Population Growth

Population

and Development A Survey

Research

in the Philippines:

Alejandro N. Herrin

[. INTRODUCTION

This paper attempts to review the state of social science research on population and development relationships in. the Philippines with the aim of: (I) taking stock of what we know about such relationships as a guide to development planning, and (2) providing a basis for formulating recomII)endations to guide future research on the topic. The many population-related studies by independent investigators often contain reviews of past studies and suggestions for further research. In addition, efforts to bring together experts in various fields to focus on the status and directions of population-related research have also been made in the past (Concepcion 1966, 1969 ; Bulatao 1976). However, neither singly nor in combination do these efforts readily provide a unIfied view of the larger perspective needed for the development of a systematic knowledge base and a national research agenda specifically geared to the needs of policy-makers and planners in the 1980's.
Professor of Economics, University of the Philippines

Population

and Development

Research

289

Background

The increased concern about the role of population in development led many governments at the turn of the 1970's to adopt an official population poli~y whose main fo~us was the reduction of the rapid population growth. The main population program was the family planning program aimed at providing contraceptive technology to reduce fertility. The rapid declines in fertility noted in some countries have been attributed in varying extent to the impact of such programs. It was difficult, however, to assessthe impact such program has had on the population growth since the countries that had achieved the most rapid declines in fertility were also the ones which underwent rapid economic and social transformation. Among other factors, this led many governments to view rapid socioeconomic transformation as an important factor for the rapid and sustained decline in fertility. This view, further reinforced by discussions at international forums (e.g., the 1974 Bucharest Conference), has led to the recognition of the necessity of formulating population policies and programs as integral parts of the social and economic developmen~strategy . The present concern in th~ Philippines for such integration has been more clearly stated in the report of the Special Committee to Review the Philippine Population Program (1978). Noting that while some efforts have been taken to link the Philippine Population Program with other economic and social dimensions of development, the Committee found that, to a large extent, the. program has remained essentially a family planning program. Moreover, the Committee observed that whenever population was considered in the formulation of development plans, it was often treated. more as a demand variable than as a factor that can be influenced by economic and social development; Hence, the Committee recommended that "the Philippine Population Program should be designed on a broader scale and be fully integrated in the national development plans of the country. Economic, social and institutional policies and programs should be evolved with a conscious consideration of their impact on demographic behavior and objectives" (p. 122). Part of the failure to fully integrate population into the overall development plan is the difficulty of such integration at the operational level. This, in turn, is partly due to the complexity of the interaction between population and socioeconomic development,

290

Alejandro N. Herrin

and partly due to the inadequacy of the empirical knowledge base both internationally and nationally for the assessment of such interrelationships for development planning. An additional reason has been suggested, namely that "until recently, there has been an unclear institutional responsibility for advocating and overseeing the fuller integration of population in development concerns. This has resulted in a rather weak and uncoordinated effort towards integration" (Pante and Morales 1980). It is within the context of the above developments that this present effort to review social science research on population and development relationships in the Philippines is made. The ultimate aim is to hasten the operational integration of population in development concerns by expanding the knowledge basenecessary to support such effort.
Coverage and Approach

A note on the coverage and approach of this review is in order . Population-related research has sometimes been classified into: (a) research on demographic levels, trends and patterns; (b) research on socioeconomic-demographic relationships; and (c) action or program-oriented research. In the Philippines, considerable work has been done with respect to the first and third of these categories and some recent attempts at synthesis have been made. (See Concepcion and Smith 1977 for the first category, and Laing 1979 for the third category.) The main focus of this review, therefore, will be on the second category of researches as indicated by the title of this paper. Within this second broad category of researches, the review is organized around the examination of empirical studies focusing primarily on the.major demographic variables, namely: (a) fertility, (b) mortality, and (c) migration. Within the fertility variable, studies dealing with nuptiality and other immediate correlates of fertility are examined, while within the migration variable, studies dealing with internal and international migration are distinguished. Emphasis was placed on those empirical studies which analyzed data whose coverage included the nation as a whole (e.g., census, vital registration, national survey data) to reflect findings that would be representative of the national situation. F or each of the major variables, the review is further organized around the following questions, modifying somewhat the approach

