<<

The Pennsylvania State University

The Gradate School

College of the Liberal Arts

HOUSEHOLDS, LAND AND LABOR: POPULATION DYNAMICS IN THE

NORTHERN ISLAND, , 1851 to 2003.

A Thesis in

Anthropology and Demography

By

Corey Shepard Sparks

© 2007 Corey Shepard Sparks

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

December 2007

ii

The thesis of Corey S. Sparks was reviewed and approved ∗ by the following:

James W. Wood Professor of Anthropology Thesis Adviser Chair of Committee

Patricia L. Johnson Associate Professor of Anthropology, Demography and Women’s Studies

Jeffrey A. Kurland Associate Professor of Anthropology and Human Development

Robert Schoen Hoffman Professor of Family Sociology and Demography

Kenneth M. Weiss Evan Pugh Professor of Anthropology and Genetics

George R. Milner Professor of Anthropology Special Signatory

Nina Jablonski Professor of Anthropology Head of the Department of Anthropology

∗Signatures are on file in the Graduate School

iii

ABSTRACT

The North Orkney Population History Project was started to understand the

determinants of population change in the northern half of the Orkney

from the 18 th to the 20 th centuries. After four years of fieldwork much data on the

population and landscape of Orkney have been collected. This thesis is the first

detailed analysis of the demographic data collected by the project. The purpose of

this thesis is to analyze the determinants of family fertility and childhood mortality in

the Northern Orkney Islands from the mid 19 th to the late 20 th centuries.

I begin by providing a thorough description of the population of the Northern

Orkney Islands from 1851 to 2001. I describe changes that have occurred in fertility, mortality and population composition during this time period. To analyze fertility I estimate the effects of individual, household and family characteristics on birth interval data generated from marriage and birth records collected from the Registrar

General of Scotland’s vital registration system using Cox proportional hazards models. Results indicate that farm families tend to have longer birth intervals than nonfarm families and the presence of post-reproductive grandparents can also lengthen birth intervals. Building on the results from the fertility analysis I estimate the effects of individual and household level characteristics on infant and childhood mortality, again relying on the Cox model for all analyses. Results indicate a decreased risk of death for children living in farm families relative to nonfarm families and that higher order and twin births face a substantial increase in risk of death relative to lower order and singleton births. iv

Based on the data for Northern Orkney, I suggest that the farm household by solidifying its economic interests and managing its household production system has out competed nonfarm families and led to the mass emigration of many of these families from the over the past century.

v

TABLE OF CONTENTS

LIST OF FIGURES ...... vi LIST OF TABLES ...... vii ACKNOWLEGEMENTS ...... viii Chapter 1 Introduction ...... 1 Objectives and outline ...... 1 Chapter 2 The Demographic and Ecological Setting of the Northern Orkney Islands ...... 3 The Orkney Archipelago ...... 3 Climate (1841-1918) ...... 7 Demographic Data ...... 10 Population Size ...... 13 Age and sex structure ...... 16 Occupational Structure ...... 23 Comparative Demographic Rates ...... 26 Household Structure and Dynamics ...... 39 Discussion ...... 44 References ...... 47 Chapter 3 Analysis of birth spacing behavior in the Northern Orkney Islands: 1855-2003...... 52 Introduction ...... 52 Data and Methods ...... 54 Results ...... 58 Discussion ...... 66 Conclusion ...... 69 References ...... 69 Chapter 4 The costs of large families: An analysis of infant and childhood mortality in the Northern Orkney Islands...... 74 Introduction ...... 74 Household growth and formation dynamics ...... 75 Data and methods ...... 78 Results ...... 83 Discussion ...... 92 Conclusion ...... 95 References ...... 95 Chapter 5 General Discussion and Summary ...... 108 General Conclusions ...... 108 Future Directions ...... 112 References ...... 114 Appendix: Database and Record Linkage ...... 129

vi

LIST OF FIGURES

Figure 2.1 1 Map of the Northern Orkney Islands, Scotland, showing areas identified in the text……………………………………………………………………….……4 Figure 2.2 Yearly average rainfall and temperature recorded at , Orkney 1841-1918……………………………………………………………………………9 Figure 2.3 Average monthly rainfall and temperature recorded at Deerness, Orkney, 1841-1918……………………………………………………………………………9 Figure 2.4 Sex ratios for the Northern Isles 1851-1901. Ratios are number of females per male………………………………………………………………………………18 Figure 2.5 Age-sex pyramid for Northern Orkney, 1851…………………………....19 Figure 2.6 Age-sex pyramid for Northern Orkney, 1861…………………………....20 Figure 2.7 Age-sex pyramid for Northern Orkney, 1871. …………………………..20 Figure 2.8 Age-sex pyramid for Northern Orkney, 1881. …………………………..21 Figure 2.9 Age-sex pyramid for Northern Orkney, 1891. …………………………..21 Figure 2.10 Age-sex pyramid for Northern Orkney, 1901. ………………………....22 Figure 2.11 Percentage of the male work force in agriculture, 1851 to 1901……….26 Figure 2.12 Comparative crude birth rate (CBR) for the Northern Isles and other areas of Scotland 1851-1911………………………………………………………………29 Figure 2.13 Comparative crude death rate (CDR) for the Northern Isles and other areas of Scotland 1851-1911…………………………………………………………30 Figure 2.14 Age-specific fertility rates calculated from the vital registration data for the Northern Isles, 1861-1901………………………………………………………..32 Figure 2.15 Average age of first marriages for men and women in Northern Orkney 1855-2003……………………………………………………………………………34 Figure 3.1 Density estimates for the first birth interval for farm and nonfarm households……………………………………………………………………………60 Figure 3.2 Density estimates for various birth orders, indicating whether the previous child had died in the interval…………………………………………………………63 Figure 3.3 Mean inter-birth interval length by completed family size and parity...... 68 Figure 4.1 Consumer/Worker ratios under the basic model of household formation (modified from Chayanov 1966). ……………………………………………...……76 Figure 4.2 Farm and nonfarm child survivorship to age 1 (Model 3). Note: Survivorship is truncated at .7 for visual purposes………...……………………...…90 Figure4.3 Child survivorship to age 1 for varying birth orders (Model 3). Note: Survivorship is truncated at .7 for visual purposes………………………………..…90 Figure 4.4 Farm and nonfarm child survivorship to age 15 (Model 3)…...……...... 91 Figure 4.5 Child survivorship to age 15 for varying birth orders (Model 3)...………92

vii

LIST OF TABLES Table 2.1 Areas of Northern Orkney Islands…………………………………………4 Table 2.2 Population sizes of the Northern Orkney Island 1851-2001………………13 Table 2.3 Population densities of the Northern Orkney Islands 1851-2001. Figures represent persons/km 2…………………………………………………..………..... 14 Table 2.4 Annual Population growth rates (%) of the Northern Orkney Islands 1851- 2001. ……………………………………………………..………...... 15 Table 2.5 Total, young age and old age dependency ratios for the Northern Isles, 1851-1901.Ratios represent the number of people supported by every 100 producers……………………………………………………………………………..17 Table 2.6 Occupational structure of the male workforce (% of total workforce) of the Northern Isles, 1851-1901…………………………………………………………...24 Table 2.7 Crude annual birth and death rates for the Northern Isles 1851-1911……28 Table 2.8 Period total fertility rates (TFRs) and total marital fertility rates (TMFR) for the Northern Isles, 1861-1901……………………………………………………….33 Table 2.9 Singulate mean age at marriage, and observed mean age at marriage for males and females in Northern Orkney, 1851 – 1901…………………………….....35 Table 2.10 Estimated life expectancy values for Northern Orkney, 1861 to 1901…..36 Table 2.11 Estimated life expectancy values for Northern Orkney, 1861 to 1901 based on indirect standardization to and ’s mortality rates…………………37 Table 2.12 Estimated life expectancy values for Northern Orkney, 1861 to 1901 based on a mixture of the observed and standardized mortality rates……………………...38 Table 2.13 Total number of households in the Northern Isles 1851-1901…………..40 Table 2.14 Average number of people living in a household in the Northern Isles, 1851-1901……………………………………………………………………………41 Table 2.15 Average number of people living at a farmstead in the Northern Isles, 1851-1901……………………………………………………………………………42 Table 2.16 Average household size of farm and nonfarm households, 1851-1901.....43 Table 2.17 Average age of household head, 1851-1901……………………………..43 Table 2.18 Percentage of male household heads, 1851-1901………………………..44 Table 2.19 Percentage of extended family households, 1851-1901………………....45 Table 3.1 Descriptive statistics for first birth interval analysis……………………...57 Table 3.2 Descriptive statistics for higher order birth interval analysis……………..57 Table 3.3 Results of Cox regression analysis of first birth interval…………………59 Table 3.4 Results of Cox regression analysis of higher birth intervals…………...... 64 Table 4.1 Descriptive statistics for household and individual level variables used in regression models…………………………………………………………………….82 Table 4.2 Results of Cox regression analysis of infant mortality…………………...84 Table 4.3 Results of Cox regression analysis of children’s age at death……………88

viii

ACKNOWLEGEMENTS

I would like to thank my advisor Dr. Jim Wood for his support, encouragement, friendship and guidance during the course of my graduate studies at

Penn State. When I began my studies at Penn State, I had never heard of Orkney, little did I know that six years later I would have probably become the foremost expert on the demography of the Orkney Islands and well versed in various names for muck. I am also very grateful to the members of my committee. Jeff Kurland has provided so much support and has often been the rock on which my various research musing have crashed. Pat Johnson, while often providing valuable nutritional supplements has been a valuable colleague during the past few years in Orkney and helped me overcome my shyness towards the methods of ethnography. George

Milner has provided valuable resources and training during the early part of my career at Penn State. Bob Schoen was instrumental in my demographic education and inspired me to purse interests in formal mathematical demography. Ken Weiss has been the epitome of the kind of scientist I would like to become in my career.

I also owe a debt of gratitude to Dr. Gordon De Jong and Dr. Leif Jensen at the Penn State Population Research Center who have provided both financial and moral support over the years. Dr. Tim Murtha has also been a valuable friend and colleague while conducting my fieldwork in Orkney.

During my graduate residency at Penn State many friends and colleagues have helped me and provided many a shoulder to lean on and long hours of theoretical musings on life. I will always be grateful to Dr. Sharon DeWitte, Kirk French, Jason ix

Deleon, Craig Goralski, Abigail Bigham, Mike Aitkenhead, Thomas Nielsen and

Erick Rochette for all their friendship and occasional much needed distractions.

While doing the fieldwork which forms the substance for this dissertation, the people of have been colleagues, friends and teachers. John, Isabell, Michael and Teenie Harcus took me in and treated me as family during my first summer in

Orkney and have been valuable friends ever since. Billy and Jean Brown, while not only allowing me to basically squat in their house for hours on end collecting data, sustained me with gallons of rocket hot tea and delicious treats. Douglas Leslie has provided valuable knowledge on the inner workings of the traditional Orcadian farming system, and cleared up many daft ideas of how the world works that academics are prone to. Sam Harcus has shown me that many Orcadians are truly scientists in their own right; doctorates be buggered! The staff of the Orkney Library and Archive, Sarah Jane Grieve-Gibbon, Lucy Gibbon, and Allison Fraser, provided great assistance in acquiring much of the accompanying information on Orkney not only for this thesis but for all of the side work that has gone into it. Finally, I would like to thank the General Registrar Office of Scotland for allowing me to access the vital registration data for Orkney. The North Orkney Population History Project has provided much valuable financial support over the past four years through two

National Science Foundation grants NSF REU #0353527 and NSF BCS# 0527539 and for providing much needed student labor. Also the Penn State Population

Research Institute provided funding through a NICHD traineeship (NICHDT-

32HD007514 Penn State Population Research Institute). x

Last but not least I owe my wife, Dr. P. Johnelle Sparks, more than anyone.

She has given me such inspiration to complete this task, from discussing the inner assumptions of inclusive fitness and Chayanovian household cycles to making me write out a work schedule and stick to it. She is my best friend, my colleague and the love of my life, and I could not have accomplished this without her.

1

Chapter 1 Introduction

Often the discussion of human populations focuses solely on one of the major population process (mortality, fertility or migration) in isolation. However, if we are to understand the intricacies of the people we study, should we not at least try to be holistic in our approaches? Demography, by its nature is holistic as much of its roots are not in social, but biological sciences (Lotka 1922, 1989) and, as a field, has been influenced by theories from economics, sociology, biology and anthropology. So as demographers, why should we attempt to pigeonhole ourselves by relying solely on a single theoretical perspective or by focusing solely on a single outcome? This thesis, presented in three parts, attempts to describe and explain the demographic variation seen in a rural Scottish population from 1855 to 2003. Instead of only examining mortality or fertility, I consider both as part and parcel of the same set of dynamics: those of the farming household.

Objectives and outline

The objectives of this dissertation are to examine the population dynamics of the Northern Orkney Islands, Scotland, over the last century and a half and by relying on theories from sociology, biology, anthropology and economics, attempt to explain the demographic rates observed over this time period. The guiding principle of this thesis is that the nature of the Orkney social environment influences the inequality of vital rates in the archipelago and has lead to long term demographic decline over the last century. To examine these effects I consider the influences of individual and 2 household conditions on fertility and mortality outcomes using individual level longitudinal data derived from the Scottish vital registration system.

This thesis is organized in the following manner. Chapter 2, “The

Demographic and Ecological Setting of the Northern Orkney Islands,” presents the general demographic setting of the study area over the period from which data are available. This section begins with a description of the general environmental and climatic setting of North Orkney. Also included in this section are descriptions of the data sources and collection methods, comparable measures of fertility and mortality, and descriptions of the age, sex, and household composition of the Northern Isles.

Building on the general data presented in Chapter 2, Chapter 3, “Analysis of birth spacing behavior in the Northern Orkney Islands: 1855-2003,” presents a hazards analysis of inter-birth intervals for the Northern Isles over the period when vital registration data are available. The analysis focuses on two main factors, farm household fertility and the impact of post-reproductive grandparents on the timing of fertility.

Chapter 4, “The costs of large families: An analysis of infant and childhood mortality in the Northern Orkney Islands,” is a hazards analysis of the influences of family size and household composition on infant and childhood mortality, focusing on the impact of the household consumer/producer ratio and the effect of being a high order or twin birth on two measures of child mortality.

Chapter 5 summarizes the conclusions from Chapters 2 through 4 and present some general conclusions on the demography of the Northern Isles based on the above analysis. Future directions for research on these data are also discussed. 3

Chapter 2 The Demographic and Ecological Setting of the Northern Orkney Islands

The purpose of this chapter is to describe the general ecological and demographic characteristics of the northern Orkney Islands during the mid nineteenth to early twentieth centuries. This will include a brief description of the Orkney archipelago, a brief history of the area focusing on local economic patterns and settlement ecology, a general description of the total population size, the age and sex structure, the occupational structure, comparative measures of fertility and mortality and a description of the household variation for each of the islands in the study area.

The Orkney Archipelago

The northern islands of Orkney are located at roughly 59˚N 3˚W (the meeting ground of the and the North Atlantic) and are composed of six major islands and approximately twelve smaller islands often referred to as holms or calves.

The total surface area of the northern islands is 177.92 km 2 with the six main islands representing 97% of the area. The six islands included in the study area are ,

North Ronaldsay, , Pharay, Sanday and Westray. The respective areas and shoreline lengths of each are given in Table 2.1. Figure 2.l is a map of the study area.

4

Table 2.1 Areas of Northern Orkney Islands.* Island Area in km 2 Length of shoreline in km Eday 29.7 46.7 9.6 27.5 Papa Westray 9.9 23.6 Pharay 1.7 8.4 Sanday 66.6 127.8 Westray 53.8 80.1 *Data are derived from a Geographic Information System of the Northern Isles created by the author, and derived from OSGB 1:25000 scale Explorer base maps.

Figure 2.1 Map of the Northern Orkney Islands, Scotland, showing areas identified in the text.

Settlement within the islands is extremely discontinuous, with individual settlements called farmsteads scattered widely across the landscape and only weakly 5 clustered into higher-order settlements – traditionally called townships but purely rural in character. Higher order settlement is extremely restricted with the only real

“villages” on the northern islands being on Westray, and to a lesser degree

Lady and Kettletoft Villages on Sanday. Although there has been much rebuilding of farm structures over the years, the great majority of modern farms are located on named farm sites that predate the eighteenth century, often by several hundred years

(Marwick 1952; Scott et al. 2003). The majority of pre-twentieth century farmsteads have (O.N.) names that describe their setting on the landscape (e.g.

Hammar = rock outcrop O.N.) or English names that describe, almost sarcastically, their often unfavorable settings (e.g. Mount Pleasant and Purgatory). Thus, while the human population of the islands may have been mobile, farmsteads appear to be remarkably stable.

The traditional system of production in northern Orkney was geared primarily

to meet household subsistence needs and secondarily to pay rents, mostly in kind. It

was based on a delicate balance of arable grain production (black oats and bere, a

primitive form of barley) and livestock raising (cattle, sheep, pigs and chickens) – a

balance that was critical to long-term demographic homeostasis since grazing both

competed with arable for land itself and provided essential inputs into arable in the

form of animal manures (Dodgshon 1994). Archaeological and historical evidence

suggests that this traditional system had not changed in any fundamental way since

the early medieval period (Berry 1985; Davidson and Simpson 1994; Fenton 1978;

Flinn 1977; Pearson and Collier 1998; Thomson 2001; Whyte 1979). Beginning in

the late eighteenth century, there occurred several major economic changes to the 6 traditional agrarian system that are known to have had dramatic effects on the northern Orkney isles, even though the demographic details are not yet clear. The first of these periods of change was the 50-year kelp boom beginning c. 1780, during which several species of seaweed were collected and burned to make kelp on a commercial scale to provide alkali for the British glass-, soap-, and dye-making industries. 1 Within a few decades, The Northern Isles and other parts of Orkney underwent a near doubling of population in response to the increased demand for labor in kelp making (Thomson 1983, 2001), a remarkable expansion for what had been up to that time a near-subsistence-level agrarian system. Although part of this population increase was attributable to net in-migration, there is circumstantial evidence that it may also have involved changes in fertility and mortality (Bowers

1983; Brennan 1979; Brennan 1983b; Gibson and Smout 1995). This situation presents an excellent opportunity for studying the effects of a large but short-lived

(and datable) change in the demand for labor on vital rates, including migration rates.

The kelp boom persisted until 1830, when the British government repealed restrictive tariffs on the importation of high quality alkali from other countries (e.g. Spanish barilla). As a result, the price of Orkney kelp collapsed, and the islands entered a period of agricultural stagnation largely reflecting the disincentives to agricultural innovation that had been associated with the kelp boom itself (Thomson 1983, 2001).

Little is known about the effects of this sudden downturn on the population of

Orkney; it is clear, however, that the collapse of the kelp market was followed by some 20-30 years of low returns on labor caused primarily by poor land management and a dearth of marketable exports (Thomson 2001). The modern agricultural

1 In British usage, kelp is the marketable material produced by burning seaweed, not the seaweed itself. 7 improvement movement finally reached Westray in the mid-nineteenth century, ushering in a period of farm reorganization, enclosure of common pasture, introduction of liming, construction of new field drainage systems, and general intensification of agricultural production (Schrank 1995). This period was again accompanied by population growth, although nowhere near as rapid as during the early stages of the kelp boom. The period of agricultural expansion continued until about 1880, when the prices of Orkney cattle and grain exports experienced another sudden downturn owing to competition from overseas, mainly North American and

Australia (Thomson 2001). Because of the ensuing agricultural depression, Orkney entered a prolonged period of population decline caused by increased out-migration to Scotland, , the United States, Australia, and New Zealand. The increase in migration associated with this same agricultural depression in other parts of the has been the subject of recent research (Pooley and Turnbull

1998), but detailed information on the north of Scotland in general, and Orkney in particular, is lacking.

Climate (1841-1918)

The climate of the northern isles is relatively stable if not (based on personal experience) always pleasant. The islands tend to be quite temperate owing to the influence of the Mid-Atlantic Drift. Detailed climatic data were collected daily and exist in logbooks housed at the Orkney Library and Archive in . These data are based on climatic observations from the 19 th and early 20 th centuries for the parish

of Deerness in eastern Mainland, Orkney, and are summarized below. Figure 2.2 8 gives the total observed rainfall in inches and the average yearly temperature recorded at Deerness between 1841 and 1918. Five year moving averages of the respective series are provided to aid in visualization. With respect to temperature, there is remarkably little variation, with the series mean being 45.7ºF, the lowest yearly average temperature being 43.8ºF in 1860 and the highest average temperature being

48.3ºF in 1852. Daily temperatures from 1890-1910 were examined and the lowest recorded temperature was 20ºF and the maximum daily temperature was 70ºF. For comparison, the average temperature at Greenwich, England, from 1841 to 1960 was

49.8ºF. The climate of Orkney is quite wet, again reflecting the influence of the Mid-

Atlantic Drift and its position in the North Atlantic. The average yearly rainfall between 1841 and 1918 was 40.4 inches; for comparison Greenwich, England, received an average of 24.1 inches of rainfall between 1941 and 1960 (WorldClimate,

2007). Monthly temperatures over this period display more variability than the yearly data. Figure 2.3 gives the average monthly temperature and rainfall over the same period; note confidence bars represent the variance in monthly temperature or rainfall over the period. The average peak temperature occurs in July and is 51.7ºF. The coldest month was February, with an average of 38.6ºF. The wettest month was

December with an average of 4.3 inches of rainfall. The driest month was May with and average of 1.8 inches of rainfall. One other pattern worth noting is the high variability in autumn and winter rainfall, with October having a monthly variance of

2.2 inches over the period.

9

Figure 2.2 Yearly average rainfall and temperature recorded at Deerness, Orkney 1841-1918.

50 49

45 48

40 47 35

46 30

25 45

20

Total Rainfall (in) 44 Average Tempreture(F) 15 Yearly Rainfall 43 10 AvgTemp

5 Year Moving Average (Temperature) 42 5 5 Year Moving Average (Rainfall) 0 41

1 6 87 88 1841 1844 1847 1850 1853 1856 1859 1862 1865 1868 1 1874 1877 1880 1883 1 1889 1892 1895 1898 1901 1904 1907 1910 1913 1916 Year

Source: Orkney Library and Archive. Meteorological observations at Deerness. Figure 2.3 Average monthly rainfall and temperature recorded at Deerness, Orkney, 1841-1918.

70 7

60 6

50 5

40 4 Temp (F) Temp

30 3 Rainfall(in)

20 2

10 1 Avg Temp Avg Rainfall 0 0 Jan Feb Mar Apr May Jun Jul Aug Sept Oct Nov Dec Month

Source: Orkney Library and Archive. Meteorological observations at Deerness. 10

Demographic Data

Two primary data sources are used throughout this dissertation: vital registration records and census data. The vital registration data are administered by the General Registrar of Scotland (GROS). Beginning in summer 2003, data were collected for the islands of Eday, Pharay, Papa Westray, North Ronaldsay, Sanday and Westray. Three primary record types form the basis of the data for this dissertation. The first are individual birth records. These contain information on the name of the child, day, month and year of birth, house and island of birth, names of the child’s parents, occupation of the child’s father, marital statuses and places and dates of marriage for the child’s parents. The second record type is marriage records.

These contain names of the bride and groom, date and place of marriage, occupational status of the bride and groom, names, occupations, marital statuses and information on whether the parents of the bride and groom were alive at the time of marriage. In addition, they contain indicators of blood relationship between bride and groom. The third record type is death records; these contain the name of the deceased, date and time of death, age at death, marital status, name of the spouse of the deceased, names and occupations of the deceased’s parents, cause of death (if the death was attended by a physician) and length of sickness. These data are used in chapters 3 and 4 for detailed individual level analysis.

The census data described in this chapter come primarily from the decennial census returns for Orkney from 1851 to 1901. These data were transcribed from microfilm copies of the original census returns by volunteers at the Orkney Family

History Society in Kirkwall between 1995 and 2001 and by the author from returns 11 from the General Registrar of Scotland’s annual reports on population. Beginning in

2003, the author began entering these transcriptions into a standardized form for analysis. The data entry was completed by the author and student research assistants employed by the North Orkney Population History Project in the summer of 2005.

