 1, % Maricao 1, % % 8, % % Maunabo 1, % % 7, % % Mayaguez 9, % % 41, % 4, % Moca 2, % % 15, % 1, % Morovis 2, % % 14, % 1, % Naguabo 2, % % 15, % 1, % Naranjito 1, % % 10, % 1, % Patillas 2, % % 14, % 1, % Caribbean Peñuelas Studies 2, % % Vol. 35, 13,598 No. 2 (July 1.05% - December 1, ), 1.04% 3-29 Ponce 16, % 1, % 71, % 6, % Quebradillas 1, % % 9, % % Rincon 1, % % 8, % % Rio Grande 2, % % 13, % 1, %