Population

and Development

Research

291

adopted by the International Review Group of Social Science Research on Population and Development ( 1978) : (a) How much is known about the determinants, including individual, household and community level determinants, of the variable and its components? (b) How much is known about the consequences of the particular demographic behavior for individuals and families; for geographical, communities and specific groups; and for the nation as a whole? ( c) How much i8 known about the impact of public policies and programs on the variable in question? In view of the existing synthesis of studies dealing with the levels, trends and patterns of the demographic variables in question, only a brief introduction is made in this review to put the subsequent discussions into proper perspective, and to identify issues that could later be clarifed by further research. The subsequent sections of this report are organized as follows. In the succeeding four sections, a review of studies is made focusing respectively on mortality, fertility, internal migration, and international migration. The final section highlights the more important findings regarding population and development relationships ~d the most critical gaps in knowledge requiring serious investigatio~. Both of these aspects are discussed within a simple framework of population-development relationships geared towards a policymaker's and planner's viewpoint.
II. MORTALITY

Writing in 1974, F1ieger (1976) lamented the fact that, compared to fertility, information regarding mortality was quite scanty, and that whatever available information there was referred almost totally to the national population, leaving "an almost complete lack of mortality information for regions and provinces." Part of the problem has been that the major sources of data for the estimation of mortality were either deficient or defective. Vital registration data, for example, were oniy about 60 per cent complete for the Philippines in 1970, and the completeness of registration varied from 30 per cent in Western and Southern Mindanao to 80 per cent in Southern Tagalog (Abenoja and Flieger 1979). Censuses, on the other hand, contained defects in the age-sex structure data, making the application of

292

Alejandro

N .Herrin

indirect estimation techniques difficult. The above assessment of the state of mortality information is still generally valid today. However, notable gains have been achieved since then, and studies in the last five years (although accumulating slowly and unevenly) have provided us more information than has heretofore been available. These developments include the emergence of new sources of data such as the sample vital registration system project of the National Census and Statistics Office implemented during the period 1971-73~ and the 1968 and 1973 National Demographic Surveys. Paralleling these were the increased efforts to apply indirect techniques of mortality estimation from census, vital regi~tration, and survey data to produce not only national mortality estimates but also differential mortality estimates by region, p"rovince, and social group. In spite of this progress, however, importan~ gaps in knowledge still exist especially in the area of mortality-development relationships, as the subsequent discussion will reveal. A review of available studies, however, provides some concrete basis for making inferences regarding their relationship. Levels, Trends and Differentials National levels and trends. What we know about national mortality levels and trends in tentlS of such indicators as the crude death rate or the life expectancy at birth are based mostly upon the estimates of several investigators using different sources of data and measurement techniques (e.g., Aromin 1961, Lorimer 1966, Madigan and Avancefia 1965, Mijares 1976, Flieger 1976, Zablan 1975a). These sets of estimates compiled recently by Zablan (1978) revealo a pattern of gradually declining mQrtality from the earlier part of the century up to the beginning of World War II, a rapid decline during the postwar period up to-the end of the 1960's, and a slackenihg somewhat of the decline thereafter up to the early 1970's. Estimates for the mid- and late 1970's are hard to come by. The emerging pattern is roughly illustrated as follows. Estimates of life expectancy at birth place an upperbound value of 38 years in 1918 which increased to years in 1968, and ° to around 61 45 years in 1948, it rose to 59 40 years in 1938. From a level of years in 1973. The average °annual increase in the respective periods are 0.10, 0.70 and 0.40 years. The rapid decline in mortality after World War II is a common