The first census for Orkney was done in 1831, but only the parish of

Stromness was recorded. In 1841 a census of all parishes in Orkney was completed,

but many problems exist with this information and it is not included in this

dissertation. Primary concerns with the 1841 data are the process of age heaping

where anyone over the age of 15 was rounded down to the nearest 5 or 10 year

interval (Flinn 1977). This presents a major problem for any attempted record linking

procedure relying on year of birth as a key identifying field, especially if a particular

surname is common. Secondly, the only information on place of birth was the

question “Was the individual born in Orkney?” Unlike later censuses, which asked

for specific birthplaces, this does not allow for individuals from other parishes to be

followed between census years. Thirdly, the census failed to ask for information on

relationship of individuals to household head, which makes statements about

household and family structure uncertain at best. It was judged by the author that,

because of these inherent problems in the 1841 census it would not be used at the

present time. This chapter will rely on the more complete decennial census returns

from 1851 – 1901. Individual-level census data are unavailable for the years 1911 to

2001 because of the 100 year rule, which forbids dissemination of individual

information that could be used for identification purposes. With a few exceptions,

these census years relied on the same census questionnaire, making for easier 12 comparability between years. The exceptions mentioned above relate to the recording of rented or owned land and the household labor pool. From 1851 to 1881 the census form asked all household heads if they owned or rented land and information on the servants employed by the household (specifically information on number, sex, and job duties of servants). Unfortunately, these questions were removed in the 1891 and

1901 census forms for uncertain reasons. Inconsistencies also exist in census enumeration districts within parishes. The delineations of the individual islands used in the 1851 census are completely different from their 1861-1901 counterparts making temporal comparisons of individual districts impossible.

Despite the problems listed above, the census records represent a valuable set of information on the microdemographic structure of the northern isles during this period. The census returns provide information at the individual level, with identifiers allowing for construction of individual households. For each individual information on place of birth, place of residence at the time of the census, occupation, relationship to household head (including kin relations), and from 1851-1881 the amount of land held by individual households. Presumably, the census records everyone living on the island at the time of census, generally around the first week of

April, and if someone was absent from the household the enumerator mentioned the fact on the first page of the return. The following sections describe general demographic characteristics of each of the northern islands between 1851 and 1901

13

Population Size

Over the last century and a half, the Orkney archipelago has undergone both population growth and decline. The period of population growth generally peaked in the mid to late 19 th century, with sustained decline following. Table 2.2 gives the total population sizes for the northern isles between 1851 and 2001.

Table 2.2 Population sizes of the Northern Orkney Island 1851-2001. Year Eday North Papa Pharay Sanday Westray Total Ronaldsay Westray

1851 944 481 337 67 1892 1791 5512 1861 947** 526 371 69 2004 2088 6005 1871 897 532 392** 82 2006 2153** 6062** 1881 822 539 370 83** 2053 2090 5957 1891 730 547** 345 72 2075** 2190 5959 1901 647 501 337 58 1929 2044 5516 1911 592 442 295 47 1727 1817 4920 1921 508 436 258 51 1529 1659 4441 1931 474 349 247 60 1403 1507 4040 1941* 430 298 237 60 1160 1249 3434 1951 308 224 184 0 866 1080 2662 1961 198 161 139 0 670 872 2040 1971 179 134 106 0 592 735 1746 1981 147 109 92 0 525 701 1574 1991 166 92 85 0 533 704 1580 2001 121 70 65 0 478 563 1297 *No census, population i nterpolated using midpoint of 1931 and 1951 population counts. **Maximum recorded population. Data from Barclay (1965).

The peak population for the northern isles occurred in 1871, with the various islands peaking between 1861 and 1891. The first island to peak was Eday in 1861 with 947 people, and the latest to peak were North Ronaldsay and Sanday in 1891 with 547 and

2075 people respectively. While the level of population growth and decline is obvious (see Table 2.4 for population growth rates), the one feature that deserves 14 further note is the abandonment of the island of Pharay. From the numerical evidence this occurred sometime between the 1931 and 1951 census (we would not know for sure because there was not 1941 census of ), but based on ethnographic interviews the island was abandoned in 1947 (unpublished data).

It is obvious that there are marked differences in population size between the islands, with Sanday and Westray being the largest and Pharay being the smallest on average. To standardize this figure for more accurate comparison, the population densities for each island and time period are presented in Table 2.3.

Table 2.3 Population densities of the Northern Orkney Islands 1851-2001. Figures represent persons/km 2. Year Eday North Papa Pharay Sanday Westray Total Ronaldsay Westray

1851 31.8 49.9 33.9 39.7 28.4 33.3 32.2 1861 31.9 54.5 37.3 40.9 30.1 38.8 35.0 1871 30.2 55.1 39.4 48.6 30.1 40.0 35.4 1881 27.7 55.9 37.2 49.2 30.8 38.8 34.8 1891 24.6 56.7 34.7 42.7 31.2 40.7 34.8 1901 21.8 51.9 33.9 34.4 29.0 38.0 32.2 1911 19.9 45.8 29.7 27.9 25.9 33.7 28.7 1921 17.1 45.2 25.9 30.2 23.0 30.8 25.9 1931 16.0 36.2 24.8 35.6 21.1 28.0 23.6 1941 14.5 30.9 23.8 35.6 17.4 23.2 20.0 1951 10.4 23.2 18.5 0.0 13.0 20.1 15.5 1961 6.7 16.7 14.0 0.0 10.1 16.2 11.9 1971 6.0 13.9 10.7 0.0 8.9 13.6 10.2 1981 5.0 11.3 9.2 0.0 7.9 13.0 9.2 1991 5.6 9.5 8.5 0.0 8.0 13.1 9.2 2001 4.1 7.3 6.5 0.0 7.2 10.5 7.6

We see the most densely occupied island was North Ronaldsay with a density of nearly 57 people/km 2 in 1891. North Ronaldsay remained the most densely occupied

island until 1971 when Westray became most densely occupied. Eday was 15 consistently the island with the lowest density of population, owing most likely to its highly variable landscape and low proportion of land of agricultural value.

The rates of population growth are given in table 2.4. These rates were calculated as ln(N t+10 /N t)/10 (Preston et al. 2000), and expressed as the percentage

annual growth rate since the previous census, where t is the census year.

Table 2.4 Annual Population growth rates (%) of the Northern Orkney Islands 1851- 2001. Eday North Papa Pharay Sanday Westray Total Ronaldsay Westray

1851 ------1861 0.03 0.89 0.96 0.29 0.58 1.53 0.86 1871 -0.54 0.11 0.55 1.73 0.01 0.31 0.09 1881 -0.87 0.13 -0.58 0.12 0.23 -0.30 -0.17 1891 -1.19 0.15 -0.70 -1.42 0.11 0.47 0.00 1901 -1.21 -0.88 -0.23 -2.16 -0.73 -0.69 -0.77 1911 -0.89 -1.25 -1.33 -2.10 -1.11 -1.18 -1.14 1921 -1.53 -0.14 -1.34 0.82 -1.22 -0.91 -1.02 1931 -0.69 -2.23 -0.44 1.63 -0.86 -0.96 -0.95 1941 -0.97 -1.58 -0.41 - -1.90 -1.88 -1.63 1951 -3.34 -2.85 -2.53 - -2.92 -1.45 -2.55 1961 -4.42 -3.30 -2.80 - -2.57 -2.14 -2.66 1971 -1.01 -1.84 -2.71 - -1.24 -1.71 -1.56 1981 -1.97 -2.06 -1.42 - -1.20 -0.47 -1.04 1991 1.22 -1.70 -0.79 - 0.15 0.04 0.04 2001 -3.16 -2.73 -2.68 - -1.09 -2.23 -1.97

We see a general trend of slight annual growth for most of the islands during the mid to late 19 th century, but beginning in 1901 we see consistent levels of population decline that continue in most islands until 1991 when we see slight positive population growth for Eday, Sanday and Westray. This appears to be a short term growth, because negative growth rates return again in 2001. By examining the general trajectory of Northern Isle’s population growth, the trend is clear: continued 16 depopulation and likely abandonment of smaller communities with larger islands struggling to remain sustainable.

Age and sex structure

One of the most important attributes of a population is its age and sex distribution. This is of importance because age specific rates of fertility and mortality depend on the age-specific population at risk of experiencing an event. The age distribution is also necessary for the fitting of model schedules of fertility and mortality and for determining total stability and stationarity of the population under consideration. These notions of stability and stationarity have important implications because the age distribution of a population can reflect past demographic disturbances such as mortality crises, age-specific migratory events or sudden changes in demographic rates.

Additionally, the general productivity of population depends heavily on the consumer/producer balance, typically referred to as the population dependency ratio.

The dependency ratio is typically calculated as the number of people under age 15 and the number over age 65 (presumed to be non-producers) divided by the number of people aged 15 to 65 (producers) and is interpretable as the fraction of the total population supported by the working-age segment of the population. While these definitions of producer and consumer are not without problems, the dependency ratio is a standard and hence comparable measure of population productivity when better measures (e.g. per capita income) are absent. Subsequent to the total dependency ratio, other commonly used quantities are the young and old age dependency ratios. 17

These are calculated as the fraction of the population under age 15 (young age) or over 65 (old age) relative to the population aged 15 to 65, respectively. Table 2.5 gives the dependency ratios for each of the northern isles between 1851 and 1901. It also lists the young and the old age dependency ratios.

Table 2.5 Total, young age and old age dependency ratios (DR) for the Northern Isles, 1851-1901. Ratios represent the number of people supported by every 100 producers. Total DR 1851 1861 1871 1881 1891 1901 Eday 73.9 82.3 87.4 78.9 73.2 71.8 North Ronaldsay 74.8 68.0 87.2 84.2 66.8 59.8 Papa Westray 81.9 109.6 88.8 73.4 66.9 56.9 Pharay 72.5 86.4 102.4 100.0 114.8 51.6 Sanday 69.5 67.9 79.2 73.9 67.4 61.8 Westray 75.5 76.5 81.6 81.4 79.8 79.7

Young Age DR Eday 59.7 66.1 70.6 60.0 55.2 50.4 North Ronaldsay 62.5 55.4 68.4 69.0 54.2 43.1 Papa Westray 66.2 92.0 70.4 59.8 52.3 39.4 Pharay 65.0 81.8 90.2 77.8 92.6 48.4 Sanday 56.9 58.3 65.8 60.8 50.7 42.8 Westray 62.4 62.5 65.2 64.6 60.3 57.6

Old Age DR Eday 14.2 16.3 16.8 18.9 18.0 21.3 North Ronaldsay 12.3 12.7 18.8 15.2 12.6 16.7 Papa Westray 15.7 17.6 18.4 13.6 14.5 17.6 Pharay 7.5 4.5 12.2 22.2 22.2 3.2 Sanday 12.6 9.6 13.5 13.2 16.8 19.0 Westray 13.1 13.9 16.4 16.8 19.5 22.2

In looking at general trends in the dependency ration, the average total ratio displays little total change across the sixty year period despite some variation in some of the smaller islands. Notably Papa Westray and Pharay tend to display a high degree of census to census variability, most likely attributable to their small total population 18 sizes. The young age dependency ratio on average is declining across islands. This in combination with the old age dependency ratio, which is on average increasing, suggests a general trend of declining fertility and population aging across the islands over this period. For the sake of comparison, the average total dependency ratio in

1851 was 74.7; in 2006 countries having similar ratios are Botswana and Haiti. The maximum average dependency ratio of 87.8 occurred in 1871; in 2006 countries that had this ratio were Namibia, Togo and Guinea. The lowest average dependency ratio of 63.6 was in 1901; in 2006 countries around this value were Uzbekistan and Egypt.

The age-sex structures of the isles show some remarkable similarities across islands and years, with a notable deficit of working-aged males Figure 2.4 shows the average sex ratios for each age class and census year for the Northern Isles.

Figure 2.4 Sex ratios for the Northern Isles 1851-1901. Ratios are number of females per male.

1.6

1.4

1.2

1 1851 1861 1871 0.8 1881

Sex Ratio Sex 1891 1901 0.6

0.4

0.2

0 0_9 10_19 20_29 30_39 40_49 50_59 60_69 70_79 80+ Age

19

To further illustrate the age-sex composition of the islands over the period, I present the age-sex pyramids in Figures 2.5 to 2.10 for the Northern Isles for each census year. In 1851 the population looks relatively stable, and the general shape is characteristic of high fertility and mortality. There does appear to be a slight deficit of males relative to females in the 20 to 40 year old range.

Figure 2.5 Age-sex pyramid for Northern Orkney, 1851

100

90

80

70

60

Males 50 Age Females

40

30

20

10

0

-500 -400 -300 -200 -100 0 100 200 300 400 500 Count

20

Figure 2.6 Age-sex pyramid for Northern Orkney, 1861

100

90

80

70

60

Males 50 Age Females

40

30

20

10

0

-500 -400 -300 -200 -100 0 100 200 300 400 500 Count

Figure 2.7 Age-sex pyramid for Northern Orkney, 1871

90

80

70

60

50 Males

Age Females 40

30

20

10

0

-500 -400 -300 -200 -100 0 100 200 300 400 500 Count

21

Figure 2.8 Age-sex pyramid for Northern Orkney, 1881

90

80

70

60

50 Males

Age Females 40

30

20

10

0

-500 -400 -300 -200 -100 0 100 200 300 400 500 Count

Figure 2.9 Age-sex pyramid for Northern Orkney, 1891

90

80

70

60

50 Males

Age Females 40

30

20

10

0

-500 -400 -300 -200 -100 0 100 200 300 400 500 Count

22

Figure 2.10 Age-sex pyramid for Northern Orkney, 1901

90

80

70

60

50 Males

Age Females 40

30

20

10

0

-500 -400 -300 -200 -100 0 100 200 300 400 500 Count

Beginning in 1861, this gap becomes more pronounced generating an even larger gap between the numbers of males and females. This gap widens in 1871, and by 1881 we see not only a deficit of males, but a reduction in the number of females as well.

By the 1891 and 1901 censuses, it is apparent that the population’s fertility has declined, as evidenced by the narrow base of the pyramid and the population has also dramatically shrunk in total size. What appears to be an initial sex-based out migration during the 1860’s and 70’s eventually turned into both a male and female out migration in the late 1800’s. This massive loss in the number of marital and reproductive age people most likely led to a reduction in total fertility and in marriage. I will return to these issues later in this chapter.

23

Occupational Structure

In this section I describe the occupational structure of the population from 1851 to

1901 as recorded in the census enumerations. Since the census returns list 2,452 unique occupations for the area I employ the coding system of the Integrated Public

Use Microdata Series (IPUMS) to combine occupations into a total of 11 categories

(Ruggles et al. 2004). These categories are: professional/technical, farmers, managers/proprietors, clerical/kindred, sales workers, craftsmen, operatives, private household service workers, non household service workers, farm laborers, laborers, non-occupational responses, and no occupation/missing. These categories are then lumped into four higher levels of classification for ease of comparison and because many islands have extremely low or zero frequencies of the categories (e.g. .12% of the population of Sanday in 1861 were kindred workers). These higher level categories are Agricultural, nonagricultural laborers, skilled laborers, and professionals. I exclude non-occupational responses and no occupation responses from further calculations. Table 2.6 lists the proportions of the male work force in each of these categories for the Northern Isles between 1851 and 1901.

Unfortunately, unless a female was a house servant or the head of a household she was typically not given an occupation by the census enumerator, despite the integral part women played in the island economy ( 1974; Wenham 2001).

24

Table 2.6 Occupational structure of the male workforce (% of total workforce) of the Northern Isles, 1851-1901. 1851 1861 1871 1881 1891 1901 Eday Agriculture 40.2 39.2 39.6 40.5 37.4 51.6 NonAgLabor 29.8 32.4 35.4 34.9 40.5 27.9 Skilled Labor 28.7 26.5 23.1 23.0 19.1 14.6 Professional 1.3 1.9 1.9 1.6 3.1 5.9

North Ronaldsay Agriculture 77.1 61.9 60.0 60.4 49.1 44.2 NonAgLabor 11.1 19.7 26.5 21.3 31.2 41.6 Skilled Labor 10.4 16.3 11.8 15.5 16.2 11.2 Professional 1.4 2.0 1.8 2.9 3.5 3.0

Papa Westray Agriculture 67.4 50.0 46.6 58.4 47.5 64.8 NonAgLabor 19.4 29.0 39.7 21.8 41.6 23.9 Skilled Labor 10.1 18.5 11.5 15.8 7.9 8.0 Professional 3.1 2.4 2.3 4.0 3.0 3.4

Pharay Agriculture 44.4 37.5 36.4 50.0 76.9 72.7 NonAgLabor 44.4 46.9 39.4 27.8 15.4 18.2 Skilled Labor 7.4 12.5 24.2 16.7 0.0 0.0 Professional 3.7 3.1 0.0 5.6 7.7 9.1

Sanday Agriculture 58.8 61.4 67.0 62.5 57.4 55.0 NonAgLabor 22.6 19.4 19.4 19.6 22.5 26.1 Skilled Labor 16.6 16.7 12.2 15.3 17.6 15.6 Professional 2.0 2.4 1.4 2.6 2.6 3.3

Westray Agriculture 51.8 47.1 55.8 51.6 44.6 56.3 NonAgLabor 29.1 26.1 22.8 29.1 36.1 26.8 Skilled Labor 16.7 24.2 18.6 16.8 16.6 14.0 Professional 2.4 2.7 2.8 2.5 2.6 2.9

As revealed in the above table, and unsurprisingly for anyone who has ever visited the Orkney archipelago, the occupational structure of the islands is dominated by 25 agriculture. Typically only 1-3% of the population was Professionals, and these were generally teachers and schoolmasters. Similarly a low but variable proportion of the population was skilled laborers; these would primarily be smiths, millers and weavers 2. Generally the second most common occupational category was non agricultural laborers, primarily represented by fishermen and quarrymen. This is not surprising, as Orcadians have often been referred to as farmers with boats during good agricultural times and as fishermen with plows during times when became the dominant industry. The Northern Isles have long been an agricultural community, and this easily recognizable from the proportion of the population in agriculture. This category represents both farmers and agricultural laborers, and as shown in Figure 2.11 decreases in frequency for North Ronaldsay and Sanday between 1851 and 1901. This category increases or does not substantially change for the other islands, save Pharay whose workforce becomes dominated (72-76%) by agriculture between 1891 and 1901.

Lastly, with respect to the occupational structure of the islands, it should be said that trying to apply a rigid scheme of occupational classification to Orcadians is not a simple task. As mentioned above the notion that Orcadians are often said to be farmers with boats implies that it is the rule rather than the exception that an individual or a household participates in many work activities throughout the year.

For example someone might participate in plowing a field in April, join a group of neighbors to cut grass in July, and be on a fishing boat in the North Sea in October. It

2 Although much of the population would have contributed to the agricultural labor pool at various times during the year, it is assumed that individuals listed as farmers in the census would be more likely to base the majority of their household income on farming alone, versus other people who draw an income from smithing or weaving. 26 is a large, and a very weak assumption that if we were to ask someone from Orkney what their profession was that we would get in response a single discrete category.

Instead we would most likely be told that the person did whatever it took to survive as an answer - a reply as true today as it was in the 19 th century.

Figure 2.11 Percentage of the male work force in agriculture, 1851 to 1901.

100.0

90.0

80.0

70.0

60.0 Eday North Ronaldsay Papa Westray 50.0 Pharay Sanday

% in Agriculture in% 40.0 Westray

30.0

20.0

10.0

0.0 1851 1861 1871 1881 1891 1901 Year

Comparative Demographic Rates

The basics of demographic comparison lie in the differences in population vital rates. Although much has been written with regard to general trends in demographic rates for (Bourgeois-pichat 1981; Coale and Watkins 1986;

Guinnane et al. 1994; Gunnlaugsson 1988; Knodel 1977; Knodel 1988; Laslett and

Wall 1972), Great Britain (Anderson 1998; Gage 1993; McKeown et al. 1975; 27

Russell 1948; Smith 1981; Woods et al. 1988, 1989; Wrigley and Schofield 1983,

1989), and to a lesser extent Scotland (Anderson and Morse 1993a, b; Flinn 1977), our knowledge of the specific demographic experience of the Northern Isles is limited

(Barclay 1965). Previous work on microdemographic variation in the Northern Isles is extremely limited (Bowers 1983; Brennan 1979; 1983a, b; Brennan et al. 1982;

Harrison 1976; Macbeth and Boyce 1987; Relethford and Brennan 1982; Roberts

1983; Roberts and Roberts 1983) and has dealt mainly with the population genetic aspects of surnames and unique genetic markers. So we know little of the actual processes of birth, death and marriage and needless to say household composition and social networks. Macrodemographic work has traditionally lumped Orkney together with , and as the Northern area of Scotland, choosing to ignore any small scale variation that is of utmost concern to anthropological demographers.

This section focuses on the comparative macrodemographic conditions of the

Northern Isles during the late 19 th and early 20 th century. Specifically I will present crude rates of birth and death for each of the isles and for comparative areas of

Scotland over the period. To accurately discuss the macrodemographic circumstances of these areas, we would ideally have information on immigration and emigration, but unfortunately these data do not exist for this time period.

Table 2.7 presents the crude birth and death rates for the northern isles between 1851 and 1911. Because for some years either no vital event occurred, or the events were recorded on the neighboring island for Pharay and Papa Westray, they are combined with the larger island closest to them, namely Eday and Westray. 28

Table 2.7 Crude annual birth and death rates for the Northern Isles 1851-1911. Eday North Sanday Westray and Ronaldsay and Papa Pharay Westray Year CBR CDR CBR CDR CBR CDR CBR CDR 1851 43.1 16.0 28.4 17.0 24.6 13.5 28.8 11.6 1861 33.7 20.4 13.2 7.5 28.3 12.6 28.7 12.2 1871 21.0 8.8 37.1 7.4 30.2 11.2 26.7 17.9 1881 21.2 10.0 34.7 16.5 24.5 11.0 25.9 13.0 1891 24.1 12.8 22.0 14.0 21.3 15.0 29.4 9.8 1901 20.2 23.3 22.6 20.4 19.7 12.2 20.4 19.1 1911 19.7 12.5 20.6 20.6 20.9 14.4 23.4 16.6 *Rates are per 1000 people.

While there is a high degree of variation between the islands with respect to fertility and mortality, two trends emerge from the data. There is a slight decreasing trend in the crude birth rate (CBR) from an average value of 32 in 1851 to 20 in 1911, and the crude death rate remains remarkably constant over time from an average value of 15.5 in 1851 to 15.8 in 1911. But from a comparative perspective, how do these rates compare with other areas of Scotland at these time periods? Figures 2.12 and 2.13 present the crude birth and death rates for the northern isles and other areas of

Scotland computed from the annual reports of the General Registers Office of

Scotland (RGS) from 1851 to 1911. With respect to fertility, the northern isles show the second or third lowest level of fertility of the 10 regions over the period, tending to resemble other northern and northwestern areas of Scotland, with Orkney as a whole showing the lowest level of fertility of any area. The picture of fertility presented here seems to correspond to what we would expect from the findings from other areas of Europe, showing the general decline in fertility over the latter 29 nineteenth century, a fact that met with confusion from some of the only comparative demographic analysis of Scotland during this period (Anderson and Morse 1993a, b).

Figure 2.12 Comparative crude birth rate (CBR) for the Northern Isles and other areas of Scotland 1851-1911.

4.50

4.00

3.50

NorthernIslesAvg 3.00 Orkneytotal EastMidland 2.50 NorthEastern Northern

CBR NorthWestern 2.00 SouthEastern Southern 1.50 SouthWestern WestMidland

1.00

0.50

0.00 1851 1861 1871 1881 1891 1901 1911 Year

30

Figure 2.13 Comparative crude death rate (CDR) for the Northern Isles and other areas of Scotland 1851-1911.

4.50

4.00

3.50

NorthernIslesAvg 3.00 Orkneytotal EastMidland 2.50 NorthEastern Northern

CDR NorthWestern 2.00 SouthEastern Southern 1.50 SouthWestern WestMidland

1.00

0.50

0.00 1851 1861 1871 1881 1891 1901 1911 Year

With respect to mortality, the northern isles show the lowest level of mortality of all the comparative groups until 1901 and 1911 when they increase to the highest level of mortality. From 1851 to 1891, the northern isles show a remarkably low level of general mortality, suggesting the population was not overly exposed to adverse conditions like their counterparts in the urban areas of the southeast ( and the Lothians) and the southwest (Glasgow and Strathclyde), indicating the possible benefit of living in a remote rural area. We have to wonder, though, why do we see the increase in CDR in 1901 and 1911 in the northern isles? The primary constituents of this are the high CDR’s in Westray and North Ronaldsay in 1901 and 1911, but upon examination of the civil registration of deaths for these islands no immediate 31 cause presents itself (e.g. a large shipwreck or outbreak of small pox), and perhaps this is just a statistical artifact of such small populations, or based on the data presented in Figures 2.5 through 10 this slight increase could be the result of the dramatically changing age structure in the islands (see above). These fluctuating age structures could artificially decrease the crude death rates, making them seem lower than they actually are. Only by age-standardizing the death rates can we really gain an appreciation for the mortality experience during this time period. Unfortunately, the vital registration data seem very incomplete at this time (see below), and do not allow us to construct an age-standardized rate.

Continuing with our discussion of fertility, Figure 2.14 provides a graphical representation of the age-specific fertility rates for census years and Table 2.8 provides the estimates of the total fertility rate derived from the vital registration data on births for those same years. Owing to the random fluctuations between years and the small size of the populations of each island, the age-specific fertility data were pooled across islands and the number of births in each age interval was averaged over a 5 year window surrounding the census year. Even after calculating the average number of births for the 5 year window surrounding the census years, the age specific fertility rates and TFRs seem very low, especially when we consider that we have no evidence suggesting effective contraceptives were used at this time in Orkney. The

TFRs, which range between 2.3 and 2.8 are highly questionable for a population characterized by natural fertility (Leridon, 1977; Wood, 1994). Since these estimates were judged to be unrealistic, assuming the population was not actively contracepting, the model fertility methods of Coale and others (Coale 1971; Coale 32 and Tye 1961; Coale and Watkins 1986; Sardon 1996) that use the Hutterite standard fertility rates and the observed population’s female age structure. The method uses the age-sex distribution of women and the total number of births in a given year in the observed population, the Hutterite age-specific fertility rate to calculate an expected age-specific fertility rate for the observed population (for a historical demographic example see Brown 2000). Since previous work (Anderson and Morse, 1993a;

1993b) suggested the nuptiality rate for Orkney was one of the highest rates in

Scotland, I also calculate the Total Marital Fertility Rate (TMFR). The calculations of the model TFR and the TMFR are also reported alongside the observed TFR in

Table 2.9.

Figure 2.14 Age-specific fertility rates calculated from the vital registration data for the Northern Isles, 1861-1901.

0.2

0.18

0.16

0.14

0.12 1861 1871 0.1 1881

ASFR 1891 0.08 1901

0.06

0.04

0.02

0 15_19 20_24 25_29 30_34 35_39 40_44 45_49 50_54 Women's age

33

Table 2.8 Period total fertility rates (TFRs) and total marital fertility rates (TMFR) for the Northern Isles, 1861-1901. TFR (Vital TMFR (Vital TFR (Model Year Registration Registration Estimate) Estimate) Estimate) 1861 2.71 3.84 8.38 1871 2.82 3.48 8.81 1881 2.81 3.28 11.89 1891 2.72 2.83 10.01 1901 2.34 2.84 8.49 Average 2.68 3.25 9.51

The average TFR estimated from the vital registration information over the period

1861 to 1901 is 2.68, almost certainly a poor estimate for a population without modern contraception available. The TFR estimated via the Hutterite standard fertility schedule gives an average of 3.25 children per woman, and while this is still a low value it is somewhat more realistic than the first calculation. The average TMFR, or the average number of children had by a married woman, over the period was 9.51.

This value is similar to those found in small German populations around this time

(Knodel, 1988). If the TMFR calculations are accurate, this indicates that marital fertility was high during this period, a conclusion that agrees with Anderson and

Morse’s (1993a,b) conclusions about Orkney’s high nuptiality levels. If childbearing was primarily confined to married couples, as implied by the high nuptiality values, then some primary determinants of fertility would be the age at marriage and the fraction of women in a married union (Bongaarts 1978; Wood 1994). Figure 2.15 shows the average age of first marriages for grooms and brides over the period of vital registration, along with a ten year running average. Over the course of the nineteenth century we see an increasing age at marriage for both sexes, followed by a 34 declining age at marriage during the years surrounding the first and second world wars (Orkney played a large role in the wars, as it was home to the British fleet).

Figure 2.15 Average age of first marriages for men and women in Northern Orkney 1855-2003.

40

35

30

Average Age 25

20 Groom Age Marriage Bride Age Marriage 10 Year Moving Average (Bride's Age) 10 Year Moving Average (Groom's Age) 15 1855 18 18 1885 18 19 1915 1925 19 1945 19 1965 1975 19 1995 65 75 95 05 35 55 85

Year

Since these averages were calculated from the vital registration data, I also

supplement them with measures of marital age calculated from the available censuses.

The singulate mean age at marriage is used here. This measure of marital age, first

proposed by Hajnal (1953), calculates the average age of marriage from a distribution

of population by ever-married status derived from a census (United Nations, 1983).

These calculations were done for each census in which detailed age-sex distributions

were available (1851 to 1901), and are reported in Table 2.9. The observed mean

ages, calculated from the vital registration data, are also given for comparison.

35

Table 2.9 Singulate mean age at marriage, and observed mean age at marriage for males and females in Northern Orkney, 1851 – 1901. Female Female Male Male Mean Age Mean Age Mean Age Mean Age at Marriage at Marriage at Marriage at Marriage Year (Singulate) (Observed) (Singulate) (Observed) 1851 30.46 - 31.48 - 1861 29.93 24.64 30.28 26.91 1871 28.99 25.64 32.25 27.67 1881 29.69 24.67 31.46 27.58 1891 30.72 25.60 33.57 29.08 1901 31.64 24.73 33.51 27.87

The singulate mean age at marriage is always higher than the observed data by an average of 5.1 years for females and 4.4 years for males, although both show a general trend toward an increasing age at marriage over the course of the 19 th century.

Upon examining the trends in the age at marriage, the female singulate age at

marriage drops to a low of 28.99 in 1871 and then increases to 31.64 in 1901. In

comparison the male mean does not experience such a consistent pattern, and

generally increases over the period. For natural fertility populations, the mean age at

marriage is one of the primary determinants of fertility, and since the ages of

marriage were high, and tended to be increasing over the course of the late 19 th century in the Northern Isles, it is not surprising that the TFR is low.

To expand the analysis of mortality beyond the crude death rates, an attempt was made at constructing an abridged life table for the Northern Isles for the periods covered by available census data. Since in some years none, or a very small number of deaths occurred in the isles, I took a 5 year window around each census year (e.g. for 1861, the deaths in 1859-1863 were used) and averaged the number of deaths for

5-year age categories. By averaging the number of deaths in a single age category 36

(e.g. 0 to 4.999 years) I hope to avoid any random yearly fluctuations in number of deaths by age. I do the averaging for 5 year intervals up to age 85+, and combine all deaths over age 85 into this final age category. The only years I could complete this calculation were 1861, 1871, 1881, 1891, and 1901. The 1851 life table could not be estimated because deaths were not registered until 1855. Table 2.10 shows the calculated life expectancy values at birth, age 15, age 45 and age 65 for each census year 1861- 1901.

Table 2.10 Estimated life expectancy values for Northern Orkney, 1861 to 1901. 1861 1871 1881 1891 1901 Birth 52.5 57.2 58.9 54.5 54.0 Age 15 44.7 48.1 49.2 47.5 47.4 Age 45 28.2 28.9 28.2 27.4 28.3 Age 65 13.9 13.6 14.2 14.2 12.9

We see the life expectancy values at birth, as estimated from the observed data to have an average value of 55.4 years for this period. This figure is obviously not a reliable estimate of life expectancy, because the Registrar General of Scotland estimates the life expectancy for the country of Scotland to be 42.5 for males and 45.4 for females (GROS, 1996), a figure more than 10 years less than the one estimated from the Orkney data. Because the Northern Isles life expectancy values are so much higher than the national average for the period 1861 to 1901 (GROS, 1996), I attempt an indirect standardization of the mortality rates (see Preston et. al. 2000). Since the

GROS does not provide a life table estimate from the nineteenth century, I use the period mortality rates (M-type rates) for England and Wales from the Human

Mortality Database (HMD, 2007) for the following time periods: 1860-64, 1870-74,

1880-84, 1890-94 and 1900-1904. Even though the standard of living was probably 37 higher in England and Wales at this time, the English life table life expectancies are very close to those of Scotland for these periods (HMD, 2007; GROS, 1996). The estimated life expectancies at birth, age 15, age 45 and age 65 are presented in Table

2.11.

Table 2.11 Estimated life expectancy values for Northern Orkney, 1861 to 1901 based on indirect standardization to England and Wales’s mortality rates. 1861 1871 1881 1891 1901 Birth 50.6 53.8 55.7 51.6 51.0 Age 15 52.5 54.0 54.8 53.8 53.0 Age 45 28.8 29.5 29.6 28.3 27.5 Age 65 13.8 14.3 14.3 13.5 13.0

The indirect standardization leads to lower life expectancy at some ages and higher life expectancies at other ages. For example at birth, the average standardized life expectancy is 52.5 years compared to 55.4 for the observed data, a difference of 2.9 years. At the older ages, however we see higher life expectancies produced from the standardization procedure, and across all ages life expectancy is 1.9 years higher than the observed values. The result of the standardization procedure suggest that, although it provides a more realistic infant mortality rate than the observed data, the mortality at the higher ages is too low to compare to the Northern Isles. To get the best estimate of the mortality experience of the Northern Isles for this time period, I combine the observed and standardized mortality rates. I use the infant and child mortality rates from the standardized data, and the adult mortality rates from the observed data. I do this because it is obvious that there is a systematic under- reporting of infant and childhood deaths in the Northern Isles, thus making the observed mortality rates at the young ages too low. But because of the harsh 38 environment and the lower standard of living of the area, the mortality rates of

England and Wales at the adult ages are too low. I use the standardized mortality rates based on the England and Wales life tables for the youngest ages to represent the childhood mortality situation, and the observed mortality rates for the Northern

Isles over age 5 in order to capture the higher mortality rate for Orkney. The resulting life expectancies are reported in Table 2.12.

Table 2.12 Estimated life expectancy values for Northern Orkney, 1861 to 1901 based on a mixture of the observed and standardized mortality rates. 1861 1871 1881 1891 1901 Birth 44.4 49.2 50.6 45.2 45.4 Age 15 44.7 48.1 49.2 47.5 47.4 Age 45 28.2 28.9 28.2 27.4 28.3 Age 65 13.9 13.6 14.2 14.2 12.9

Based on the figures in Table 2.12 we still see life expectancy values that are still high for this time period. The data from England and Wales (HMD, 2007) show values that are much lower than these estimates up until 1891, when the estimated life expectancies are more in agreement. The estimated values are still higher than the

Scottish national life expectancies at all ages for this time period.

Based upon the various methods employed in this analysis it is obvious that mortality is under-reported in the Northern Isles for this time period. The systematic under-reporting of infant deaths is a major problem, but we should also be suspect of the under-reporting of deaths at all ages. The under-reporting of deaths is a common occurrence in many developing nations. After 1855 registration of all vital events

(births, marriages and deaths) were supposed to be reported by the interested parties and recorded by local registrars, but this reporting system was obviously incomplete 39 in its efforts. Further detailed analysis of infant and childhood mortality is presented in Chapter 4, and it should be noted that based on the problems encountered in this descriptive analysis of mortality, that care and caution should be taken in any interpretations from the analysis of the mortality data for the Northern Isles.

Household Structure and Dynamics

This section focuses on the description of households in the Northern Isles between 1851 and 1901. The chapters that follow deal with the in-depth analysis of effects of household dynamics on fertility decision making and childhood mortality, but first it is important to describe the general setting of households in Northern

Orkney. As much work has been done concerning household variation and typology in Europe and other areas of the world (Hajnal 1982; Laslett 1969, 1970; Laslett and

Wall 1972), it is necessary to first describe what is meant by households in this context. For this analysis households are defined as co-resident groups of people who routinely share resources (food, shelter, etc.). Also, since the area of Northern

Orkney is predominantly agricultural the individuals that comprise a household are assumed to share evenly (according to their age and sex) in household labor and productivity, unless compromised by sickness or other limiting factors. Another important distinction to make is the difference between households and farmsteads.

We will use the term household to refer to the people that inhabit the physical location of a farmstead, so in this setting farmstead will describe the physical collection of buildings and other accoutrements that comprise the area where a household spends the majority of its existence. It should be noted that it is a common occurrence that multiple households occupy a single farmstead, and that these 40 households are often joined by some degree of biological or marital connection. It was also very common for unrelated servants to live with families, often in small out buildings, or in the case of larger farms to live in a separate house called a bothy.

We begin with the total number of households in the study area during the late nineteenth century, given in Table 2.13. We see that Sanday and Westray are the largest of the islands with respect to the total number of households, followed by

Eday, North Ronaldsay, Papa Westray and finally Pharay. These differences mimic those of general population size as seen in Table 2.2

.

Table 2.13 Total number of households in the Northern Isles 1851-1901. Island 1851 1861 1871 1881 1891 1901 Eday 220 219 178 166 143 132 North Ronaldsay 89 97 100 91 89 88 Papa Westray 84 82 78 80 70 67 Pharay 14 17 18 13 9 8 Sanday 422 437 420 427 415 395 Westray 432 482 471 486 455 440

Next, Table 2.14 presents the average number of people living in a household on each of the Northern Isles over the period 1851-1901. The largest average household sizes are typically found on North Ronaldsay, which also exhibited the highest population density. The smallest average household sizes typically occur on Eday, the island with the lowest general population density.

41

Table 2.14 Average number of people living in a household in the Northern Isles, 1851-1901. Island 1851 1861 1871 1881 1891 1901 Eday 4.3 4.1 5.6 4.4 4.5 4.5 North Ronaldsay 5.9 5.5 5.4 6.0 5.6 5.0 Papa Westray 4.4 4.8 4.7 4.3 4.1 4.4 Pharay 4.9 4.8 4.6 5.5 6.4 5.9 Sanday 4.7 4.9 4.9 4.9 4.6 4.4 Westray 4.8 4.5 4.4 4.5 4.3 4.1

Since we defined the difference between households and farmsteads in the introductory paragraph, we should also present the differences in average farmstead sizes for the Northern Isles; this is done in Table 2.15. By farmstead size, I am referring to the total number of people living at a given farmstead. This may include related families, servants, or visitors that were in occupation at the time of census enumeration. In comparison to the average sizes of households, farmsteads tend to be nearly twice as large on most islands, the exception to this appearing to North

Ronaldsay, with little difference between the size of households and farmsteads.

While no immediate cause of this is known, we could infer a higher degree of household autonomy on North Ronaldsay in comparison to other islands, or that the households on North Ronaldsay rarely, or employed to a lesser degree, agricultural wage laborers that lived at the individual farms. Sanday displays the largest average farmstead sizes, probably a result of the nucleation of agriculture that has occurred on that island over the nineteenth and twentieth centuries.

42

Table 2.15 Average number of people living at a farmstead in the Northern Isles, 1851-1901. Island 1851 1861 1871 1881 1891 1901 Eday 9.4 8.0 10.2 7.0 6.4 6.1 North Ronaldsay 7.2 7.3 7.3 7.3 7.1 6.2 Papa Westray 8.6 8.3 7.7 7.2 6.8 6.1 Pharay 8.6 9.1 8.3 8.0 7.3 5.9 Sanday 11.9 13.1 11.8 11.0 11.0 9.5 Westray 10.0 9.5 8.6 8.3 7.8 7.4

Another primary focus of the chapters that follow is the differences between farm and nonfarm households. While those familiar with the Northern Isles might see an immediate problem with the distinction between households that farm and those that do not, based on occupational listing in census returns and vital registration records, many households are consistently referred to as being occupied by farmers versus other nonfarm occupational categories. In general the distinction used below and in the chapters that follow is that farm households are assumed to base the majority of their household labor expenditures on the active production of agricultural products

(in Orkney these are generally animal husbandry and grain production) versus those households that focus the lion’s share of their efforts on weaving, smithing, teaching, fishing or wage labor. The average sizes of these farm and general nonfarm households are given in Table 2.16.

We see that, with few exceptions, farm households tend to be larger compared to nonfarm households in Northern Orkney by an average of 1.5 people. North

Ronaldsay and Papa Westray share the largest average difference (1.9 people/household) and Pharay has the smallest average difference (1 person/household).

43

Table 2.16 Average household size of farm and nonfarm households, 1851-1901. Island Farm Status 1851 1861 1871 1881 1891 1901 Eday Nonfarm 3.4 3.3 5.6 3.9 3.9 3.7 Farm 5.5 5.2 5.6 5 5.2 5 North Ronaldsay Nonfarm 3.9 4.1 4 5 4.6 3.8 Farm 6.7 6.1 6.1 6.3 6.2 5.6 Papa Westray Nonfarm 3.6 3.7 3.6 3.2 3.1 3 Farm 5.1 5.6 5.4 5.4 5 5.1 Pharay Nonfarm 3 3.6 4.1 4.8 8 5 Farm 6 5.3 4.9 5.9 6.3 6 Sanday Nonfarm 4 4.2 3.9 3.7 3.8 3.9 Farm 6 5.8 5.8 5.9 5.5 4.8 Westray Nonfarm 3.9 3.8 3.8 4 3.9 3.6 Farm 5.7 5.3 5 5.1 4.8 4.6

We turn now to the description of household composition. Table 2.17 gives the average ages of household heads in Northern Orkney between 1851 and 1901.

The island with the youngest average age of household head is Pharay, and the oldest is North Ronaldsay.

Table 2.17. Average age of household head, 1851-1901. Island 1851 1861 1871 1881 1891 1901 Eday 51.2 50.4 50.7 50.7 51.4 50.6 North Ronaldsay 55.3 53.8 56.7 56.1 54.2 58.1 Papa Westray 45.6 47.6 50.0 50.3 52.6 56.3 Pharay 49.6 42.2 41.9 44.8 53.1 50.0 Sanday 48.6 48.6 51.1 51.5 53.2 54.9 Westray 49.3 49.4 51.3 50.9 51.3 53.0

Pharay also has the highest frequency of male household heads with an average value of 87.4%, while North Ronaldsay has the lowest average percentage of male household heads with 81.2%.

44

Table 2.18 Percentage of male household heads, 1851-1901. Island 1851 1861 1871 1881 1891 1901 Eday 76.36 82.38 83.62 82.42 83.10 94.66 North Ronaldsay 79.31 89.47 82.00 81.32 73.26 81.82 Papa Westray 92.86 84.81 89.61 87.50 81.43 83.08 Pharay 100.00 88.24 80.00 92.31 88.89 75.00 Sanday 84.21 83.53 83.90 84.71 82.28 82.99 Westray 86.74 85.47 82.56 85.23 89.64 90.59

With respect to household types, Table 2.19 gives the percentages of extended family households in Northern Orkney. Extended family households have at their core a nuclear family of a husband-wife pair, possibly but not necessarily with children, and at least one either vertical or horizontally related kin member residing with the nuclear family at the time of census enumeration. Pharay by far has the highest percentage of extended family households, while Sanday has the lowest percentage.

Table 2.19 Percentage of extended family households, 1851-1901. Island 1851 1861 1871 1881 1891 1901 Eday 3.2 5.9 7.3 6.6 6.3 5.3 North Ronaldsay 7.9 5.2 8.0 6.6 4.5 3.4 Papa Westray 4.8 8.5 5.1 6.3 2.9 3.0 Pharay 7.1 5.9 16.7 38.5 22.2 12.5 Sanday 5.5 4.6 5.5 4.0 3.6 2.8 Westray 4.6 4.4 4.7 6.6 4.0 3.9

Discussion

This goal of this chapter was to describe the demographic setting of the Northern

Isles, mostly focusing on the mid 19 th to early 20 th centuries. In aggregate we see a

small population, rural in character with a strong reliance on agriculture and fishing 45 as the primary industries. Using comparative information from other areas of

Scotland, we see the Northern Isles having nearly the lowest rates of fertility and mortality of any area of the country, with a few exceptions. At most ages and time periods there is a shortage of marriage age males in the islands, and in turn we see a low fraction of the female population in martial unions. If the data from the vital registration system are correct, the Northern Isles have an extremely high survival rate for the 19 th century, possibly owing to their relative isolation from infectious

disease outbreaks. The poor marriage market, little access to farm land and higher

than average longevity could combine to produce what we have seen as the dominant

trend in the Northern Isles during the 20 th century: emigration.

To illustrate this we need only look as far as the standard population growth

model. In 1861, there were 3620 people living in the Northern Isles (excluding

Sanday because the vital registration data were not in the possession of the author),

between 1861 and 2001 there were 6472 births and 4749 deaths. This translates into

a natural increase of 1723 people, producing a population size in 2001 of 5343,

assuming population growth is only attributable to natural increase. The observed

population of these islands in 2001 was 819 people, a difference of 4524 indicating a

net loss of 2801 people from the 1861 figure; or, expressed as a fraction, the 2001

population was 15.3% of what it would have been in the absence of emigration.

The people of Orkney have always had to deal with what nature and their

geographic location has given them. During the late 18 th century, when the British

glass markets demanded cheap alkali to fuel their production, the people of Orkney

jumped at the chance to make their location turn into a real cash economy through the 46 processing of their naturally occurring seaweed. When the kelp market crashed in the mid 19 th century, they reorganized their farms and modified their lands to make the

most of the poor soils they have and started raising sheep and cattle as their primary

produce. As land was bought up by larger farms to sustain the production of sheep

and cattle, many families could not keep up because they quite literally had no money

to compete, and many left for the Orkney Mainland or places further afield. Those

that remained during the early 20 th century saw two wars take many of the young men

and women to the south of the British Isles to aid in the war effort, many of whom

were never to return to their native Orkney. Coinciding with the inter-war period, the

large estates that had made up the majority of Orkney began to be broken up by their

owners, allowing the residents of Orkney (or at least those with the means) to actually

own the land their families had worked for generations. Many families that remained

used their small plots of land in the rolling hills of the Northern Isles to place large

chicken coops, and thanks to the demand for eggs for the second and the

markets to the south, made (what could be considered by Orcadian standards) small

fortunes selling their eggs. These families used the proceeds of their household labor

to purchase more of the commodity most valuable to farmers, land. Families without

land faced a bleaker future, and indeed many families left Orkney for Australia, New

Zealand, and America. Small communities like Pharay became hopelessly

unsustainable and were abandoned, other small islands still face very difficult

decisions on whether they face an unsustainable future. Still other islands were left

mostly devoid of native Orcadians, and were resettled by migrants from England.

The native Orcadians that remain in the Northern Isles however represent the heartiest 47 of the hearty because they and their families saw the bad times, the really bad times, and the times they do not even want to discuss.

With the increased ferry and steamer service of the 20 th century and the access to electricity, cable and phone service, many more Orcadians have decided to seek their opportunities elsewhere and during the 20 th century the dominant demographic phenomena that characterized Orkney was emigration. So it seems that the ingenuity of the Orcadians once again overcame their environmental and social setting and the choice of many to leave versus face an uncertain future in their homeland. For those who remain, a new Orkney has emerged from the old subsistence system. Tourism, , large scale farming, and the growing global economy have once again given Orcadians opportunities to carve out their own niche and make the most of what they have.

References

Anderson M (1998) Fertility decline in Scotland, England and Wales, and : comparisons from the 1911 census of fertility. Popul Stud (Camb) 52: 1-20.

Anderson M, and Morse DJ (1993a) High fertility, high emigration, low nuptiality: adjustment processes in Scotland's demographic experience, 1861-1914, Part I. Popul Stud (Camb) 47: 5-25.

Anderson M, and Morse DJ (1993b) High fertility, high emigration, low nuptiality: adjustment processes in Scotland's demographic experience, 1861-1914, Part II. Popul Stud (Camb) 47: 319-43.