Population

and Development

Research

293

observation in many developing countries. The major factor for such dramatic decline, especially in the earlier part of the postwar period, has often been attributed to effect of the introduction of relatively inexpensive public health measures more than to tne effect of genf;ral economic development. While the relative quantitative contribution of each factor to mortality decline in the immediate postwar period has yet to be firmly established in the Philippines, one additional and perhaps more important finding that needs explanation is the slackening of the mortality decline in the most recent past. While one can expect an eventual slowing down of the rate of mortality decline once son;te low level of mortality has been achieved, there is a suggestion that, on the basis of observed international mortality patterns, the slowing down of the mortality decline in the early 1970's may be somewhat premature. The average annual increment in life expectancy of 0.40 years observed between 1968 and 1973 is generally expected of populations which have achieved a high level of life expectancy of 70 years or so. If our estimate of life expectancy of around 60 years for that period is correct, the expected annual increment should have been still around 0.56 years (UN 1956, cited in Zablan 1977). That we are falling below the expected rate of mortality decline suggeststhe need for a more careful quantitative study of recent mortality determinants. Differentials. Several estimates of areal mortality differentials have been made. These include: (a) estimates from the NCSO's Sample Vital Registration Project which provide direct estimates of crude death rates by region for 1971 (Flieger 1976) and' for 1971-73 (Mijares 1976), (b) indirect estimates by region and province using census and vital registration data for 1970 (Flieger 1979), and (c) indirect estimates by region using 1968 and 1973 NDS data (e.g., Smith et al. 1975, Zablan 1975b). Zablan (1978)'has provided a convenient summary of some of the results of the st~dies conducted using the 1973 NDS data. All these studies point to the fact that the mortality levels by region, and even by provinces within regions, are far from uniform, suggesting that a single national mortality estimate hides more than it reveals in terms of mortality conditions in the country .For example, the results of the analyses from the 1968 and 1973 NDS data summarized in Zablan ( 1978) show that the life expectancy at birth (a measure not affected by the age composition of the population) in 1973 ranged from a high of 62-65 years in such regions as Metro

294

Alejandro

N .Herrin

Manila, Ilocos, Central Luzon, Southern Tagalog, Bicol, anq Central Visayas to a low of 56-58 years in such regions as Northern Mindanao, Western Mindanao, Eastern Visayas, and Cagayan Valley. The Philippine average was around 61 years. In addition to the differential mortality levels, regional es.timates for 1968 and 1973 further reveal areas where life expectancy has either declined or failed to increase as fast asin other regions. Regions with already low life expectancies in 1968 which declined further in 1973 include Cagayan Valley, Eastern Visayas, and Northern Mindanao. On the other hand, regions with already high life expectancies in 1968 which increased further in 1973 include Central Luzon, Southern Tagalog, Bicol, and Metro Manila. These differential trends tended to further widen existing regional mortality differentials over time. That some regions wi th high life expectancies in 1968 (Ilocos and Southern Mindanao) showed declines in life expectancies that tended to narrow differentials among regions somewhat is no consolation at all. Even differentials among provinces within a given region vary a great deal, sometimes by as much as 11-14 years of life expectancy, as in 1970 (Rizal vs. Palawan or Ilocos Norte vs. Mt. Province) (Flieger's data, 1979). An analysis of mortality dIfferentials by subgroups of the pOpUlation based on the 1973 NDS data was made by Alcantara (1975). The results, however, are highly subject to measurement error which renders interpretation difficult. Nevertheless, the results tend to show that childhood mortality measured in terms of the probability of dying from birth to age two (q2) tends to be higher among farm wives than among professionals and related workers, and higher among wives with low than with high educational attainments. Likewise, childhood mortality is higher in the rural than in the urban areas. Determinants What the above data do not reveal, however, is why differentials occur, and why trends vary among regions, in some cases reversing a favorable trend. Practically no systematic effort has yet been made to quantitatively identify the determinants of both areal and household or individual mortality levels and differentials. The most that has been done by way of empirical analysis has been to relate regional mortality levels with a set of socioeconomic indicators in an attempt