Barclay RS (1965) The population of Orkney 1755-1961. Kirkwall: Mackintosh.

Berry RJ (1985) The natural . London: Collins and Son.

Bourgeois-pichat J (1981) Recent demographic change in western Europe: an assessment. Popul Dev Rev 7: 19-42. 48

Bowers EJ (1983) Patterns of adult mortality in the Orkney Islands. Ph. D. dissertation, University of Pennsylvania, Philadelphia.

Brennan E (1979) Kinship, demographic, social, and geographic characteristics of mate choice in a small human population. Ph. D. dissertation, Pennsylvania State University, University Park.

Brennan ER (1983a) Mortality patterns in anthropological populations. Hum Biol 55: 1-7.

Brennan ER (1983b) Pre-reproductive mortality and family structure: Sanday, Orkney Islands 1855-1974. Hum Biol 55: 19-33.

Brennan ER, Leslie PW, and Dyke B (1982) Mate choice and genetic structure Sanday, Orkney Islands, Scotland. Hum Biol 54: 477-89.

Brown DAV (2000) The political economy of fertility in the British West Indies 1891-1921. Barbados: University of The West Indies Press.

Coale AG (1971) Age patterns of marriage. Population Studies 25 :193-214.

Coale AJ, and Tye CY (1961) The significance of age-patterns of fertility in high fertility populations. The Millbank Memorial Fund Quarterly 39 :631-646

Coale A, and Watkins SC, eds. (1986) The decline of fertility in Europe. Princeton: Princeton University Press.

Davidson DA, and Simpson IA (1994) Soils and landscape history: case studies from the Northern Isles of Scotland. In A Foster and TC Smout (eds.): The history of soils and field systems. : Scottish Cultural Press.

Dodgshon RA (1994) Budgeting for survival: Nutrient flow and traditional farming. In S Foster and TC Smout (eds.): The history of soils and field systems. Aberdeen: Scottish Cultural Press.

Fenton A (1978) The Northern Isles: Orkney and Shetland. Edinburgh: John Donald.

Firth J (1974) Reminiscences of an Orkney parish. : Orkney Natural History Society.

Flinn M, ed. (1977) Scottish population history from the 17th century to the 1930s. Cambridge: Cambridge University Press. 49

Gage TB (1993) The decline of mortality in England and Wales 1861 to 1964: decomposition by cause of death and component of mortality. Popul Stud (Camb) 47: 47-66.

Gibson AJS, and Smout TC (1995) Prices, food, and wages in Scotland 1550-1780. Cambridge: Cambridge University Press.

GROS (1996) 1996 Annual report of the Registrar General of Scotland. Accessed online at www.gro-scotland.gov.uk/statistics/annrep/.

Guinnane TW, Okun BS, and Trussell J (1994) What do we know about the timing of fertility transitions in Europe? Demography 31: 1-20.

Gunnlaugsson GA (1988) Family and household in 1801-1930. Stockholm: Almqvist and Wiksell International.

Hajnal (1953) Age at marriage and proportions marrying. Population Studies 7: 111- 136.

Hajnal J (1982) Two kinds of preindustrial household formation system. Popul Dev Rev 8: 449-94.

Harrison GA (1976) Genetic and anthropological studies in the human adaptability section of the International Biological Programme. Philos Trans R Soc Lond B Biol Sci 274: 437-45.

HMD (2007) Human Mortality Database: University of California, Berkeley (USA), Max Planck Institute for Demographic Research (Germany). Accessed online at www.mortality.org.

Knodel J (1977) Family limitation and the fertility transition: Evidence form the age patterns of fertility in Europe and Asia. Population Studies 31: 219-249.

Knodel JE (1988) Demographic behavior in the past: A study of fourteen German village populations in the eighteenth and nineteenth centuries. Cambridge: Cambridge University Press.

Laslett P (1969) Size and Structure of the Household in England Over Three Centuries. Population Studies 23: 199-223.

Laslett P (1970) The Comparative History of Household and Family. Journal of Social History 4: 75-87. 50

Laslett P, and Wall R, eds. (1972) Household and family in past time. Cambridge: Cambridge University Press.

Leridon H (1977) Human Fertility. Chicago: University of Chicago Press.

Macbeth HM, and Boyce AJ (1987) Anthropometric variation between migrants and non-migrants: Orkney Islands, Scotland. Ann Hum Biol 14: 405-14.

Marwick H (1952) Orkney Farm Names. Kirkwall: W. R. Mackintosh.

McKeown T, Record RG, and Turner RD (1975) An interpretation of the decline of mortality in England and Wales during the twentieth century. Popul Stud (Camb) 29: 391-422.

Pearson AW, and Collier P (1998) The integration and analysis of historical and environmental data using a geographic information system: landownership and agricultural productivity in Pembrokeshire c. 1850. Agricultural History Review 46: 162-176.

Pooley C, and Turnbull J (1998) Migration and mobility in Britain since the eighteenth century. London: University College of London Press.

Preston SH, Heuveline P, and Guillot M (2000) Demography: Measuring amd modeling population processes. London: Basil Blackwell.

Relethford JH, and Brennan ER (1982) Temporal trends in isolation by distance on Sanday, Orkney Islands. Hum Biol 54: 315-27.

RGS Detailed annual report of the Registrar-General of Scotland. Edinburgh and London: HMSO.

Roberts DF (1983) Genetic epidemiology. Am J Phys Anthropol 62: 67-70.

Roberts DF, and Roberts MJ (1983) Surnames and relationships: an Orkney study. Hum Biol 55: 341-7.

Ruggles S, Sobek M, Alexander T, Fitch CA, Goeken R, Hall PK, King M, and Ronnander C (2004) Integrated Public Use Microdata Series: Version 3.0: Minnesota Population Center.

Russell JC (1948) British Medieval Population. Albuquerque: Univerisity of New Mexico Press. 51

Sardon J-P (1996) Coale’s indices, comparative indices, mean generation, total fertility rate and components. Population: An English Selection 8:252-257.

Schrank G (1995) An Orkney estate: Improvements at Graemeshall 1827-1888. East Linton: Tuckwell Press.

Scott N, Stevenson C, and Stout A, eds. (2003) Fae Quoy tae Castle: The buildings of Westray: Westray Building Preservation Trust.

Smith RM (1981) Fertility, economy, and household formation in England over three centuries. Popul Dev Rev 7: 595-622, 728-30.

Thomson WPL (1983) Kelp making in Orkney. Kirkwall: The Orkney Press.

Thomson WPL (2001) The new history of Orkney. Edinburgh: Mercat Press.

United Nations (1983) Manual X: Indirect techniques for demographic estimation. New York: United Nations Press.

Wenham S (2001) A More Enterprising Spirit: The Parish and people of in the 18th century Orkney. Kirkwall: Bellavista Publications.

Whyte I (1979) Agriculture and society in seventeenth century Scotland. Edinburgh: John Donald.

Wood JW (1994) Dynamics of human reproduction: Biology, biometry, demography. New York: Aldine De Gruyter.

Woods RI, Watterson PA, and Woodward JH (1988) The causes of rapid infant mortality decline in England and Wales, 1861-1921, Part I. Popul Stud (Camb) 42: 343-66.

Woods RI, Watterson PA, and Woodward JH (1989) The causes of rapid infant mortality decline in England and Wales, 1861-1921. Part II. Popul Stud (Camb) 43: 113-32.

WorldClimate (2007) WorldClimate.com. www.worldclimate.com.

Wrigley EA, and Schofield RS (1983) English population history from family reconstitution: summary results 1600-1799. Popul Stud (Camb) 37: 157-84.

Wrigley EA, and Schofield RS (1989) The population , 1541-1871: A reconstruction. Cambridge: Cambridge University Press. 52

Chapter 3 Analysis of birth spacing behavior in the Northern Orkney Islands: 1855-2003.

Introduction

Fertility within agricultural populations has often been suggested to be higher than other areas (Firebaugh 1982; Vanlandingham and Hirschman 2001) primarily because of the demand for child labor in agricultural settings (Becker 1981; Bentley et al. 1993; Boserup 1987; Clay and Johnson 1992; Johnson 1990; Markle and Pasco

1977; Rosenzweig 1977; Schutjer et al. 1983; Singh 1979) and often a result of high infant mortality (Galloway et al. 1994; Palloni and Rafalimanana 1999; Sanderson and Dubrow 2000; Tolnay and Glynn 1994). Although we know little about the demographic dynamics of the population of Orkney, Scotland (Bowers 1983;

Brennan 1983b; Brennan et al. 1982; Brennan and Relethford 1983; Roberts and

Roberts 1983), some authors have noted Orkney as being the exception rather than the rule with respect to high agricultural fertility levels (Anderson and Morse 1993a, b; Flinn 1977). Indeed, Anderson and Morse note that because the level of fertility in

Orkney was so low, fertility must have fallen early in contrast to similar

(small-holder farming) regions of Scotland sharing similar economic conditions

(Anderson and Morse, 1993a, p. 18) during the mid 19 th century. It should also be noted that, in comparison to other crofting counties, Orkney displayed higher nuptiality levels, more similar to more urbanized and industrial areas to the south

(Anderson and Morse 1993a, b). So why, given the high nuptiality levels, was the level of fertility so low in Orkney at this time? This chapter will analyze two measures of fertility: the first being length of the interval between marriage and first 53 birth, and the second being the intervals of all subsequent births in hopes of answering some of the inconsistencies from previous analyses mentioned above. I will also test whether agricultural families show higher demand for children, as evidenced by shorter inter-birth intervals, and if there is evidence for kin-based support networks allow for a faster pace of reproduction.

Two main hypotheses will be tested in the following analysis. First, based on previous work analyzing the economic aspects of fertility behavior in agricultural communities we expect farm families to have shorter birth intervals in order to increase their household labor supply. Evidence has often shown the connection between farm status, amount of land held and number of offspring needed to work the family farm (Johnson 1990; Schutjer et al. 1983). While currently we have no data on size of land holding, we assume that households whose primary means of income is farming (see methods section) need more children to offset the need for household labor. Secondly we expect the presence of both maternal and paternal grandparents will tend to shorten birth intervals because they provide a social network resource parents may exploit in order to maximize their reproductive output. While grandparents have shown a significant impact on child survivorship and nutritional status (Sear et al. 2000; Sear et al. 2002; Strassmann and Gillespie 2002; Voland and

Beise 2002), their influence has been investigated to a lesser degree on patterns of birth spacing (Mace and Sear 1997; Sear et al. 2001; Tymicki 2004; Voland 1998).

Research that has shown effects of kin in reproduction indicate an increase in the fertility rate brought on by kin acting as caretakers for children thus improving their survival chances (Sear et al. 2000; Voland and Beise 2002) and by increasing the 54 nutritional status and decreasing the workload of the mother, although the authors themselves state that the evidence is for this by no means certain (Tymicki 2004).

Data and Methods

The North Orkney Population History Project (NOPH) has entered into a

database all birth, marriage and death registered in the Northern Orkney Islands

between 1855 and 2001. The resulting database included 7410 births, 2025 marriages

and 5141 deaths for the islands of Eday (and Pharay), North Ronaldsay, Papa

Westray and Westray. Although another island, Sanday, is part of the Northern Isles,

at the time this analysis was undertaken the vital registration data had not been linked

and were unavailable to the author. All marriages that occurred on the islands

between 1855 and 2003 were linked to birth records for all children produced from

the marriage. The linkage was generated by using father’s first and surname,

mother’s first and maiden name, and child’s year of birth as linking criteria. This

generated a total of 4585 births that could be linked to marriages between 1855 and

2001, and 1672 censored open birth intervals, yielding a total of 6256 birth intervals

generated by 1991 different first marriages for analysis. These open intervals are

important because they represent marriages that produced no children either because

of sterility or out-migration. Marriage dissolution represents another possible cause

for the open birth intervals. If the vital registration data are complete (many

indications point to them not being complete) there were no occurrences of divorce in

the Northern Isles until the late 1970’s. Although formal divorces may have been

uncommon, it is not uncommon for couples in the Northern Isles to undergo

voluntary mutual separations where the husband and wife no longer live together, but 55 are not technically divorced. Maternal or paternal death would also lead to the premature termination of a marriage. As of the time this analysis was undertaken, the appropriate linkages have not been completed to estimate the level of maternal or paternal mortality. All open birth intervals, regardless of cause, will be treated as right censored observations in the analysis which follows.

I am examining the influences of various individual and household level variables on the lengths of birth intervals. First I will discuss the individual and household level variables, and then characteristics of the child’s mother and father.

For the variable child’s sex, 1 represents males and 0 represents females. Farm household status is 1 if the child’s father’s occupation was listed as “farmer” or

“crofter” in the child’s death record, else the code is 0.3 In the Orkney setting this

distinction is a very tenuous one, as most everyone living in the islands participates in

agriculture at some time of the year, and likewise many of the farmers participate in

activities other than farming. Another distinction that should be used instead of

farm/nonfarm is landowner verses renter verses crofter, but this distinction cannot be

made based on the information available in the vital registration records. The

farm/nonfarm distinction is used here as a proxy measure of land renters or owner

compared to everyone else. Number of paternal and maternal grandparents alive at

the index 4 child’s birth does not indicate that the child’s grandparents were living in

the household with the child, only that they were alive at the time of the child’s birth.

3 While most of the population of Orkney was probably participating in agriculture as part of their daily lives, I identify farmers here are those whose occupations were listed as such. They are presumably less likely to pursue other forms of work than fathers listed as agricultural laborer or fisherman. The farm household category is intended to represent those families who were participating solely in agriculture. 4 Index child, as it is used throughout this chapter refers to the child whose birth closed the birth interval. 56

As little work has been done on the impacts of these post-reproductive individuals on fertility, I am including these measures to test for any association with length of birth interval. I assume here that the presence of these individuals in the community will have one of two possible impacts: 1, they will reduce birth intervals by adopting childcare responsibilities from the mother, presumably allowing here to shorten the period of post-partum amenorrhea; or 2, they will increase birth intervals because their presence could represent an additional cost for the household, effectively becoming a net burden on the household instead of a net benefit. For the analysis of higher order births, I include birth order, or parity of the index child, in the analysis. I also include a dummy variable indicating whether the previous child died in the inter- birth interval; this variable takes the value 1 if the previous child died before this child was born and 0 otherwise. The death of the previous child has been shown in a variety of settings to shorten the inter-birth interval by allowing the mother to resume her ovulatory cycle faster (see Wood, 1994). Parents’ variables include (for both the mother and the father) age (in years) at the time of child’s birth, the square of age at the time of child’s birth (included to test for nonlinear effects of age). For the higher order births the duration of the parents’ marriage measured in years is included.

Mother and father’s birth cohort is used to test for trends in cohort spacing behavior, and is the decade of birth of the mother or father. For second and higher order births, first birth interval length is included to control for heterogeneity in fecundability

(Larsen and Vaupel, 1993; Wood 1994; Wood et. al. 1994). Descriptive statistics are given for all variables included in both analyses in Tables 3.1 and 3.2.

Table 3.1 Descriptive statistics for first birth interval analysis. 57

Variables Mean Std Min Max n Sex, male=1 .495 0 1 1297 Event Censored (1=uncensored) .998 0 1 1298 Farm/Nonfarm, Farm=1 .341 0 1 1298 # Paternal Grandparents Alive 1.56 .62 0 2 1271 # Maternal Grandparents Alive 1.68 .54 0 2 1275 Mother’s Age 23.31 5.32 16 51 1298 Mother’s Age 2 666.88 297.67 245.44 2601.00 1298 Father’s Age 28.32 6.75 16.33 62.5 1297 Father’s Age 2 847.80 445.31 266.66 3906.25 1297 Mother’s Birth Cohort 1880.17 37.15 1820 1980 1297 Father’s Birth Cohort 1876.89 36.91 1800 1970 1279 Islands Eday 257 North Ronaldsay 142

Papa Westray 89 Westray 810

Table 3.2 Descriptive statistics for higher order birth interval analysis. Variables Mean Std Min Max n Sex, male=1 .503 0 1 3262 Event Censored (1=uncensored) .762 0 1 4281 Farm/Nonfarm, Farm=1 .356 0 1 4281 # Paternal Grandparents Alive 1.61 .59 0 2 3229 # Maternal Grandparents Alive 1.69 .51 0 2 3219 Duration of Marriage (years) 8.98 5.37 -2.83 35 3262 Mother’s Age 36.17 9.86 16 57 4281 Mother’s Age 2 1403.22 753.85 266.77 3249.00 4281 Father’s Age 34.68 7.58 16.33 66.33 3262 Father’s Age 2 1260.52 565.94 266.66 4400.11 3262 Mother’s Birth Cohort 1869.69 31.81 1820 1970 3262 Father’s Birth Cohort 1866.24 31.59 1800 1970 3262 First birth interval length (months) 12.43 19.81 -84.00 252.00 4281 Death of previous child .047 .213 0 1 3262

Islands Eday 717 North Ronaldsay 570

Papa Westray 266 Westray 2728 Birth Order 0 to 3 2688 4 to 6 1087 7+ 506

I use Cox proportional hazard models (Cox 1972) to examine the effects of family and child characteristics on inter-birth intervals. The proportional hazard model estimates the effect of observed covariate patterns on the risk of a birth 58 occurring at a certain time for a family with a given set of covariates, relative to families with different patterns of covariates. The model is generally written:

p

hi (t) = λ0 (t)exp( ∑βk zk ) k =1 with hi(t) being the hazard of experiencing the event at time t, λ0(t) being a baseline

hazard rate that is left unspecified (this feature distinguishes the proportional hazard

model from fully parametric hazard models (Wood et al. 1992) and is generally

represented by an individual having 0’s for all covariate values, β being the estimated

regression coefficients and being the individual covariate values. This model produces an estimate of the individual’s hazard at time t , given the baseline risk and the linear effect of the covariates.

For both analyses I estimate three nested models, the first incorporating characteristics of the child (sex, birth order and death of the previous child) and general characteristics of the household (farm status and number of living paternal and maternal grandparents). The second model incorporates only characteristics of the mother and father (age, age 2, parents’ birth cohort, duration of marriage and

length of the first birth interval). The full model incorporates child, household and

parents’ variables. To test for the relative information content of the models, I

construct likelihood ratio tests of each of the nested models (Model 2 vs. Model 1,

and Model 3 vs. Model 2).

Results I will present the results of the two analyses in turn starting with the analysis

of time to first birth. Table 3.3 presents the parameter estimates, standard errors and

model fit statistics for the three Cox proportional hazard models fit to the first birth 59 interval data. It should be noted that since the Cox model is estimating the effects of each covariate on the hazard function, a negative regression coefficient decreases the hazard rate, in this case lengthening the birth interval, and a positive regression coefficient increases the hazard rate, thereby shortening the birth interval.

When only considering the first model, farm household status is the only statistically significant effect. Farm households have a reduced hazard for the interval between marriage and first birth compared to nonfarm households, indicating they typically have longer intervals than their nonfarm counterparts. Farm households have 19% decrease in the hazard rate (hazard ratio=.811, 95%CI .720-

.914). This model fits the data rather poorly with a model R 2 of only .001, suggesting

that the majority of variation in first birth interval length is not explained by these

factors.

Table 3.3 Results of Cox regression analysis of first birth interval. Variable Model 1 (s.e.) Model 2 (s.e.) Model 3 (s.e.) Sex, male=1 -.024 (.058) -032(.057) Farm/NonFarm, Farm=1 -.209(.061) -.161(.061) #Paternal Grandparents .032(.046) -.076(.048) #Maternal Grandparents -.025(.051) -.192(.054) Mother’s Age -.154(.039) -.154(.040) Mother’s Age 2 .001(.000) .001(.000) Father’s Age -.170(.028) -.179(.028) Father’s Age 2 .002(.000) .002.(.000) Mother’s Birth Cohort -.005(.007) -.006(.007) Father’s Birth Cohort -.004(.007) -.002(.007)

n 1206 1253 1206 Model -2LL 9440.093 9393.205 9000.711 x2 (Wald) 13.38 418.79 423.44 df, p 4, .0096 6,<.0001 10,<.0001 Model R 2 .001 .045 .047 Likelihood Ratio Test for model improvement x2, p 93.78, <.0001 784.98, <.0001

60

Figure 3.1 Density estimates for the first birth interval for farm and nonfarm households.

As an illustration of the effect of farm household status, the estimated densities of the

first birth interval, stratified by farm household status are given in Figure 3.1.

The second model includes only parent’s characteristics in the analysis. Not

surprisingly, both mother’s and father’s ages and ages 2 have significant effects. Older mothers and fathers typically have longer intervals. Mother’s age reduces the hazard by 15% (hazard ratio .857, 95% CI .793-.927) and father’s age reduces the hazard by slightly more than maternal age, 16% (hazard ratio .844, 95% CI .798-.891). For women, these hazard rates are primarily interpreted as the difference in effective fecundability (including possible pregnancy loss) between older and younger women.

As an example, using the estimated regression coefficients from Model 2, a woman who is 40 years old has only about 15.3% of the effective fecundability of a 20 year 61 old woman, all other things being equal. This effect is estimated by calculating the

2 e(−.154 *40 +.001 *40 ) hazard ratios for two different ages of women 2 = .153 . For fathers, a e(−.154 *20 +.001 *20 )

40 year old father has 36.8% of the effective fecundability (if this term holds for a

male) than his 20 year old counterpart. The higher fecundability for males illustrates

the ability of older males to still sire offspring at a high rate, relative to women, as

well as the age difference between the spouses. This model represents a significant

improvement in model fit in comparison to Model 1, with a R 2 value of .04 and a

significant likelihood ratio test. This suggests that mother and father’s age effects are

the important factors controlling the pace of first birth, which is to be expected.

The full model shows a different effect pattern than the previous models. The

effects of farm household status and completed family size decline somewhat relative

to Model 1, but remain significant. In the full model the number of living maternal

grandparents becomes a significant effect and tends to increase the birth interval by

18% (hazard ratio .825, 95% CI .741-.919), while the number of paternal

grandparents remains insignificant. Mother and father’s age and age 2 stay approximately the same as in Model 2, with no change in the coefficient for mother’s age and a slight increase in the coefficient for father’s age. Model fit is significantly improved with a significant likelihood ratio test over Model 2, although Model 3’s R 2

value remains at only 4.7%. The results of these models indicate a substantial portion

of the variation in first birth interval is left unexplained by the variables included

here.

62

Table 3.4 presents the results of the analysis of higher order birth intervals.

Many of the variable from the first birth interval analysis are included in this analysis, with the addition of child’s birth order, duration of parents’ marriage, death of the previous child in the inter birth interval, and the length of the first birth interval. The first model indicates significant effects of farm households, index child’s birth order and death of the previous child. Once again, we see the negative coefficient for farm households, reducing the hazard by 9% (hazard ratio .911, 95% CI .850-.978) compared to nonfarm households. Higher order births tend to have shorter birth intervals. As an example of this effect, if the child is of parity 4 to 6, compared to 2 to 3, the interval is 7% shorter on average (hazard ratio 1.072, 95%CI 1.020-1.125).

We see a major influence of the death of the previous child on the index child’s birth interval, decreasing it by 53% (hazard ratio 1.535, 95%CI 1.303 – 1.809). This presumably represents the effects of the premature end to breastfeeding brought on by the previous child’s death and the return to ovulation by the mother. This effect is illustrated in Figure 3.2, which gives the density of the birth interval lengths for the various birth orders and for the births that followed deaths of the previous birth.

63

Figure 3.2 Density estimates for various birth orders, indicating whether the previous child had died in the interval.

Although we see these impacts, as a whole model fit is very poor, with an R 2 value of

only 1%.

The second model, again considering only parental characteristics, indicates

significant effects of marital duration, mother’s age, mother’s age 2 and first birth interval length. Higher marital durations tend to lengthen birth intervals by 9%

(hazard ratio .919, 95% CI .908-.929). An example of this would be if a couple was married 10 years they would have birth intervals 65% longer than a couple married 5

 (−.085 10* )   e  years  ( .085 )5* = .653  .  e − 

64

Table 3.4 Results of Cox Regression Analysis of Higher Birth Intervals. Variable Model 1 (s.e.) Model 2 (s.e.) Model 3 (s.e.) Sex, male=1 .018(.035) .080(.035) Farm/NonFarm, Farm=1 -.093(.035) .001(.037) #Paternal Grandparents -.006(.030) .009(.031) #Maternal Grandparents -.002(.034) .043(.036) Birth order .069(.024) .839(.019) Death of previous child .428(.083) .588(.085) Duration of Marriage -.085(.005) -.444(.011) Mother’s Age -.122(.025) -.123(.027) Mother’s Age 2 .001(.000) .001(.000) Father’s Age -.026(.019) -.044(.019) Father’s Age 2 .000(.000) .000(.000) Mother’s Birth Cohort -.004(.004) .009(.004) Father’s Birth Cohort -.000(.004) -.013(.004) First birth interval length (months) .002(.001) .034(.001)

n 3186 3208 3139 Model -2LL 28164.075 27673.781 25072.476 x2 (Wald) 41.56 668.94 2009.59 df, p 6,<.0001 8,<.0001 14,<.0001 Model R 2 .001 .026 .097 Likelihood Ratio Test for model improvement x2, p 980.58, <.0001 5202.61,<.0001

As in the previous analysis, we see a negative impact of maternal age (hazard ratio .885, 95% CI .841-.931) that tends to lengthen birth intervals, and a slight positive effect of the quadratic term mother’s age 2 (hazard ratio 1.002, 95% CI 1.001-

1.002) that decreases interval length slightly. Following the example set in the analysis of first birth interval length, we see that a 40 year old woman has roughly 9% of the effective fecundity of a 20 year old woman, all other things being equal. We also see a significant effect of the length of the first birth interval, suggesting some heterogeneity in general fecundability in this sample (hazard ratio 1.002, 95%CI

1.000-1.005), although the effect is small. Following the example of mother’s age above, we see that the relative fecundity of a family with a first birth interval length 65 of 18 months has 97.9% of the fecundity of a couple with a first birth interval length of 9 months. The model R 2 for Model 2 is .026, indicating a small amount of the

variance in higher birth intervals is attributable to parental effects only, but a

noticeable improvement over Model 1.

Model 3 once again represents the full model combining both child’s and

parent’s characteristics. We see significant effects of sex, child’s birth order, death of

the previous child, marital duration, mother’s age, square of mother’s age, both

parent’s birth cohorts and the length of the first birth interval. First, we see an effect

of child’s sex, indicating a reduction in the birth interval by 8.4% (hazard ratio 1.084,

95%CI 1.010 – 1.163) for male births compared to females. While this suggests

some sex preference, we would ideally like to add the information on the sex of the

previous offspring to the model to measure this effect. Once again we see the effect

of birth order decreasing the birth interval length by a factor of 2.3 (hazard ratio

2.315, 95%CI 2.228 – 2.406). This means that if, for example, a child was of birth

order 2 or 3, their birth interval would be 2.3 times as long as a child of birth orders 4

to 6. If the previous child died in the interval, we see a reduction of the index child’s

birth interval by 80% (hazard ratio 1.80, 95%CI 1.521 – 2.131). The effect of marital

duration continues to show an effect of lengthening the birth interval by 36% (hazard

ratio .641, 95%CI .626 - .656). This effect translates into a couple that has been

married 10 years has a birth intervals that are on average, 11% longer than a couple

 (−.444 10* )   e  that has been married 5 years  ( .444 )5* = .108  . Mother’s age and the square of  e −  mothers age has approximately the same effect that it did in Model 2. Father’s age becomes significant in Model 3 indicating a lengthening of birth intervals for older 66 fathers. For example, a father who is 50 years old has 33% of the effective fecundability of a male aged 25, all other things being equal. Both Mother’s and father’s birth cohort have significant influences on higher order birth orders.

Mother’s birth cohort tends to shorten the birth interval by 1% (hazard ratio 1.010,

95%CI 1.001 – 1.019). An example of this would be that a child born to a woman who was born in 1910 would have a birth interval 1.64 times shorter than a woman born in 1860. This result seems peculiar, as we should expect fertility to decrease in the 20 th century compared to the 19 th century, but we see the opposite effect here.

Father’s birth cohort indicates children born to fathers who born later tend to have 2% longer birth intervals (hazard ratio .986, 95%CI .978 - .995). Again, for example a child born to a father born in 1910 would have a birth interval 49% longer than a child born to a father born in 1860. The length of the first birth interval again has an effect on subsequent birth intervals indicating an effect of heterogeneity in fecundability. The estimated coefficient in Model 3 indicates that, for example, a couple whose first birth interval length was 18 months has a birth interval 1.36 times longer than a couple whose first birth interval was 9 months long. Model 3 represents a significant improvement over Model 2 as indicated by the likelihood ratio test, but the Model fit is still questionable because the R 2 value remains at only 9.7%.

Discussion

When we consider our two primary hypotheses, we see no support for the first and some suggestive support for the second. After controlling for other family and individual characteristics, farm families actually tend to have longer first birth 67 intervals and there is some suggestive evidence that the same effect holds for higher order births as well. With respect to the effect of grandparents on first birth spacing, we only see an effect in the full model (Model 3) for maternal grandparents. This effect is negative indicating that having more maternal grandparents alive tends to increase the time to the first birth, all things being equal. While number of paternal grandparents alive stayed non-significant in the full model, the effect was in the same direction as the maternal grandparent coefficient. Associated with these findings, we see the negative effects of maternal age and marital duration being major effects.

Since fecundity is often thought to decline with age (Wood 1994), the effect of maternal age is not surprising since it is associated with longer birth intervals in both analyses. The model estimates presented here indicated that for the first birth interval, a woman 40 years old had 15% of the effective fecundability of a woman 20 years her junior, indicating the major influence of a woman’s age on her fertility. The effect of marital duration could be indicative of permanent sterility of some couples, whereby some couples may have very long birth intervals or may never produce a full term birth. Like mother’s age, father’s age also shows the effect of lengthening the first birth interval. While to some degree mother’s and father’s ages at the birth of the child should be correlated, there is perhaps less biological reason why we should see this negative effect. Like the effect of age and marital duration in general, this probably speaks to the lowering of coital frequency with age, the rising incidence of pregnancy loss, and the decline of other behavioral factors associated with marital fertility and possibly the decline in sperm quality with age (Wood 1994). The effect of birth order shows the expected pattern, and to illustrate this finding, I present in 68

Figure 3.3 the mean completed inter-birth interval by parity and completed family size (Leridon 1977; Wood, 1994). Upon examination of the average intervals, we see low completed family sizes having longer intervals on average, while couples of higher completed family size have shorter birth intervals at all parities. We also often see that the final parity for a given completed family size is generally longer than any of the proceeding intervals.

Figure 3.3 Mean inter-birth interval length by completed family size and parity.

60 3 4

50 10 5 6 2 8 2 3 40 13 4 9 7 11 5 6 12 7 30 8 9 14 10 11 12 Mean birthMean interva l(months) 20 13 14

10

0 1 2 3 4 5 6 7 8 9 10 1 1 13 14 1 2 Parity of birth that opens interval

With respect to period difference, we see no effects of mother or father’s birth cohort on timing of first birth, although we do see effects on higher order birth intervals.

While women born in more recent birth cohorts tend to have shorter birth intervals, father’s born to recent cohorts tend to have longer intervals. This could reflect some 69 differences in male versus female demand for children, but at this time we can only speculate.

Conclusion

The initial hypotheses stated in the introductory portion of this paper suggested that we should see support for the economic demand for children we see none when birth intervals are used as the measure of fertility. This could point to some balancing mechanism practiced by farm households to manage production and consumption in the Orkney archipelago. We also see some suggestive effects of post reproductive kin on birth intervals. While it appears that maternal grandparents have a significant impact on lengthening the first birth interval, the effect disappears for higher order births. This is suggestive of the relative independence of households in northern Orkney, where the household is the locus of ultimate fertility decision making and childbearing, with little to no direct impact of kin on timing of fertility.

If anything, the effects of maternal grandparents could be attributable to the nuclear family having to balance their resources between having a child and taking care of aging parents, thus delaying the first birth. Finally, the results of this analysis indicate that, although there were most likely no modern methods of contraception during the 19 th and early 20 th centuries in northern Orkney, families could have been limiting their family size, possibly by prolonged periods of post-partum abstinence, periods of spousal separation, or prolonged breastfeeding.

References

Anderson M, and Morse DJ (1993a) High fertility, high emigration, low nuptiality: adjustment processes in Scotland's demographic experience, 1861-1914, Part I. Popul Stud (Camb) 47: 5-25. 70

Anderson M, and Morse DJ (1993b) High fertility, high emigration, low nuptiality: adjustment processes in Scotland's demographic experience, 1861-1914, Part II. Popul Stud (Camb) 47: 319-43.

Becker GS (1981) A treatise on the family. Cambridge: Harvard University Press.

Bentley GR, Goldberg T, and Jasienska G (1993) The Fertility of Agricultural and Non-Agricultural Traditional Societies. Population Studies 47: 269-281.

Boserup E (1987) Population and technology in preindustrial Europe. Popul Dev Rev 13: 691-701, 764, 76.

Bowers EJ (1983) Patterns of adult mortality in the Orkney Islands. Ph. D. dissertation, University of Pennsylvania, Philadelphia.

Brennan ER (1983) Pre-reproductive mortality and family structure: Sanday, Orkney Islands 1855-1974. Hum Biol 55: 19-33.

Brennan ER, Leslie PW, and Dyke B (1982) Mate choice and genetic structure Sanday, Orkney Islands, Scotland. Hum Biol 54: 477-89.

Brennan ER, and Relethford JH (1983) Temporal variation in the mating structure of Sanday, Orkney Islands. Ann Hum Biol 10: 265-80.

Clay DC, and Johnson NE (1992) Size of farm or size of family: Which comes first? Population Studies 46: 491-505.

Cox DR (1972) Regression models and life tables (with discussion). Journal of the Royal Statistical Society B34: 187-220.

Firebaugh G (1982) Population density and fertility in 22 Indian villages. Demography 19: 481-94.

Flinn M, ed. (1977) Scottish population history from the 17th century to the 1930s. Cambridge: Cambridge University Press. 71

Galloway PR, Hammel EA, and Lee RD (1994) Fertility Decline in Prussia, 1875- 1910: A Pooled Cross-Section Time Series Analysis. Population Studies 48: 135-158.

Johnson PL (1990) Changing household composition, labor patterns, and fertility in a highland New Guinea population. Hum Ecol 18: 403-16.

Knodel J (1987) Starting, stopping, and spacing during the early stages of fertility transition: the experience of German village populations in the 18th and 19th centuries. Demography 24: 143-62.

Knodel JE (1988) Demographic behavior in the past: A study of fourteen German village populations in the eighteenth and nineteenth centuries. Cambridge: Cambridge University Press.

Larsen U and Vaulpel JW (1993) Hutterite fecundability by age and parity: strategies for frailty modeling of event histories. Demography 30: 81-102.

Leridon H (1977) Human Fertility. Chicago: University of Chicago Press.

Mace R, and Sear R (1997) Birth interval and the sex of children in a traditional African population: an evolutionary analysis. J Biosoc Sci 29: 499-507.

Markle GE, and Pasco S (1977) Family limitation among the Old Order Amish. Population Studies 31: 267-280.

Palloni A, and Rafalimanana H (1999) The effects of infant mortality on fertility revisited: new evidence from Latin America. Demography 36: 41-58.

Roberts DF, and Roberts MJ (1983) Surnames and relationships: an Orkney study. Hum Biol 55: 341-7.

Rosenzweig MR (1977) The demand for children in farm households. The Journal of Political Economy 85: 123-146.

Sanderson SK, and Dubrow J (2000) Fertility decline in the modern world and in the original demographic transition: testing three theories with cross-national data. Popul Environ 21: 511-37. 72

Schutjer WA, Stokes CS, and Poindexter JR (1983) Farm size, land ownership, and fertility in rural Egypt. Land Economics 59: 393-403.

Sear R, Mace R, and McGregor IA (2000) Maternal grandmothers improve nutritional status and survival of children in rural Gambia. Proc Biol Sci 267: 1641-7.

Sear R, Shanley D, McGregor IA, and Mace R (2001) The fitness of twin mothers: evidence from rural Gambia. Journal of Evolutionary Biology 14: 433-443.

Sear R, Steele F, McGregor IA, and Mace R (2002) The effects of kin on child mortality in rural Gambia. Demography 39: 43-63.

Singh S (1979) Demographic variables and the recent trend in fertility in Guyana, 1960-1971. Population Studies 33: 313-327.

Strassmann BI, and Gillespie B (2002) Life-history theory, fertility and reproductive success in humans. Proc Biol Sci 269: 553-62.

Tolnay SE, and Glynn PJ (1994) The persistence of high fertility in the American South on the eve of the baby boom. Demography 31: 615-31.

Tymicki K (2004) Kin influence on female reproductive behavior: The evidence from the reconstitution of the Bejsce parish registers, 18th to 20th centuries, Poland. American Journal of Human Biology 16: 508-522.

Vanlandingham M, and Hirschman C (2001) Population Pressure and Fertility in Pre- Transition Thailand. Population Studies 55: 233-248.

Voland E (1998) Evolutionary ecology of human reproduction. Annual Review of Anthropology 27: 347-374.

Voland E, and Beise J (2002) Opposite effects of maternal and paternal grandmothers on infant survival in historical Krummhörn. Behavioral Ecology and Sociobiology 52: 435-443.

Wood JW (1990) Fertility in anthropological populations. Ann Rev Anthropol 19: 211-242. 73

Wood JW (1994) Dynamics of human reproduction: Biology, biometry, demography. New York: Aldine De Gruyter.

Wood JW, Holman DJ, Weiss KM, Buchanan AV, and Lefor B (1992) Hazards models for human population biology. Yearbook of Physical Anthropology 35: 43-87.

Wood JW, Holman DJ, Yashin AI, Peterson RJ, Weinstein M, and Chang MC (1994) A multistate model of fecundability and sterility. Demography 31: 403-426.

74

Chapter 4 The costs of large families: An analysis of infant and childhood mortality in the Northern Orkney Islands.

Introduction

Childbearing is a costly process, both to the individual mothers and offspring and to the household at large. Costs may be defined many ways; popular definitions are those corresponding to energetic costs (Bronson 1995; Hytten 1980; Ulijaszek 1995), economic costs (Cleland and Wilson 1987; Easterlin 1975; Vlassoff 1982), and opportunity costs (Cramer 1979; Geronimus and Korenman 1993; Hoffman et al.

1993). For families to minimize the costs of children from an evolutionary perspective, they must optimize their reproductive efforts. To do this they must balance the survival of their current children and themselves with costs of new children; this represents the generalized life history problem in evolutionary biology

(Stearns 1992) and has been reviewed extensively in the human context (Voland

1998). But, as a practical matter, just how do human families balance these needs?

Understanding how this balancing act is performed, and indeed if it is consciously performed at all, could allow great insight into the dynamics of households, families and populations. More importantly, the impacts of the balance should have readily noticeable effects on infant and childhood mortality, since this is the end result of the balancing act. The purpose of this chapter is to examine the products of this balancing act on infant and childhood mortality in the Northern Islands of Orkney,

Scotland. Specifically I will investigate the influences of family size, birth order and household composition on infant and childhood mortality. I begin by outlining the dynamics of households from an economic perspective. I will then discuss possible impacts of the household formation and growth process on infant mortality and child 75 quality. I will also consider the problem of intra household competition and the constraints of parental choice on child well being.

Household growth and formation dynamics

The principles of household formation are relatively simple, although they may take a variety of forms and abide by different sets of formation rules (Burch and Matthews

1987; Goody 1996; Gunnlaugsson 1988; Hajnal 1982; Hammel 2005a, b, c; Hammel and Laslett 1974; Laslett 1969; Smith 1981). Traditionally, at least for the western

European household, a household begins with a marriage between a man and a woman; they then leave their respective parents’ homes and form a new household, generally assumed to be in a different physical structure from where their natal families live. Although this example has numerous exceptions, and this western concept of household is definitely becoming antiquated in various settings (Axinn and

Thornton 1996; Bauman 1999; Bumpass and Raley 1995; Manning and Smock 1995), for our purposes here we are speaking in terms of models and gross oversimplifications (e.g. Hammel 2005a). Once this new household is formed, it begins producing children at a given rate, generally assumed - at least in the household economics literature (Chayanov, 1966; Hammel, 2005a) - to be every 3 years. This process continues until a large number of offspring is produced: Hammel

(2005a) and Chayanov (1966) each assume 9 children. This resulting process of age- specific childbearing, child maturation and household growth can be summarized by a set of numeric relationships, first used by Chayanov (1966) and later formally derived by Durrenberger (1984), that equate the stage of the household growth cycle to the 76 relative productivity and consumption patterns of the household. The pattern of consumers relative to producers over the life cycle of the household takes the general form presented in Figure 4.1.

Figure 4.1 Consumer/Worker ratios under the basic model of household formation (modified from Chayanov 1966).

8 2.5

7

2 6

5 1.5

4 C/W C/W Ratio

1 3 # Consumers or Workers Consumers #

2 Consumers 0.5 Workers 1 C/W Ratio

0 0 0 5 10 15 20 25 Years of Family's Existence

The ratio of consumers to workers was central in Chayanov’s work and his aim was to show that a peasant household should work no more that it needs in order to support its consumers. Ideally, the household would come to a balancing point whereby the needs of its consumers are balanced by the production of its workers.

This model of household dynamics seems simple because it assumes that everything external to the household is equal among households, including the total amount of land and equality of access to land. It also assumes that all households were also homogeneous in their ultimate desires, an assumption that even in 1920’s Russia was 77 unlikely to be true. Recently Hammel (Hammel 2005a, b, c) revitalized the ideas in

Chayanov’s work and used simulations to provide more extended cases of multiple generation households, competition within households and availability of kin. Most important for the current analysis is Hammel’s (2005a) argument for competition within households. In his discussion of the inequalities in complex households he brings up the point of different aged siblings having very different household economic scenarios. He states:

“For example, one brother may have many children who are immature, whereas another brother may have only a few children, but approaching maturity.” (Hammel, 2005a, p. 7045)

He goes on to say:

“In a sense, the complex household, although offering advantages over the nuclear, contains the seeds of its own destruction. Other issues may also stimulate fission, such as crowded housing conditions or insufficiency of exploitable resources such as farmland or pasture.” (Hammel, 2005a, p.7046)

The statement “seeds of its own destruction” is rather poignant, especially if we are interested in the negative impacts of such overcrowded, competitive environments on reproductive success and mortality.

If households were to balance their current state, economic or otherwise, in order to maximize their present levels of consumption and production, what then would be the impact on the household of the addition of another mouth to feed?

Additionally what if an even more disadvantageous scenario were to occur: twins? In many settings the additional stress of twin births often leads to cases of infanticide of one of the infants (Ball and Hill 1996; Cronk 1989; Scrimshaw 1978). If the household was behaving as life history theory (Lack, 1947; Stearns, 1992) and

Chayanov (1966) say it should, what would be the impacts on the current and future well being of children, the most vulnerable individuals in the household? 78

In agrarian communities, children may represent both a positive (Becker and

Lewis 1973; Kaplan 1994; Kramer 2005a, b; Kramer and Boone 2002; Lee and

Kramer 2002) and negative (Ho 1979; Lee 1986; Muhuri 1995; Muhuri and Preston

1991; Tiefenthaler 1997) impact on the household time and energy budget. If children can represent both these aspects, having a certain number of children may be beneficial to the family, and while births beyond this number may create such an economic or energetic deficit that these addition may have a serious impact on the other children in the family (Hagen et al. 2006; Pebley and Stupp 1987; Winikoff

1983). In addition to this effect, if other relatives are living in the household, the resources of the family could be stretched even further. Lastly, if the couple is practicing sex-based fertility decisions, these could lead to an impact on child mortality because the could invest differentially in male verses female children

(Casterline et al. 1989; Desai 1992; Gupta 1997). In the analysis that follows I will examine the influences of family size and household composition on infant and childhood mortality, focusing on the impact of the household consumer/producer ratio and the effect of being a high order birth on two measures of child mortality. I will also test for the rare, but most certainly disadvantageous situations of twin births.

Data and methods

The North Orkney Population History Project (NOPH) has entered into a database all births, marriages and deaths registered in the Northern Orkney Islands between 1855 and 2001. The resulting database included 7410 births, 2025 marriages and 5141 deaths for the islands of Eday (and Pharay), North Ronaldsay, Papa 79

Westray and Westray. Data for the island of Sanday were unavailable at the time of this analysis, but will be incorporated in later work. All marriages that occurred on these islands between 1855 and 2003 were linked to birth records for all children produced from the marriages that were likewise born on one of the Northern Isles. At the present time, marriages that occurred at any place other than the Northern Isles have been excluded from the analysis since no appropriate marriage records have been entered into the database. Likewise, if a child was born at any place other than the Northern Isles, they are not included in this analysis since no birth record exists as of the time of the analysis. The linkages were generated by using father’s first and surname, mother’s first and maiden name, and child’s year of birth as linking criteria.

This generated a total of 4583 births that can be linked to marriages that occurred on the Northern Isles between 1855 and 2001, and 1992 right censored open birth intervals. These birth records were then linked to death records using the same criteria as above and child’s name and year of birth, producing a total of 1313 complete records for analysis. Although many other death records exist in the database, if either the parents’ marriage record or the child’s birth record was missing, the death was not included in the analysis. Because many people emigrated from the islands during this time period, the linkage rate between births and deaths is low; in addition linkages between islands have not yet been completed. This presents a particularly large deficiency in the data, because the movement of people between islands in the Orkney archipelago is very common, as is spending portions of ones life on several islands. By performing these linkages more records will certainly be obtained for future work along these lines. While these inter-island linkages should 80 increase the number of complete observations for analysis, many more will still be incomplete because many families moved to other parts of Britain, North America or

Australia. These individuals are effectively lost from any analysis.

This analysis focuses on two main outcomes. First I investigate the influence of individual and family variables on a child’s age at death before age 1, and secondly

I estimate the effects on a child’s age at death before age 15. Age 1 is used because it represents surviving beyond the high period of infant mortality, and age 15 is used because it represents the approximate age of sexual maturity.

Variables considered in this analysis represent both household and individual level factors that could influence child survivorship. Household variables include farm household status, age in years of mother and father at time of child’s birth, number of grandparents alive at the time of a child’s birth, and a proxy calculation of the household consumer/producer ratio. Farm household status is 1 if the child’s father’s occupation was listed as “farmer” or “crofter” in the child’s death record, otherwise the code is 0.5 In the Orkney setting this distinction is a very tenuous one, as almost everyone living in the islands participates in agriculture at some time of the year, and likewise many of the farmers participate in activities other than farming.

Another distinction that should be used instead of farm/nonfarm is landowner verses renter verses crofter, but this distinction cannot be made based on the information available in the vital registration records. The farm/nonfarm distinction is used here as a proxy measure of land renters or owner compared to everyone else. Number of

5 While most of the population of Orkney was probably participating in agriculture as part of their daily lives, I identify farmers here are those whose occupations were listed as such. They are presumably less likely to pursue other forms of work than fathers listed as agricultural laborer or fisherman. The farm household category is intended to represent those families who were participating solely in agriculture. 81 grandparents alive does not indicate that the child’s grandparents were living in the household with the child, only that they were alive at the time of the child’s birth.

Previous work (Voland and Beise 2002) has shown some evidence of residential location of grandparents on grandchild survival, but at present I am unable to control for this locality effect. The household consumer/producer ratio was estimated as follows:

((#of live parents+ #of live siblings)/ #of live parents).

This is not the ideal specification for this ratio because it ignores the age and sex of siblings (c.f. Hammel, 2005a; Kramer, 2005b) and the relative contribution to consumption and production of children of varying ages and sexes, although the basic assumption I believe is a valid one. Higher ratios indicate large numbers of surviving offspring per parent in the household at the time of the index child’s birth, so this ratio should still represent a proxy for how economically stressed the parents were prior to the index child’s birth. In future work the vital registration data should be linked to census information on precise household composition, and be able to incorporate more refined measures of household production and consumption.

Individual level variables included are child’s sex (male=1, female=0), child’s birth order (0 to 3, 4 to 6, 7+), child’s twin status (twins=1, singleton=0), and child’s birth cohort (measured as the decade the child was born). I also include the length of the child’s inter-birth interval (measured in months since previous birth) for second and higher birth orders. In other settings (Brittain 1992; Curtis et al. 1993; Lehrer 1984;

Miller et al. 1992; Palloni and Tienda 1986) the length of the inter-birth interval has 82 displayed a strong influence on child mortality. Descriptive statistics for all variables included in the analytical models are presented in Table 4.1.

Table 4.1 Descriptive statistics for household and individual level variables used in regression models. Mean or Percent Std Min Max N Family Variables Farm/Nonfarm, Farm=1 .46 1313 Mother's age at child's birth 30.35 6.51 16 49 1313 Father's age at child's birth 33.47 8.16 17 66 1313 # Grandparents Alive 3.22 0.84 0 4 1289 Household C/P ratio 2.13 1.23 0.5 8 1313 Individual Variables Sex, Male=1 .51 1313 Child's birth order 1313 0 to 3 .64 4 to 6 .24 7+ .12 Twin status, Twin=1 .04 0.18 1313 Inter-birth interval (in months) 28.48 23.09 -72 186 1313

% censored at age 15 74.02 % censored at age 1 79.05 Island North Ronaldsay .17 Papa Westray .07 Westray .65 Eday .11

To test for effects of family and individual level variables on child’s age at death I use nested Cox proportional hazard models (Cox 1972). The proportional hazard model estimates the effect of observed covariate patterns on the risk of an individual experiencing some event, in this case death, relative to individuals with different patterns of covariates. The model is generally written:

p

hi (t) = λ0 (t)exp( ∑βk zk ) k =1 with hi(t) being the hazard of experiencing the event at time t, λ0(t) being a baseline

hazard rate that is left unspecified (this feature distinguishes the proportional hazard 83 model from fully parametric hazard models (Wood et al. 1992) and is generally represented by an individual having 0’s for all covariate values, β being the estimated

regression coefficients and z being the individual covariate values. This model produces an estimate of the individual’s hazard at time t, given the baseline risk and the linear effect of the covariates. The attraction of the Cox model is that the researcher does not have to be forced into specifying a parametric form for the baseline hazard and the ease of interpretation of the regression coefficients.

I estimate the effects of the covariates described above on the risk of dying before age 1 and before age 15. I estimate three models per analysis. The first model estimates effects of family level variables on the child’s age at death. Secondly I estimate only characteristics of the child on their age at death, and finally I estimate a full model incorporating both the family and child level effects. The Cox regressions were performed using the phreg procedure in SAS v.9.1 (SAS 2003) and the

“survival” routines in R 2.5.1 (CRAN 2007; Lumley 2007).

Results

I will discuss the results of the two analyses separately, beginning with the

Cox regression analysis of infant mortality. I will first discuss the results of each model then summarize the general findings.

Table 4.2 presents the estimated regression coefficients and model fit criteria for the three nested models. When considering only family variables (Model 1), farm household status, mother’s age at child’s birth and household C/P ratio have significant effects. Children who are born into farm households have a 39% decreased chance of dying before age 1 (hazard ratio=.609, 95%CI .403 - .921). 84

Children born to older mothers have a slight (4.5%), but significant decrease in risk

(hazard ratio=.955, 95% CI .918 - .913). As an example of the maternal age effect, a woman age 40 has 39.6% as much risk of losing a child before age 1 compared to a woman aged 20. The total number of grandparents alive at the time of child’s birth has a positive, but insignificant impact on reaching age 1. The consumer/producer ratio has a negative impact on reaching age 1, raising the risk by 29.8% (hazard ratio=1.298, 95% CI 1.089 – 1.547). As an example if a child was born into a household that already had 3 children (C/P ratio = 2.5), versus being the first birth in the household (C/P ratio=1), that child would be 1.48 times more likely to die before age 1. With respect to model fit, the R 2 for Model 1 was only 1.5%, this indicates

that the factors chosen in the first model do a poor job of explaining the variability in

infant age at death and other factors beside those of the household the child was born

should be considered in the analysis.

Table 4.2 Results of Cox regression analysis of infant mortality. Model 1 (s.e.) Model 2 (s.e.) Model 3 (s.e.) Family Variables Farm/Nonfarm, Farm=1 -.496(.211) -.492(.213) Mother’s age at child’s birth -.046(.020) -.035(.021) Father's age at child's birth .026(.013) .031(.014) # Grandparents Alive -.194(.111) -.206(.112) Household C/P ratio .261(.089) -.188(.215) Individual Variables Sex, Male=1 .181(.193) .174(.195) Child's birth order .413(.123) .789(.341) Twin status, Twin=1 .963(.377) .938(.383) Inter-birth interval -.011(.005) -.009(.004) Child birth cohort .003(.004 ) .006(.004)

n 1182 1201 1182 Model -2LL 1251.31 1251.85 1233.852 x2 (Wald) 18.96 24.22 37.25 df, p 5, .002 5,.0002 10,<.0001 Model R 2 .015 .017 .028

85

Model 2 indicates a significant increase in risk to children of high birth order (51% increase in risk, hazard ratio 1.512, 95% CI 1.188-1.924). An example of this is, if a child was of parity 4 to 6, they would have a 1.51 higher risk of dying before age 1 than a child of birth order parity 1 to 3. Twin births were also 2.62 times more likely to die before reaching age 1 than singletons (hazard ratio 2.620, 95% CI 1.252-5.481).

There appears to be no effect of child’s sex on mortality before age 1. Lastly the length of the inter-birth interval indicates a decrease in risk by 1.1% (hazard ratio

.989, 95% CI .980 - .998). To illustrate the effect of inter-birth interval, if a child had an interval of 24 months, they would have a 16% lower risk of dying before age 1 than a child with an interval of 9 months. Model fit for the second model is also poor, as shown in the model R 2 of .017. No likelihood ratio tests were performed on Model

2 versus Model 1 because they have different sample sizes, and different null model likelihoods.

Model 3, the full model, controls for both the family and child level effects, and aside from a few changes the estimated effects agree with the previous models.

Factors that decrease the child’s risk of death are being born into a farm household

(39% decrease in risk, hazard ratio 0.612, 95% CI .402-.930), and having a longer inter-birth interval (1% decrease in risk, hazard ratio .990, 95% CI .980 – 1.000). To illustrate the effect of inter-birth interval, if a child had an interval of 24 months, they would have a 14% lower risk of dying before age 1 than a child with an interval of 9 months. Factors increasing the risk of a child dying before age 1 were father’s age at child’s birth (a 3.2% increase in risk, hazard ratio 1.032, 95% CI 1.004-1.060). This effect translates into a 2.18 times greater risk of infant death for a father aged 50 86 compared to a father aged 25. Child’s birth order was also a significant factor increasing the hazard rate (higher birth orders were 2.2 times more likely to die before age 1 than lower birth orders, hazard ratio 2.202 95% CI 1.129-4.298), and twin status also increased the hazard rate (twin births were 2.56 times more likely to die before reaching age 1, hazard ratio 2.556, 95% CI 1.206-5.417). It should be noted that the effect of the household consumer/producer ratio becomes insignificant and changes sign after controlling for individual level covariates in Model 3. While lacking statistical significance, speculation on this change in effect is necessary.

After controlling for index child’s birth order, the negative sign of the consumer/producer ratio could reflect the potential benefits of older children in the household on the index child’s risk of death (Kramer 2005a, b; Lee and Kramer

2002). Without a better specification of the household C/P ratio, this effect (while supported by findings from other settings) remains speculative in this setting. While each of the three models represents statistically significant attempts to understand the effects of these covariates on infant mortality, the amount of variation explained by them is rather poor. The full model only accounted for 2.8% of the variation in infant mortality, with Models 1 and 2 performing worse (1.5 and 1.7% respectively). A likelihood ratio test of Model 3 versus Model 1 indicates a significant increase in model fit ( χ2=34.9, df=5, p=<.0001). Based on this poor performance of these

models it is obvious that other factors need to be controlled for in further analyses of

these data. Another possibility is that the Cox regression model used in the analysis

does not represent the ideal specification for the analysis of infant mortality, and

perhaps a parametric model should be pursued. 87

I now turn to the analysis of total childhood mortality and the analysis of risk factors for child death up to age 15. Table 4.3 presents the Cox regression coefficients and model fit statistics for the three nested models estimated in this analysis. As in the analysis of infant mortality, Model 1 estimates the effects of family variables, Model 2 estimates effects of child’s variables, and Model 3 is a full model considering both sets. When we examine model fit, Model 1 has an R 2 value of only .8%, showing a very poor performance in explaining the variability in age at death.

Results of Model 1 indicate a significant 33% reduction in risk for children born into farm families (hazard ratio .667, 95% CI .506-.880). The only significant negative effect is the household consumer producer ratio, showing a nearly 22% increase in risk of dying before age 15 (hazard ratio 1.217, 95% CI 1.076-1.377). As an example, if a child was born into a household that already had 3 children (C/P ratio

= 2.5), versus being the first birth in the household (C/P ratio=1), that child would be

1.35 times more likely to die before age 15. All other effects were insignificant in the model.

Model 2 indicates that children of higher birth order have a nearly 45% higher risk of dying before age 15 (hazard ratio 1.408, 95% CI 1.193-1.661). An example of this is, if a child was of parity 4 to 6, they would have a 1.46 higher risk of dying before age 15 than a child of birth order parity 1 to 3. Inter-birth interval has a negative impact on the hazard, decreasing it by .7% (hazard ratio .993, 95%CI .987 -

.999). To illustrate the effect of inter-birth interval, if a child had an interval of 24 88 months, they would have a 10% lower risk of dying before age 15 than a child with an interval of 9 months.

Model 3 maintains the advantage of farm households, indicating a 33% decrease in risk (hazard ratio .673, 95% CI .508 - .892) of death, while child’s birth order had an 75% increase in risk (hazard ratio 1.748, 95% CI 1.112 – 2.747).

Similar to Model 2, longer inter-birth intervals tend to have a reduced risk of death of

.8% (hazard ratio .992, 95% CI .986 - .999). To illustrate the effect of inter-birth interval, if a child had an interval of 24 months, they would have a 12% lower risk of dying before age 15 than a child with an interval of 9 months.

Table 4.3 Results of Cox regression analysis of children’s age at death. Model 1 (s.e.) Model 2 (s.e.) Model 3 (s.e.) Family Variables Farm/Nonfarm, Farm=1 -.401(.141) -.395(.143) Mother’s age at child’s birth -.023(.014) -.013(.015) Father's age at child's birth .016(.010) .018(.010) # Grandparents Alive -.055(.082) -.052(.081) Household C/P ratio .199(.063) -.111(.140) Individual Variables Sex, Male=1 -.043(.132) -.037(.133) Child's birth order .375(.086) .558(.231) Twin status, Twin=1 .558(.325 ) .529(.330) Inter-birth interval -.006(.003) -.008(.003) Child birth cohort -.001(.003) .000(.003)

n 1182 1201 1182 Model -2LL 2605.57 2624.96 2590.95 x2 (Wald) 22.06 25.41 36.92 df, p 5, .0005 5, .0001 10,<.0001 Model R 2 .008 .008 .013

Again we see a poorly fitting model with only 1.3% of the variation explained in

Model 3. A likelihood ratio test of Model 3 versus Model 1 indicates a significant increase in model fit ( χ2=29.24, df=5, p=<.0001), although this increase translates

into a relatively small amount of increased information on the mortality process.

89

The results of the two analyses described above indicate several significant sources of variation in child survivorship in the Northern Orkney Islands. The results of infant survival indicate more familial and individual factors play a determining role in survivorship than the analysis of survival to age 15. Infant survivorship tends to decrease for higher birth order children in nonfarm households with older fathers, with this statement being especially true for twin births. The effects of the household consumer/producer ratio, while appearing to be a significant detriment to child survival when only family variables are considered, disappear in the full model once individual level factors are accounted for. As in the first analysis the effect of the consumer/producer ratio becomes insignificant and changes signs. This might represent the impact of older siblings reducing the index child’s risk of death. The effect of the length of the inter-birth interval also plays an important role, whereby children who were born further apart tend to have better survival chances.

To illustrate the effects of birth order and farm household status on infant mortality, survivor function estimates were calculated from the full (Model 3) Cox

Regression models and are given in Figures 4.2 and 4.3

90

Figure 4.2 Farm and nonfarm child survivorship to age 1 (Model 3). Note: Survivorship is truncated at .7 for visual purposes.

Figure 4.3 Child survivorship to age 1 for varying birth orders (Model 3). Note: Survivorship is truncated at .7 for visual purposes.

91

The analysis of child’s ages at death up to age 15 revealed similar effects to those found in the first analysis. The dominant effects here are farm household status and birth order, with childhood survivorship tending to be lower in nonfarm families and for children of higher birth orders. These effects are again presented as survivorship curves derived from the full (Model 3) Cox regression models.

Figure 4.4 Farm and nonfarm child survivorship to age 15 (Model 3).

92

Figure 4.5 Child survivorship to age 15 for varying birth orders (Model 3).

Discussion

The general findings of this analysis indicate definite impacts of birth order

and farm household status on infant and childhood mortality. Not surprisingly, twin

births have dramatically higher risks of early death compared to singleton births.

This probably shows the negative impact of higher energetic demands of twins in

utero, leading to either premature or low birth weight infants, or the higher economic

costs of raising two children of the same age instead of one. This effect is also

present in families with a large number of total children, as reflected by the effects of

child’s birth order. The observed pattern indicates that children in large families tend 93 to be at higher risk than those born into smaller families. While the immediate stress of being a twin birth has both proximate biological and social causes, most notably maternal depletion (Miller and Huss-Ashmore 1989; Scott and Duncan 2000; Wood

1994), prematurity and close birth intervals (Miller 1991), the stress of being a high order birth is probably related to the limitations on total household productivity.

While there are currently no data on individual household productivity in this setting, we can suggest that large families tend to face a harder time providing for all of their children based on other data from other cases (Hagen et al. 2006; Winikoff 1983).

The testing of the actual causality in this argument would be an interesting future analysis if appropriate data exist. Interestingly, when we take account of the index child’s birth order the negative impact of the household consumer/producer ratio disappears. This should not be surprising because both variables are part of the same phenomena, this concept that keeps reoccurring: the negative impact of large families on child health. While this finding suggests a relative unimportance of household consumer/producer ratios if we include the interaction of child’s birth order and household consumer/producer ratio, the effect of birth order likewise disappears. We should also consider the definition of the C/P ratio used in this setting. By ignoring the age and sex of living siblings, they are given no weight at potential producers in the household, a fact that needs to be controlled for by linking the vital registration data to information from the censuses on exact household composition. The importance of the interaction between birth order and C/P ratio suggests some moderating effect of older siblings on the well being of the index child, most likely attributable to the added household productivity of the older children or the additional 94 care provided to the index child (Sear et al. 2002). By adding more detail about the

C/P ratio, however, we should be able to understand these interactions better.

In both analyses farm households appear to have reduced risks of child death relative to nonfarm households. This suggests that there may be some buffering mechanism present in farm families that may offset the effects of disadvantageous family and individual characteristics. From an analysis of birth spacing (See Chapter

3), farm families appear to have longer inter-birth intervals than nonfarm households.

This could indicate that 1) farm families could be intentionally spacing their births farther apart than nonfarm families in order to most efficiently balance their consumption/production ratios, and 2) farm families could be buffered against disadvantageous household cycle swings by having access to one of the most valuable commodities in Orkney: land. Similar results indicating the beneficial effects of farm life on infant and child mortality were found by Watterson (Watterson 1988) using life tables for various occupational groups in England and Wales in the early 20 th century. The problem with this interpretation however is that we are assuming all

“farmers” are the same qualitatively, when this is just not the case. For the most part, people did not own the land they worked, and indeed most small farmers (crofters) primarily were employed on nearby larger farms as agricultural laborers. Future work should attempt to distinguish the various types of “farmers” and examine the variability between them, because there should be a great difference between an owner of 5000 acres of land and a crofter of 2 acres who works for the land owner. 95

Other studies have also reported the negative effects of older fathers (Farah and Preston 1982; Gupta 1997; Trussell and Hammerslough 1983) and the positive effects of older mothers (Geronimus 1987) on infant mortality.

Conclusion

In conclusion, this analysis indicates a significant cost of large family size on infant and childhood mortality as measured by the probability of surviving to age 1 and to age 15 in Northern Orkney. Also evident is the very costly effect of a twin birth. This negative impact can however be offset if a child is born into a farm household, as these could be buffered against other conditions associated with poor infant and childhood health.

References

Anderson M (1998) Fertility decline in Scotland, England and Wales, and Ireland: comparisons from the 1911 census of fertility. Popul Stud (Camb) 52: 1-20.

Anderson M, and Morse DJ (1993a) High fertility, high emigration, low nuptiality: adjustment processes in Scotland's demographic experience, 1861-1914, Part I. Popul Stud (Camb) 47: 5-25.

Anderson M, and Morse DJ (1993b) High fertility, high emigration, low nuptiality: adjustment processes in Scotland's demographic experience, 1861-1914, Part II. Popul Stud (Camb) 47: 319-43.

Axinn WG, and Thornton A (1996) The influence of parents' martial dissolutions on children's attitudes toward family formation. Demography 33: 66-81.

Ball H, and Hill CM (1996) Reevaluating "Twin Infanticide". Current Anthropology 37: 865-863.

Barclay RS (1965) The population of Orkney 1755-1961. Kirkwall: Mackintosh. 96

Bauman KJ (1999) Shifting family definitions: the effect of cohabitation and other nonfamily household relationships on measures of poverty. Demography 36: 315-25.

Becker GS (1981) A treatise on the family. Cambridge: Harvard University Press.

Becker GS, and Lewis HG (1973) Interaction between quantity and quality of children. In TW Schultz (ed.): Economics of the family: marriage, children, and human capital. Chicago: University of Chicago Press, pp. 81-90.

Beckman LJ (1984) Husbands' and wives' relative influence on fertility decisions and outcomes. Popul Environ 7: 182-97.

Bentley GR, Goldberg T, and Jasienska G (1993) The Fertility of Agricultural and Non-Agricultural Traditional Societies. Population Studies 47: 269-281.

Berry RJ (1985) The natural history of Orkney. London: Collins and Son.

Bock J (2002) Introduction: evolutionary theory and the search for a unified theory of fertility. Am J Human Biol 14: 145-8.

Bongaarts J (1978) A framework for analyzing the proximate determinants of fertility. Popul Dev Rev 4: 105-32.

Boserup E (1987) Population and technology in preindustrial Europe. Popul Dev Rev 13: 691-701, 764, 76.

Bourgeois-pichat J (1981) Recent demographic change in western Europe: an assessment. Popul Dev Rev 7: 19-42.

Bowers EJ (1983) Patterns of adult mortality in the Orkney Islands. Ph. D. dissertation, University of Pennsylvania, Philadelphia.

Brennan E (1979) Kinship, demographic, social, and geographic characteristics of mate choice in a small human population. Ph. D. dissertation, Pennsylvania State University, University Park. 97

Brennan ER (1983a) Mortality patterns in anthropological populations. Hum Biol 55: 1-7.

Brennan ER (1983b) Pre-reproductive mortality and family structure: Sanday, Orkney Islands 1855-1974. Hum Biol 55: 19-33.

Brennan ER, Leslie PW, and Dyke B (1982) Mate choice and genetic structure Sanday, Orkney Islands, Scotland. Hum Biol 54: 477-89.

Brennan ER, and Relethford JH (1983) Temporal variation in the mating structure of Sanday, Orkney Islands. Ann Hum Biol 10: 265-80.

Bronson FH (1995) Seasonal Variation in Human Reproduction: Environmental Factors. The Quarterly Review of Biology 70: 141-164.

Bumpass LL, and Raley RK (1995) Redefining single-parent families: cohabitation and changing family reality. Demography 32: 97-109.

Burch TK, and Matthews BJ (1987) Household formation in developed societies. Popul Dev Rev 13: 459-511, 570, 57.

Casterline JB, Cooksey EC, and Ismail AFE (1989) Household Income and Child Survival in Egypt. Demography 26: 15-35.

Clay DC, and Johnson NE (1992) Size of farm or size of family: Which comes first? Population Studies 46: 491-505.

Cleland J, and Wilson C (1987) Demand Theories of the Fertility Transition: an Iconoclastic View. Population Studies 41: 5-30.

Coale A, and Watkins SC, eds. (1986) The decline of fertility in Europe. Princeton: Princeton University Press.

Cox DR (1972) Regression models and life tables (with discussion). Journal of the Royal Statistical Society B34: 187-220.

Cramer JC (1979) Employment trends of young mothers and the opportunity cost of babies in the United States. Demography 16: 177-97. 98

CRAN (2007) R 2.4.1.

Cronk L (1989) Low Socioeconomic Status and Female-Biased Parental Investment: The Mukogodo Example. American Anthropologist 91: 414-429.

Davidson DA, and Simpson IA (1994) Soils and landscape history: case studies from the Northern Isles of Scotland. In A Foster and TC Smout (eds.): The history of soils and field systems. Aberdeen: Scottish Cultural Press.

Desai S (1992) Children at risk: the role of family structure in Latin America and West Africa. Popul Dev Rev 18: 689-717.

Dodgshon RA (1994) Budgeting for survival: Nutrient flow and traditional highland farming. In S Foster and TC Smout (eds.): The history of soils and field systems. Aberdeen: Scottish Cultural Press.

Durrenberger EP (1984) Operationalizing Chayanov. In EP Durrenberger (ed.): Chayanov, peasants, and economic anthropology. Orlando: Academic Press.

Easterlin RA (1975) An Economic Framework for Fertility Analysis. Studies in Family Planning 6: 54-63.

Farah AA, and Preston SH (1982) Child mortality differentials in Sudan. Popul Dev Rev 8: 365-83.

Fenton A (1978) The Northern Isles: Orkney and Shetland. Edinburgh: John Donald.

Firebaugh G (1982) Population density and fertility in 22 Indian villages. Demography 19: 481-94.

Firth J (1974) Reminiscences of an Orkney parish. Stromness: Orkney Natural History Society.

Flinn M, ed. (1977) Scottish population history from the 17th century to the 1930s. Cambridge: Cambridge University Press.

Friedman D, Hechter M, and Kanazawa S (1994) A theory of the value of children. Demography 31: 375-401. 99

Gage AJ, Sommerfelt AE, and Piani AL (1997) Household structure and childhood immunization in Niger and Nigeria. Demography 34: 295-309.

Gage TB (1993) The decline of mortality in England and Wales 1861 to 1964: decomposition by cause of death and component of mortality. Popul Stud (Camb) 47: 47-66.

Galloway PR, Hammel EA, and Lee RD (1994) Fertility Decline in Prussia, 1875- 1910: A Pooled Cross-Section Time Series Analysis. Population Studies 48: 135-158.

Geronimus AT (1987) On teenage childbearing and neonatal mortality in the United States. Popul Dev Rev 13: 245-79.

Geronimus AT, and Korenman S (1993) The socioeconomic costs of teenage childbearing: evidence and interpretation. Demography 30: 281-90; discussion 291-6.

Gibson AJS, and Smout TC (1995) Prices, food, and wages in Scotland 1550-1780. Cambridge: Cambridge University Press.

Goody J (1996) Comparing family systems in Europe and Asia: are there different sets of rules? Popul Dev Rev 22: 1-20.

Guinnane TW, Okun BS, and Trussell J (1994) What do we know about the timing of fertility transitions in Europe? Demography 31: 1-20.

Gunnlaugsson GA (1988) Family and household in Iceland 1801-1930. Stockholm: Almqvist and Wiksell International.

Gupta MD (1997) Socio-Economic Status and Clustering of Child Deaths in Rural Punjab. Population Studies 51: 191-202.

Hagen EH, Barrett HC, and Price ME (2006) Do human parents face a quantity- quality tradeoff?: Evidence form a Shuar community. American Journal of Physical Anthopology 130: 405-418. 100

Hajnal J (1982) Two kinds of preindustrial household formation system. Popul Dev Rev 8: 449-94.

Hammel EA (2005a) Chayanov revisited: A model for the economics of complex kin units. Proceedings of The National Academy of Sciences, USA 102: 7043- 7046.

Hammel EA (2005b) Demographic dynamics and kinship in anthropological populations. Proceedings of The National Academy of Sciences, USA 102: 2248-2253.

Hammel EA (2005c) Kinship-based politics and the optimal size of kin groups. Proceedings of The National Academy of Sciences, USA 102: 11951-11956.

Hammel EA, and Laslett P (1974) Comparing Household Structure over Time and between Cultures. Comparative Studies in Society and History 16: 73-109.

Harrison GA (1976) Genetic and anthropological studies in the human adaptability section of the International Biological Programme. Philos Trans R Soc Lond B Biol Sci 274: 437-45.

HMD (2007) Human Mortality Database: University of California, Berkeley (USA) Max Planck Institute for Demographic Research (Germany).

Ho TJ (1979) Time costs of rearing children in the rural Philippines. Popul Dev Rev 5: 643-62.

Hoffman SD, Foster EM, and Furstenberg FF, Jr. (1993) Reevaluating the costs of teenage childbearing. Demography 30: 1-13.

Hytten FE (1980) Nutrition. In FE Hytten and g Chamberlain (eds.): Clinical Physiology in Obstetrics. Oxford: Blackwell Scientific, pp. 163-192.

Johnson PL (1990) Changing household composition, labor patterns, and fertility in a highland New Guinea population. Hum Ecol 18: 403-16.

Kaplan H (1994) Evolutionary and wealth flows theories of fertility: empirical tests and new models. Popul Dev Rev 20: 753-91. 101

Knodel J (1977) Family limitation and the fertility transition: Evidence form the age patterns of fertility in Europe and Asia. Population Studies 31: 219-249.

Knodel J (1987) Starting, stopping, and spacing during the early stages of fertility transition: the experience of German village populations in the 18th and 19th centuries. Demography 24: 143-62.

Knodel JE (1988) Demographic behavior in the past: A study of fourteen German village populations in the eighteenth and nineteenth centuries. Cambridge: Cambridge University Press.

Kramer KL (2005a) Children's help and the pace of reproduction: Cooperative breeding in humans. Evolutionary Anthropology 14: 224-237.

Kramer KL (2005b) Maya children: Helpers at the farm. Cambridge: Harvard University Press.

Kramer KL, and Boone JL (2002) Why intensive agriculturalists have higher fertility: A household energy budget approach. Current Anth 43: 511-517.

Laslett P (1969) Size and Structure of the Household in England Over Three Centuries. Population Studies 23: 199-223.

Laslett P (1970) The Comparative History of Household and Family. Journal of Social History 4: 75-87.

Laslett P, and Wall R, eds. (1972) Household and family in past time. Cambridge: Cambridge University Press.

Lee RD (1986) The value and allocation of time in high-income countries: implications for fertility. Comment. Popul Dev Rev 12: 108-10.

Lee RD, and Kramer KL (2002) Children's economics roles in the Maya family life cycle: Cain, Caldwell, and Chayanov revisited. Population and Development Review 28: 475-499.

Leridon H (1977) Human Fertility. Chicago: University of Chicago Press. 102

LeVine RA, LeVine SE, Richman A, Uribe FMT, Correa CS, and Miller PM (1991) Women's Schooling and Child Care in the Demographic Transition: A Mexican Case Study. Population and Development Review 17: 459-496.

Lotka AJ (1922) The Stability of the Normal Age Distribution. Proc Natl Acad Sci U S A 8: 339-45.

Lotka AJ (1989) Lotka on population study, ecology, and evolution. Popul Dev Rev 15: 539-50.

Lumley T (2007) The survival package, v 2.31.

Macbeth HM, and Boyce AJ (1987) Anthropometric variation between migrants and non-migrants: Orkney Islands, Scotland. Ann Hum Biol 14: 405-14.

Mace R, and Sear R (1997) Birth interval and the sex of children in a traditional African population: an evolutionary analysis. J Biosoc Sci 29: 499-507.

Manning WD, and Smock PJ (1995) Why marry? Race and the transition to marriage among cohabitors. Demography 32: 509-20.

Markle GE, and Pasco S (1977) Family limitation among the Old Order Amish. Population Studies 31: 267-280.

Marwick H (1952) Orkney Farm Names. Kirkwall: W. R. Mackintosh.

Mason KO (1987) The impact of women's social position on fertility in developing countries. Sociol Forum 2: 718-745.

Massey DS (1990) Social Structure, Household Strategies, and the Cumulative Causation of Migration. Population Index 56: 3-26.

Massey DS, Arango J, Hugo G, Kouaouci A, Pellegrino A, and Taylor JE (1993) Theories of international migration: a review and appraisal. Popul Dev Rev 19: 431-66. 103

McKeown T, Record RG, and Turner RD (1975) An interpretation of the decline of mortality in England and Wales during the twentieth century. Popul Stud (Camb) 29: 391-422.

Miller JE (1991) Birth Intervals and Perinatal Health: An Investigation of Three Hypotheses. Family Planning Perspectives 23: 62-70.

Miller JE, and Huss-Ashmore R (1989) Do reproductive patterns affect maternal nutritional status? An analysis of maternal depletion in Lesotho. American Journal of Human Biology 1: 409-419.

Muhuri PK (1995) Health programs, maternal education, and differential child mortality in Matlab, Bangladesh. Popul Dev Rev 21: 813-34.

Muhuri PK, and Preston SH (1991) Effects of family composition on mortality differentials by sex among children in Matlab, Bangladesh. Popul Dev Rev 17: 415-34.

Palloni A, and Rafalimanana H (1999) The effects of infant mortality on fertility revisited: new evidence from Latin America. Demography 36: 41-58.

Pearson AW, and Collier P (1998) The integration and analysis of historical and environmental data using a geographic information system: landownership and agricultural productivity in Pembrokeshire c. 1850. Agricultural History Review 46: 162-176.

Pebley AR, and Stupp PW (1987) Reproductive Patterns and Child Mortality in Guatemala. Demography 24: 43-60.

Pooley C, and Turnbull J (1998) Migration and mobility in Britain since the eighteenth century. London: University College of London Press.

Preston SH, Heuveline P, and Guillot M (2000) Demography: Measuring and modeling population processes. London: Basil Blackwell.

Relethford JH, and Brennan ER (1982) Temporal trends in isolation by distance on Sanday, Orkney Islands. Hum Biol 54: 315-27. 104

RGS Detailed annual report of the Registrar-General of Scotland. Edinburgh and London: HMSO.

Roberts DF (1983) Genetic epidemiology. Am J Phys Anthropol 62: 67-70.

Roberts DF, and Roberts MJ (1983) Surnames and relationships: an Orkney study. Hum Biol 55: 341-7.

Rosenzweig MR (1977) The demand for children in farm households. The Journal of Political Economy 85: 123-146.

Ruggles S, Sobek M, Alexander T, Fitch CA, Goeken R, Hall PK, King M, and Ronnander C (2004) Integrated Public Use Microdata Series: Version 3.0: Minnesota Population Center.

Russell JC (1948) British Medieval Population. Albuquerque: Univerisity of New Mexico Press.

Sanderson SK, and Dubrow J (2000) Fertility decline in the modern world and in the original demographic transition: testing three theories with cross-national data. Popul Environ 21: 511-37.

SAS (2003) SAS 9.1 for Windows. Cary, NC: SAS Institute.

Schrank G (1995) An Orkney estate: Improvements at Graemeshall 1827-1888. East Linton: Tuckwell Press.

Schutjer WA, Stokes CS, and Poindexter JR (1983) Farm size, land ownership, and fertility in rural Egypt. Land Economics 59:393-403.

Scott N, Stevenson C, and Stout A, eds. (2003) Fae Quoy tae Castle: The buildings of Westray: Westray Building Preservation Trust.

Scott S, and Duncan CJ (2000) Interacting effects of nutrition and social class differentials on fertility and infant mortality in the pre-industrial population. Popul Stud (Camb) 54: 71-87. 105

Scrimshaw SC (1978) Infant mortality and behavior in the regulation of family size. Popul Dev Rev 4: 383-403.

Sear R, Mace R, and McGregor IA (2000) Maternal grandmothers improve nutritional status and survival of children in rural Gambia. Proc Biol Sci 267: 1641-7.

Sear R, Shanley D, McGregor IA, and Mace R (2001) The fitness of twin mothers: evidence from rural Gambia. Journal of Evolutionary Biology 14: 433-443.

Sear R, Steele F, McGregor IA, and Mace R (2002) The effects of kin on child mortality in rural Gambia. Demography 39: 43-63.

Singh S (1979) Demographic variables and the recent trend in fertility in Guyana, 1960-1971. Population Studies 33: 313-327.

Smith RM (1981) Fertility, economy, and household formation in England over three centuries. Popul Dev Rev 7: 595-622, 728-30.

Stearns SC (1992) The evolution of life histories. Oxford: Oxford University Press.

Strassmann BI, and Gillespie B (2002) Life-history theory, fertility and reproductive success in humans. Proc Biol Sci 269: 553-62.

Thomson WPL (1983) Kelp making in Orkney. Kirkwall: The Orkney Press.

Thomson WPL (2001) The new history of Orkney. Edinburgh: Mercat Press.

Tiefenthaler J (1997) Fertility and family time allocation in the Philippines. Popul Dev Rev 23: 377-97.

Tolnay SE, and Glynn PJ (1994) The persistence of high fertility in the American South on the eve of the baby boom. Demography 31: 615-31.

Trussell J, and Hammerslough C (1983) A hazards-model analysis of the covariates of infant and child mortality in Sri Lanka. Demography 20: 1-26. 106

Tymicki K (2004) Kin influence on female reproductive behavior: The evidence from the reconstitution of the Bejsce parish registers, 18th to 20th centuries, Poland. American Journal of Human Biology 16: 508-522.

Ulijaszek SJ (1995) Human energetics in biological anthropology. Cambridge: Cambridge University Press.

Vanlandingham M, and Hirschman C (2001) Population Pressure and Fertility in Pre- Transition Thailand. Population Studies 55: 233-248.

Vlassoff M (1982) Economic Utility of Children and Fertility in Rural India. Population Studies 36: 45-59.

Voland E (1998) Evolutionary ecology of human reproduction. Annual Review of Anthropology 27: 347-374.

Voland E, and Beise J (2002) Opposite effects of maternal and paternal grandmothers on infant survival in historical Krummhörn. Behavioral Ecology and Sociobiology 52: 435-443.

Watterson PA (1988) Infant mortality by father's occupation from the 1911 census of England and Wales. Demography 25: 289-306.

Wenham S (2001) A More Enterprising Spirit: The Parish and people of Holm in the 18th century Orkney. Kirkwall: Bellavista Publications.

Whyte I (1979) Agriculture and socieity in seventeenth century Scotland. Edinburgh: John Donald.

Winikoff B (1983) The Effects of Birth Spacing on Child and Maternal Health. Studies in Family Planning 14: 231-245.

Wood JW (1990) Fertility in anthropological populations. Ann Rev Anthropol 19: 211-242.

Wood JW (1994) Dynamics of human reproduction: Biology, biometry, demography. New York: Aldine De Gruyter. 107

Wood JW, Holman DJ, Weiss KM, Buchanan AV, and Lefor B (1992) Hazards models for human population biology. Yearbook of Physical Anthropology 35: 43-87.

Woods RI, Watterson PA, and Woodward JH (1988) The causes of rapid infant mortality decline in England and Wales, 1861-1921, Part I. Popul Stud (Camb) 42: 343-66.

Woods RI, Watterson PA, and Woodward JH (1989) The causes of rapid infant mortality decline in England and Wales, 1861-1921. Part II. Popul Stud (Camb) 43: 113-32.

Wrigley EA, and Schofield RS (1983) English population history from family reconstitution: summary results 1600-1799. Popul Stud (Camb) 37: 157-84.

Wrigley EA, and Schofield RS (1989) The population history of England, 1541-1871: A reconstruction. Cambridge: Cambridge University Press.

Yount KM (2004) Maternal resources, proximity of services, and curative care of boys and girls in Minya, Egypt 1995-97. Popul Stud (Camb) 58: 345-55.

108

Chapter 5 General Discussion and Summary

General Conclusions

The results of the analyses presented in this thesis point to the difficult life experienced by the Orkney farm household. Even today, as the dominant industry in the archipelago, farming poses many difficult decisions in order to maximize households’ economic opportunities and returns and the maintenance and stability of the family. Indeed whenever maternal or paternal cohort effects were considered in the analyses of fertility and mortality, they made little to no difference in the outcome variables. This suggests that, although the population is declining and aging, the effects of farm households and their internal dynamics play a large role in Orcadian society. Indeed when we consider the broader nature of the household and its role in determining individual outcomes, we find a wealth of support from the demographic literature on all population processes. These include household determinants of fertility decision making (Beckman 1984; Bock 2002; Friedman et al. 1994; Mason

1987), migration decision making (Massey 1990; Massey et al. 1993) and children’s health (Bloom et al. 2001; Gage et al. 1997; LeVine et al. 1991; Teitler 2001; Yount

2004) and education (Anh et al. 1998; Blake 1981; Downey 1995) outcomes. It should come as no surprise that the findings from this work indicate that the household as a higher level structure (simply meaning it has effects on all the individuals within it) has an influence on fertility and child mortality.

The ideas presented within this thesis were concerned with how the household actively balances its current state of age, sex and extension composition with its 109 future demographic behaviors, specifically child bearing and infant mortality. From the analyses presented here I suggest that households, especially farm households, in the Northern Isles tend to balance their current way of life with their future behaviors.

Farm households, relative to nonfarm households, tend to lengthen their birth intervals at all parities and the births they eventually do have tend to have better survival chances than those of their nonfarm peers.

When looking for explanations for these phenomena, we must consider the nature of the farm household. Based on the data presented in this thesis, we see that farm households tend to be larger on average than nonfarm households most likely because they are also more likely than nonfarm households to be extended households. When we consider farm households during the 19 th and early 20 th

centuries, we must realize that these (in most cases) are not large, mechanized farms

of hundreds or even thousands of acres; these are typically small farms less than 20

acres whose power was drawn overwhelmingly from human and animal energy.

Whatever the size, the land itself was a valuable commodity for anyone who could

manage to secure even a small field, for this meant that the household could

supplement their consumption with perhaps a modicum of root crops, green

vegetables or even a single milk cow. The farms in existence today on some of the

Northern Isles are typically owned by families that have had their roots in the isles, in

some way, often for hundreds of years. The exceptions are Sanday and Eday which

have been dominated by large, non-Orcadian owned farms since the 19 th century.

The ability of these families to solidify their interests over the past two centuries by increasing their rented land or more recently by purchasing land left by families that 110 emigrated from the isles or from Orkney completely, has benefited them not only economically but demographically as well. Since these farm families typically have fewer, more widely spaced children, and these children have higher odds of surviving, the farm households can out compete nonfarm households. They are able to do this because they balance their current and future consumption and production levels so as to have the minimum number of children, but the highest chances they will survive. In addition to producing a higher fraction of surviving children, by lengthening their birth intervals the households have a more advantageous age structure. The beneficial age structure is brought about because by the time the next child is born, the previous child could be nearing the age of, at least minimal, productivity. From a biological perspective, by maximizing their reproductive success via long birth intervals and increased material well-being brought about by access to land, farm households compensate for quantity of children with quality of children. The large numbers of children produced by some households were a detriment to their own children’s well-being and their long term balance of consumption to production. Families with too many children often could not secure more land because they had to commit so many resources into their offspring, many of whom would not live past early childhood. This did not present a sound strategy for these households, and I would hypothesize that it was the children of these nonfarm households, or households that were unable to solidify their interests in acquiring more land, that formed the bulk of emigrants from the Northern Isles over the past century. With respect to these migrants, they reacted to their own living 111 situations in a rational manner by pursuing opportunities elsewhere, typically

Australia, New Zealand, or North America.

With respect to the effects of post-reproductive grandparents on birth spacing

and child mortality, the results were mixed. While maternal grandparents tended to

increase the time between marriage and first birth, they had no effect on subsequent

births and no effect on either infant or childhood survival. Likewise paternal

grandparents showed no statistically significant effects on either fertility or childhood

mortality. What should we make of these findings in light of previous work showing

dramatic increases in child survival with the presence of post-reproductive

grandparents (Blurton Jones et al. 2002; Hawkes 2003, 2004; Voland and Beise

2002)? While this study is by no means a definitive nail in the supposition that post-

reproductive grandparents represent a net benefit to child survival, we may still

speculate about possible alternatives. The only evidence that we have for either the

affirmative or negative on this debate is that by having more living maternal

grandparents families tend to increase the length of the interval between marriage and

first birth, all other things being equal. One interpretation of this would be that

grandparents influence the length of the first birth interval by simply co-inhabiting the

house with the newlyweds, possibly reducing coital frequency. This would then

prolong the interval because the couple would possibly wait to have a child until they

had their own domicile. A second possible reason is that maternal grandparents

provide information on the negative consequences of short birth intervals (Cramer

1987, 1995) based on their own personal experiences and those of their peers. This

could increase the length of the birth interval by causing the newly married couple to 112 believe it would be better to wait before having a child. Lastly, the presence of post- reproductive grandparents could actually represent a net cost rather than a benefit, simply because if the grandparents are disabled, elderly or in some way otherwise dependent on their children for support they would be net consumers in the household. Based on the survivorship curves for the Northern Isles (if they are indeed correct), it is entirely possible that newlywed couples would have to care for their parents for an extended period of time, thus dispelling the idea that a living grandparent is necessarily a beneficial grandparent. The conclusion to this is that it is probably some of each of these explanations at work. If the newlywed couple needed support they would probably look to their parents first, if they were alive, for material or cultural (here meaning informational) support, and the grandparents could represent a net benefit. But if the parents were living a long, possibly disabled, life they could very well represent a cost to their children.

Future Directions

The analyses described in this thesis are but a fraction of what can be done with the data collected under the auspices of the North Orkney Population History

Project. This thesis represents the first stage of the process of family reconstitution and individual record linkage. The data here were the result of linking all live births to their respective parents’ marriage records only. By further linking these children to their respective marriage records we could produce an intergenerational pedigree process which could yield valuable data to further test the farm family hypothesis discussed in this thesis. Along these same lines, by linking the vital registration data to the actual census records we could characterize the households in the data by their 113 actual household experiences. An extension of this is also the spatial joining of related households. Based on some preliminary interview work done by the author in the summer of 2005 and on general observations during fieldwork in Orkney, I suspect that cooperation amongst neighbors is a major tool for overcoming temporary household labor shortages, especially if the households are related. Information collected in the ethnographic interviews with the people living on the Northern Isles today should provide key insight into just how this cooperation was managed. It is possible that these cooperative effects could offset some of the negative influences observed in the analysis presented here. I believe this would prove a very valuable test of a spatially explicit model of household altruism within a setting where such a system is likely to exist. Some preliminary work on this idea was done by the author

(Sparks 2006a) and suggested a negative influence on fertility of a large number of kin in the household, but the cross-household test remains of interest.

Further data collection by the NOPH will produce island-level indicators of environmental quality that should have important implications for the analysis of mortality.

The next part of this project I plan on undertaking is the cross-island linkage of individuals considered as censored in the data discussed herein. Without a doubt, the majority of migrants during the 20 th century represent an international movement out of Orkney; it is likely that many of the 19 th century movements were to other

islands in the archipelago. During data entry for this project, many occurrences were

noted where individuals were born on one island but died or married on another,

especially with the North Ronaldsay-Sanday and Westray-Papa Westray connections. 114

Not only would this process of linkage allow for a larger sample size for hazards analysis, it would also allow us to examine migrant selectivity with respect to health and family preferences within this small system of islands. Some preliminary work by this author suggests that migrants within Orkney have higher levels of fertility and disadvantageous household consumer/producer ratios owing to the lack of extended families (Sparks 2006b).

In addition to further analysis of the Orkney data, some of the ideas presented here are easily transferable to a modern setting, specifically the idea that the household environment can have major impacts on infant and child health outcomes.

In the current era of non-traditional family formation (i.e. the move away from the nuclear, married family household) the well-being of children is of great concern.

Although much work has been done on the psychosocial and educational effects

(Brown 2000, 2006; Brown et al. 2005; Hao and Xie 2002; Lloyd and Blanc 1996) of these alternative family forms and the general differences between cohabiting and marital unions (Carlson et al. 2004; Casper and Cohen 2000; Cherlin 1999; Graefe and Lichter 1999; Waite 1995; Waite et al. 1986), the impact of these households on child health has been less researched.

References

Anderson M. 1998. Fertility decline in Scotland, England and Wales, and Ireland: comparisons from the 1911 census of fertility. Popul Stud (Camb) 52(1):1-20.

Anderson M, and Morse DJ. 1993a. High fertility, high emigration, low nuptiality: adjustment processes in Scotland's demographic experience, 1861-1914, Part I. Popul Stud (Camb) 47(1):5-25. 115

Anderson M, and Morse DJ. 1993b. High fertility, high emigration, low nuptiality: adjustment processes in Scotland's demographic experience, 1861-1914, Part II. Popul Stud (Camb) 47(2):319-343.

Anh TS, Knodel J, Lam D, and Friedman J. 1998. Family size and children's education in Vietnam. Demography 35(1):57-70.

Axinn WG, and Thornton A. 1996. The influence of parents' martial dissolutions on children's attitudes toward family formation. Demography 33(1):66-81.

Ball H, and Hill CM. 1996. Reevaluating "Twin Infanticide". Current Anthropology 37(5):865-863.

Barclay RS. 1965. The population of Orkney 1755-1961. Kirkwall: Mackintosh.

Bauman KJ. 1999. Shifting family definitions: the effect of cohabitation and other nonfamily household relationships on measures of poverty. Demography 36(3):315-325.

Becker GS. 1981. A treatise on the family. Cambridge: Harvard University Press.

Becker GS, and Lewis HG. 1973. Interaction between quantity and quality of children. In: Schultz TW, editor. Economics of the family: marriage, children, and human capital. Chicago: University of Chicago Press. p 81-90.

Beckman LJ. 1984. Husbands' and wives' relative influence on fertility decisions and outcomes. Popul Environ 7(3):182-197.

Bentley GR, Goldberg T, and Jasienska G. 1993. The Fertility of Agricultural and Non-Agricultural Traditional Societies. Population Studies 47(2):269-281.

Berry RJ. 1985. The natural history of Orkney. London: Collins and Son.

Blake J. 1981. Family size and the quality of children. Demography 18(4):421-442.

Bloom SS, Wypij D, and Das Gupta M. 2001. Dimensions of women's autonomy and the influence on maternal health care utilization in a north Indian city. Demography 38(1):67-78. 116

Blurton Jones NG, Hawkes K, and O'Connell JF. 2002. Antiquity of postreproductive life: are there modern impacts on hunter- gatherer postreproductive life spans? Am J Human Biol 14(2):184-205.

Bock J. 2002. Introduction: evolutionary theory and the search for a unified theory of fertility. Am J Human Biol 14(2):145-148.

Bongaarts J. 1978. A framework for analyzing the proximate determinants of fertility. Popul Dev Rev 4(1):105-132.

Boserup E. 1987. Population and technology in preindustrial Europe. Popul Dev Rev 13(4):691-701, 764, 676.

Bourgeois-pichat J. 1981. Recent demographic change in western Europe: an assessment. Popul Dev Rev 7(1):19-42.

Bowers EJ. 1983. Patterns of adult mortality in the Orkney Islands [Ph. D. dissertation]. Philadelphia: University of Pennsylvania.

Brennan E. 1979. Kinship, demographic, social, and geographic characteristics of mate choice in a small human population [Ph. D. dissertation]. University Park: Pennsylvania State University.

Brennan ER. 1983a. Mortality patterns in anthropological populations. Hum Biol 55(1):1-7.

Brennan ER. 1983b. Pre-reproductive mortality and family structure: Sanday, Orkney Islands 1855-1974. Hum Biol 55(1):19-33.

Brennan ER, Leslie PW, and Dyke B. 1982. Mate choice and genetic structure Sanday, Orkney Islands, Scotland. Hum Biol 54(3):477-489.

Brennan ER, and Relethford JH. 1983. Temporal variation in the mating structure of Sanday, Orkney Islands. Ann Hum Biol 10(3):265-280.

Brittain AW. 1992. Birth spacing and child mortality in a Caribbean population. Hum Biol 64(2):223-241. 117

Bronson FH. 1995. Seasonal Variation in Human Reproduction: Environmental Factors. The Quarterly Review of Biology 70(2):141-164.

Brown SL. 2000. The effect of union type on psychological well-being: depression among cohabitors versus marrieds. J Health Soc Behav 41(3):241-255.

Brown SL. 2006. Family structure transitions and adolescent well-being. Demography 43(3):447-461.

Brown SL, Bulanda JR, and Lee GR. 2005. The significance of nonmarital cohabitation: marital status and mental health benefits among middle-aged and older adults. J Gerontol B Psychol Sci Soc Sci 60(1):S21-29.

Bumpass LL, and Raley RK. 1995. Redefining single-parent families: cohabitation and changing family reality. Demography 32(1):97-109.

Burch TK, and Matthews BJ. 1987. Household formation in developed societies. Popul Dev Rev 13(3):459-511, 570, 457.

Carlson M, McLanahan S, and England P. 2004. Union formation in fragile families. Demography 41(2):237-261.

Casper LM, and Cohen PN. 2000. How does POSSLQ measure up? Historical estimates of cohabitation. Demography 37(2):237-245.

Casterline JB, Cooksey EC, and Ismail AFE. 1989. Household Income and Child Survival in Egypt. Demography 26(1):15-35.

Cherlin AJ. 1999. Going to extremes: family structure, children's well-being, and social science. Demography 36(4):421-428.

Clay DC, and Johnson NE. 1992. Size of farm or size of family: Which comes first? Population Studies 46(3):491-505.

Cleland J, and Wilson C. 1987. Demand Theories of the Fertility Transition: an Iconoclastic View. Population Studies 41(1):5-30. 118

Coale A, and Watkins SC, editors. 1986. The decline of fertility in Europe. Princeton: Princeton University Press.

Cox DR. 1972. Regression models and life tables (with discussion). Journal of the Royal Statistical Society B34:187-220.

Cramer JC. 1979. Employment trends of young mothers and the opportunity cost of babies in the United States. Demography 16(2):177-197.

Cramer JC. 1987. Social factors and infant mortality: identifying high-risk groups and proximate causes. Demography 24(3):299-322.

Cramer JC. 1995. Racial and ethnic differences in birthweight: the role of income and financial assistance. Demography 32(2):231-247.

CRAN. 2007. R 2.4.1.

Cronk L. 1989. Low Socioeconomic Status and Female-Biased Parental Investment: The Mukogodo Example. American Anthropologist 91(2):414-429.

Curtis SL, Diamond I, and McDonald JW. 1993. Birth interval and family effects on postneonatal mortality in Brazil. Demography 30(1):33-43.

Davidson DA, and Simpson IA. 1994. Soils and landscape history: case studies from the Northern Isles of Scotland. In: Foster A, and Smout TC, editors. The history of soils and field systems. Aberdeen: Scottish Cultural Press.

Desai S. 1992. Children at risk: the role of family structure in Latin America and West Africa. Popul Dev Rev 18(4):689-717.

Dodgshon RA. 1994. Budgeting for survival: Nutrient flow and traditional highland farming. In: Foster S, and Smout TC, editors. The history of soils and field systems. Aberdeen: Scottish Cultural Press.

Downey DB. 1995. When bigger is not better: family size, parental resources, and children's educational performance. Am Sociol Rev 60(5):746-761. 119

Easterlin RA. 1975. An Economic Framework for Fertility Analysis. Studies in Family Planning 6(3):54-63.

Farah AA, and Preston SH. 1982. Child mortality differentials in Sudan. Popul Dev Rev 8(2):365-383.

Fenton A. 1978. The Northern Isles: Orkney and Shetland. Edinburgh: John Donald.

Firebaugh G. 1982. Population density and fertility in 22 Indian villages. Demography 19(4):481-494.

Firth J. 1974. Reminiscences of an Orkney parish. Stromness: Orkney Natural History Society.

Flinn M, editor. 1977. Scottish population history from the 17th century to the 1930s. Cambridge: Cambridge University Press.

Friedman D, Hechter M, and Kanazawa S. 1994. A theory of the value of children. Demography 31(3):375-401.

Gage AJ, Sommerfelt AE, and Piani AL. 1997. Household structure and childhood immunization in Niger and Nigeria. Demography 34(2):295-309.

Gage TB. 1993. The decline of mortality in England and Wales 1861 to 1964: decomposition by cause of death and component of mortality. Popul Stud (Camb) 47(1):47-66.

Galloway PR, Hammel EA, and Lee RD. 1994. Fertility Decline in Prussia, 1875- 1910: A Pooled Cross-Section Time Series Analysis. Population Studies 48(1):135-158.

Geronimus AT. 1987. On teenage childbearing and neonatal mortality in the United States. Popul Dev Rev 13(2):245-279.

Geronimus AT, and Korenman S. 1993. The socioeconomic costs of teenage childbearing: evidence and interpretation. Demography 30(2):281-290; discussion 291-286. 120

Gibson AJS, and Smout TC. 1995. Prices, food, and wages in Scotland 1550-1780. Cambridge: Cambridge University Press.

Goody J. 1996. Comparing family systems in Europe and Asia: are there different sets of rules? Popul Dev Rev 22(1):1-20.

Graefe DR, and Lichter DT. 1999. Life course transitions of American children: parental cohabitation, marriage, and single motherhood. Demography 36(2):205-217.

Guinnane TW, Okun BS, and Trussell J. 1994. What do we know about the timing of fertility transitions in Europe? Demography 31(1):1-20.

Gunnlaugsson GA. 1988. Family and household in Iceland 1801-1930. Stockholm: Almqvist and Wiksell International.

Gupta MD. 1997. Socio-Economic Status and Clustering of Child Deaths in Rural Punjab. Population Studies 51(2):191-202.

Hagen EH, Barrett HC, and Price ME. 2006. Do human parents face a quantity- quality tradeoff?: Evidence form a Shuar community. American Journal of Physical Anthopology 130:405-418.

Hajnal J. 1982. Two kinds of preindustrial household formation system. Popul Dev Rev 8(3):449-494.

Hammel EA. 2005a. Chayanov revisited: A model for the economics of complex kin units. Proceedings of The National Academy of Sciences, USA 102(19):7043- 7046.

Hammel EA. 2005b. Demographic dynamics and kinship in anthropological populations. Proceedings of The National Academy of Sciences, USA 102(6):2248-2253.

Hammel EA. 2005c. Kinship-based politics and the optimal size of kin groups. Proceedings of The National Academy of Sciences, USA 102(3):11951- 11956. 121

Hammel EA, and Laslett P. 1974. Comparing Household Structure over Time and between Cultures. Comparative Studies in Society and History 16(1):73-109.

Hao L, and Xie G. 2002. The complexity of endogeneity of family structure in explaining children's misbehavior. Social Science Research 31:1-28.

Harrison GA. 1976. Genetic and anthropological studies in the human adaptability section of the International Biological Programme. Philos Trans R Soc Lond B Biol Sci 274(934):437-445.

Hawkes K. 2003. Grandmothers and the evolution of human longevity. Am J Hum Biol 15(3):380-400.

Hawkes K. 2004. Human longevity: the grandmother effect. Nature 428(6979):128- 129.

Ho TJ. 1979. Time costs of rearing children in the rural Philippines. Popul Dev Rev 5(4):643-662.

Hoffman SD, Foster EM, and Furstenberg FF, Jr. 1993. Reevaluating the costs of teenage childbearing. Demography 30(1):1-13.

Hytten FE. 1980. Nutrition. In: Hytten FE, and Chamberlain g, editors. Clinical Physiology in Obstetrics. Oxford: Blackwell Scientific. p 163-192.

Johnson PL. 1990. Changing household composition, labor patterns, and fertility in a highland New Guinea population. Hum Ecol 18(4):403-416.

Kaplan H. 1994. Evolutionary and wealth flows theories of fertility: empirical tests and new models. Popul Dev Rev 20(4):753-791.

Knodel J. 1977. Family limitation and the fertility transition: Evidence form the age patterns of fertility in Europe and Asia. Population Studies 31(2):219-249.

Knodel JE. 1988. Demographic behavior in the past: A study of fourteen German village populations in the eighteenth and nineteenth centuries. Cambridge: Cambridge University Press. 122

Kramer KL. 2005a. Children's help and the pace of reproduction: Cooperative breeding in humans. Evolutionary Anthropology 14:224-237.

Kramer KL. 2005b. Maya children: Helpers at the farm. Cambridge: Harvard University Press.

Kramer KL, and Boone JL. 2002. Why intensive agriculturalists have higher fertility: A household energy budget approach. Current Anth 43(3):511-517.

Laslett P. 1969. Size and Structure of the Household in England Over Three Centuries. Population Studies 23(2):199-223.

Laslett P. 1970. The Comparative History of Household and Family. Journal of Social History 4(1):75-87.

Laslett P, and Wall R, editors. 1972. Household and family in past time. Cambridge: Cambridge University Press.

Lee RD. 1986. The value and allocation of time in high-income countries: implications for fertility. Comment. Popul Dev Rev 12(Suppl):108-110.

Lee RD, and Kramer KL. 2002. Children's economics roles in the Maya family life cycle: Cain, Caldwell, and Chayanov revisited. Population and Development Review 28(3):475-499.

Lehrer E. 1984. The impact of child mortality on spacing by parity: a Cox-regression analysis. Demography 21(3):323-337.

Leridon H. 1977. Human Fertility. Chicago: University of Chicago Press.

LeVine RA, LeVine SE, Richman A, Uribe FMT, Correa CS, and Miller PM. 1991. Women's Schooling and Child Care in the Demographic Transition: A Mexican Case Study. Population and Development Review 17(3):459-496.

Lloyd CB, and Blanc AR. 1996. Children's schooling in Sub-Saharan Africa: The role of fathers, mothers and others. Popul Dev Rev 22:265-298. 123

Lotka AJ. 1922. The Stability of the Normal Age Distribution. Proc Natl Acad Sci U S A 8(11):339-345.

Lotka AJ. 1989. Lotka on population study, ecology, and evolution. Popul Dev Rev 15(3):539-550.

Lumley T. 2007. The survival package, v 2.31.

Macbeth HM, and Boyce AJ. 1987. Anthropometric variation between migrants and non-migrants: Orkney Islands, Scotland. Ann Hum Biol 14(5):405-414.

Mace R, and Sear R. 1997. Birth interval and the sex of children in a traditional African population: an evolutionary analysis. J Biosoc Sci 29(4):499-507.

Manning WD, and Smock PJ. 1995. Why marry? Race and the transition to marriage among cohabitors. Demography 32(4):509-520.

Markle GE, and Pasco S. 1977. Family limitation among the Old Order Amish. Population Studies 31(2):267-280.

Marwick H. 1952. Orkney Farm Names. Kirkwall: W. R. Mackintosh.

Mason KO. 1987. The impact of women's social position on fertility in developing countries. Sociol Forum 2(4):718-745.

Massey DS. 1990. Social Structure, Household Strategies, and the Cumulative Causation of Migration. Population Index 56(1):3-26.

Massey DS, Arango J, Hugo G, Kouaouci A, Pellegrino A, and Taylor JE. 1993. Theories of international migration: a review and appraisal. Popul Dev Rev 19(3):431-466.

McKeown T, Record RG, and Turner RD. 1975. An interpretation of the decline of mortality in England and Wales during the twentieth century. Popul Stud (Camb) 29(3):391-422.

Miller JE. 1991. Birth Intervals and Perinatal Health: An Investigation of Three Hypotheses. Family Planning Perspectives 23(2):62-70. 124

Miller JE, and Huss-Ashmore R. 1989. Do reproductive patterns affect maternal nutritional status? An analysis of maternal depletion in Lesotho. American Journal of Human Biology 1:409-419.

Miller JE, Trussell J, Pebley AR, and Vaughan B. 1992. Birth spacing and child mortality in Bangladesh and the Philippines. Demography 29(2):305-318.

Muhuri PK. 1995. Health programs, maternal education, and differential child mortality in Matlab, Bangladesh. Popul Dev Rev 21(4):813-834.

Muhuri PK, and Preston SH. 1991. Effects of family composition on mortality differentials by sex among children in Matlab, Bangladesh. Popul Dev Rev 17(3):415-434.

Palloni A, and Rafalimanana H. 1999. The effects of infant mortality on fertility revisited: new evidence from Latin America. Demography 36(1):41-58.

Palloni A, and Tienda M. 1986. The effects of breastfeeding and pace of childbearing on mortality at early ages. Demography 23(1):31-52.

Pearson AW, and Collier P. 1998. The integration and analysis of historical and environmental data using a geographic information system: landownership and agricultural productivity in Pembrokeshire c. 1850. Agricultural History Review 46:162-176.

Pebley AR, and Stupp PW. 1987. Reproductive Patterns and Child Mortality in Guatemala. Demography 24(1):43-60.

Pooley C, and Turnbull J. 1998. Migration and mobility in Britain since the eighteenth century. London: University College of London Press.

Preston SH, Heuveline P, and Guillot M. 2000. Demography: Measuring and modeling population processes. London: Basil Blackwell.

Relethford JH, and Brennan ER. 1982. Temporal trends in isolation by distance on Sanday, Orkney Islands. Hum Biol 54(2):315-327. 125

RGS. Detailed annual report of the Registrar-General of Scotland. Edinburgh and London: HMSO.

Roberts DF. 1983. Genetic epidemiology. Am J Phys Anthropol 62(1):67-70.

Roberts DF, and Roberts MJ. 1983. Surnames and relationships: an Orkney study. Hum Biol 55(2):341-347.

Rosenzweig MR. 1977. The demand for children in farm households. The Journal of Political Economy 85(1):123-146.

Ruggles S, Sobek M, Alexander T, Fitch CA, Goeken R, Hall PK, King M, and Ronnander C. 2004. Integrated Public Use Microdata Series: Version 3.0. Minnesota Population Center.

Russell JC. 1948. British Medieval Population. Albuquerque: Univerisity of New Mexico Press.

Sanderson SK, and Dubrow J. 2000. Fertility decline in the modern world and in the original demographic transition: testing three theories with cross-national data. Popul Environ 21(6):511-537.

SAS. 2003. SAS 9.1 for Windows. Cary, NC: SAS Institute.

Schrank G. 1995. An Orkney estate: Improvements at Graemeshall 1827-1888. East Linton: Tuckwell Press.

Schutjer WA, Stokes CS, and Poindexter JR. 1983. Farm size, land ownership, and fertility in rural Egypt. Land Economics 59(4):393-403.

Scott N, Stevenson C, and Stout A, editors. 2003. Fae Quoy tae Castle: The buildings of Westray: Westray Building Preservation Trust.

Scott S, and Duncan CJ. 2000. Interacting effects of nutrition and social class differentials on fertility and infant mortality in the pre-industrial population. Popul Stud (Camb) 54(1):71-87. 126

Scrimshaw SC. 1978. Infant mortality and behavior in the regulation of family size. Popul Dev Rev 4(3):383-403.

Sear R, Mace R, and McGregor IA. 2000. Maternal grandmothers improve nutritional status and survival of children in rural Gambia. Proc Biol Sci 267(1453):1641- 1647.

Sear R, Shanley D, McGregor IA, and Mace R. 2001. The fitness of twin mothers: evidence from rural Gambia. Journal of Evolutionary Biology 14(3):433-443.

Sear R, Steele F, McGregor IA, and Mace R. 2002. The effects of kin on child mortality in rural Gambia. Demography 39(1):43-63.

Singh S. 1979. Demographic variables and the recent trend in fertility in Guyana, 1960-1971. Population Studies 33(2):313-327.

Smith RM. 1981. Fertility, economy, and household formation in England over three centuries. Popul Dev Rev 7(4):595-622, 728-530.

Sparks CS. 2006a. Intergenerational and kinship effects on household fertility: A test from the Orkney Islans, Scotland (Presented Paper). American Association of Physical Anthropologists. Anchorage, AK.

Sparks CS. 2006b. A spatial and temporal analysis of immigrant behavior in Northern Orkney 1851-1901 (poster). Population Association of America. Los Angeles, CA.

Stearns SC. 1992. The evolution of life histories. Oxford: Oxford University Press.

Strassmann BI, and Gillespie B. 2002. Life-history theory, fertility and reproductive success in humans. Proc Biol Sci 269(1491):553-562.

Teitler JO. 2001. Father involvement, child health and maternal health behavior. Children and Youth Services Review 23(4/5):403-425.

Thomson WPL. 1983. Kelp making in Orkney. Kirkwall: The Orkney Press.

Thomson WPL. 2001. The new history of Orkney. Edinburgh: Mercat Press. 127

Tiefenthaler J. 1997. Fertility and family time allocation in the Philippines. Popul Dev Rev 23(2):377-397.

Tolnay SE, and Glynn PJ. 1994. The persistence of high fertility in the American South on the eve of the baby boom. Demography 31(4):615-631.

Trussell J, and Hammerslough C. 1983. A hazards-model analysis of the covariates of infant and child mortality in Sri Lanka. Demography 20(1):1-26.

Tymicki K. 2004. Kin influence on female reproductive behavior: The evidence from the reconstitution of the Bejsce parish registers, 18th to 20th centuries, Poland. American Journal of Human Biology 16(5):508-522.

Ulijaszek SJ. 1995. Human energetics in biological anthropology. Cambridge: Cambridge University Press.

Vanlandingham M, and Hirschman C. 2001. Population Pressure and Fertility in Pre- Transition Thailand. Population Studies 55(3):233-248.

Vlassoff M. 1982. Economic Utility of Children and Fertility in Rural India. Population Studies 36(1):45-59.

Voland E. 1998. Evolutionary ecology of human reproduction. Annual Review of Anthropology 27:347-374.

Voland E, and Beise J. 2002. Opposite effects of maternal and paternal grandmothers on infant survival in historical Krummhörn. Behavioral Ecology and Sociobiology 52:435-443.

Waite LJ. 1995. Does marriage matter? Demography 32(4):483-507.

Waite LJ, Goldscheider FK, and Witsberger C. 1986. Nonfamily living and the erosion of traditional family orientations among young adults. Am Sociol Rev 51(4):541-554.

Watterson PA. 1988. Infant mortality by father's occupation from the 1911 census of England and Wales. Demography 25(2):289-306. 128

Wenham S. 2001. A More Enterprising Spirit: The Parish and people of Holm in the 18th century Orkney. Kirkwall: Bellavista Publications.

Whyte I. 1979. Agriculture and socieity in seventeenth century Scotland. Edinburgh: John Donald.

Winikoff B. 1983. The Effects of Birth Spacing on Child and Maternal Health. Studies in Family Planning 14(10):231-245.

Wood JW. 1994. Dynamics of human reproduction: Biology, biometry, demography. New York: Aldine De Gruyter.

Wood JW, Holman DJ, Weiss KM, Buchanan AV, and Lefor B. 1992. Hazards models for human population biology. Yearbook of Physical Anthropology 35:43-87.

Woods RI, Watterson PA, and Woodward JH. 1988. The causes of rapid infant mortality decline in England and Wales, 1861-1921, Part I. Popul Stud (Camb) 42(3):343-366.

Woods RI, Watterson PA, and Woodward JH. 1989. The causes of rapid infant mortality decline in England and Wales, 1861-1921. Part II. Popul Stud (Camb) 43(1):113-132.

Wrigley EA, and Schofield RS. 1983. English population history from family reconstitution: summary results 1600-1799. Popul Stud (Camb) 37(2):157- 184.

Wrigley EA, and Schofield RS. 1989. The population history of England, 1541-1871: A reconstruction. Cambridge: Cambridge University Press.

Yount KM. 2004. Maternal resources, proximity of services, and curative care of boys and girls in Minya, Egypt 1995-97. Popul Stud (Camb) 58(3):345-355.

129

Appendix: Database and Record Linkage

The database

The data on which these analyses are based come from a variety of sources. The primary source are the old parish records (OPRs) and standardized vital registration records administered by the Register General of Scotland. Beginning in summer

2003, data were collected for the islands of Eday, Pharay, Papa Westray, North

Ronaldsay, Sanday and Westray. Three primary record types form the basis of the data for this dissertation. The first are individual birth records. These contain information on the name of the child, day, moth and year of birth, house and island of birth, names, occupations, marital statuses and places and dates of marriage for the child’s parents. The second record type is marriage records. These contain names of the bride and groom, date and place of marriage, occupational status of the bride and groom, names, occupations, marital statuses and information on whether the parents of the bride and groom were alive at the time of marriage. In addition, they contain indicators of blood relationship between bride and groom. The third record type is death records; these contain the name of the deceased, date and time of death, age at death, marital status, name of the spouse of the deceased, names and occupations of the deceased’s parents, cause of death (if the death was attended by a physician) and length of sickness.

Record linkage

The process of record linkage was only begun after all records for given island were entered into the database. Since no standardized identifying information is contained 130 in the individual records (e.g. Social Security Number, National Health Service

Number), the degree of name redundancy was high, and each record gave little information about other family members, a manual record linkage process was undertaken to minimize erroneous linkages between record types. The process was undertaken as follows:

1) Assign all births on the island of interest with a unique identifying number

(here after UID)

2) Sort marriage record by groom’s/bride’s first name (1), year of birth (2),

father’s surname (3), father’s first name (4) and mother’s maiden name (5)

Check birth record for agreement between 1-5. Assign UID to groom if (1-5)

match or were within reasonable correctness (e.g. if (4) was listed as James in

the birth record and as Jimmy in the marriage record and all other criteria

match, the UID was assigned). 6

3) Repeat the process for all grooms/brides in marriage table.

4) Sort the death record by the same criteria as in Step 2 and assign UIDs if the

same match criteria are met.

5) For each union in the marriage table, assign a unique family identifier

(intended to keep track of whole families, hereafter FamID).

6) Sort birth table by 3, 4, 5, year of marriage (6) and month of marriage (7) of

parents.

7) Assign FamID to each birth matching on 3-7.

6 It should be noted that for deaths of individuals listed as being unmarried Step 2 was omitted and Step 4 was used. 131

This process was followed until all marriage, death and birth records were located for individuals born on the island of interest and each birth was given a FamID. This process, if all records for an individual were located, generated an individual life history consisting of birth, marriage, all subsequent births had, and death for each individual born on the island of interest. VITA Corey Shepard Sparks

Education 2001 M.A. in Anthropology, Statistics minor, University of Tennessee. 1998 B.A. in Anthropology, University of Tennessee

Employment 2007 Assistant Professor, University of Texas San Antonio Department of Demography and Organization Studies 2004-2006 NICHD Traineeship in Demography, Population Research Institute, Pennsylvania State University 2001-2004 Graduate Teaching Assistant, Pennsylvania State University, Department of Anthropology. 2000 Graduate Teaching Assistant, University of Tennessee, Department of Anthropology.

Publications and Relevant Presentations 2006 Sparks CS A spatial and temporal analysis of immigrant behavior in northern Orkney, 1851-1901. Poster presented at the Population Association of America Meeting, Los Angeles, CA. 2006 Sparks CS Intergenerational and kinship effects on household fertility: A test from the Orkney Islands, Scotland. Paper presented at the 2006 American Association of Physical Anthropologists Meeting 2003 Logan MH, Sparks CS, and Jantz RL Cranial Modification among 19 th Century Osages: Admixture and Loss of an Ethnic Marker. Plains Anthropologist 48 (187): 209-224. 2003 Sparks CS and Jantz RL Changing times, changing faces: Franz Boas’ Immigrant study in modern perspective. American Anthropologist 105 (2): 333-337. 2002 Sparks CS and Jantz RL A Reassessment of Human Cranial Plasticity: Boas Revisited. Proceedings of the National Academy of Sciences, USA, 99 (23): 14636- 14639.

Awards and Honors 2005 Baker Fund Dissertation Improvement Grant 2004 Pennsylvania State University RGSO Dissertation Improvement Grant 2004 Hill Foundation Dissertation Fellowship 2002 Winner E. E. Hunt Student Award for best student paper, Human Biology Association 2002 Winner (with Michael Aitkenhead) 17 th Annual Pennsylvania State University Graduate Research Exhibition