Population

and Development

Research

295

to develop a regional typology by levels of health welfare (Zablan 1977) or by computing simple correlation coefficients between provincial mortality and a few socioeconomic variables (Abenoja and Lim 1979). Zablan ( 1977) related regional life expectancies at birth with several socioeconomic indicators to reflect average regional levels of nutrition, sanitation, income, health facilities, health manpower, infrastructure, literacy, urbanization, farm activity, and gov~rnment per capita health expenditures. On the basis of the rank orders of the regions for the indicators, three types of region by level of health welfare were identified, namely: (a) regions with low levels of health welfare as exemplified by Cagayan Valley, Northern Mindanao, Western Mindanao, Southern Mindanao, Eastern Visayas, and Bicol; (b) regions with varying levels of health welfare as exemplified by Ilocos, Western Visayas, and Central Visayas ;' and (c) regions w~th high levels of health welfare as exemplified by Metro Manila, Souihern Tagalog, and Central Luzon. Zablan suggests th~t the disparities in the levels of health welfare across regions seem to reflect to a large extent the degree of access to health services, and to a lesser extent, the level of socioeconomic development obtaining in the region (Zablan 1977). This exercise, while suggesting important socioeconomic demographic interrelationships, stopped short of attempting to quantify such possible relationships, say, thIough the application of multivariate statistical techniques. The purpose of the exercise was more to target areas where substantial mortality decline can be achieved through public interventions rather than to identify the determinants of the observed mortality differentials. In a different approach, Abenoja -and Lim (1979) applied simple correlational analysis between 1970 provincial male life expectancy at birth in the Visayan regions and a few socioeconomic indicators. High positive and significant correlations were found between life expectancy and: (a) the opportunities for nonagricultural employment, indexed by the proportion of experienced workers in secondary and tertiarty sectors and by the number of large establishments per 10,000 population; (b) the average level of living as indexed by the percentage of households with refrigerators; ~nd (c) the degree of urbanization as indexed by population density. The results ofboth the studies cited above, however, are inadequate to determine the effect of these socioeconomic factors on mortality. Both mortality and socioeconomic variables were measured eifrler in

296

Alejandro

N .Herrin

the same year as in the case of the Abenoja and Lim study or even beyond the period of reference in which mortality levels were measured as in the case of the Zablan exercise. Such correlational findin~, however, provide an initial basis for future testing of the hypothesis with respect to the determinants of areal mortality differentials. An indirect approach to the understanding of the major sets of detenninants of mortality trend would be the analysis of the different causes of deaths. This approach, which was used by Preston (1975, 1979) to examine international mortality trends, looks at those portions of the decline in mortality which can be attributed to the declines in deaths due to different diseases.The assumption is that diseases vary in the degree to which they are responsive to living standards of are capable of being controlled by medical technology. Unfortunately, statistics, if any, on the causes of death are often unreliable in view of the inherent difficulty of assessing the precise cause of death, and are most likely to be incomplete. Nevertheless, from whatever data are available, mostly those collected by the Disease Intelligence Center of the Department (now Ministry) of Health, some tentative analysis can be made concerning the possible role played by general economic development anq public health measures in reducing mortality in postwar Philippines. Data on death rates by leading causes compiled from the above source and reported but unanalyzed by Zablan ( 1978) for the period from 1946 to 1972 reveal that death rates due to such diseases as pneumonia, influenza, bronchitis, dysentery, gastroenteritis, and nutritional deficiency, which are often associated with environmental changes due to economic development, have posted major declines since 1946. However, rates due to the control of such diseases as malaria, measles and tuberculosis, which are often associated with the application of inexpensive modern medical technology, have likewise declined. Between 1946 and 1960 death rates due to all of the causes of death enumerated above declined by 57 per cent (from 888 to 383 per 100,000). Two-thirds of this decline was accounted for by the decline ~n mortality from the first set of causes. Between 1960 and 1972, mortality declined by only 18 per cent (383 to 315 per 100;000), a much slower rate, of which the f1fSt set of causes contributed again about 70 per cent of such decline. It would appear thai mortality declines associated broadly with general economic development have been more important

Population

and Development

Research

29'

than mortality declines associated with the introduction of inexpensive public health measures, especially in the more recent period. This would tend to be contrary to international opinion regarding mortality declines in the postwar era in less developed countries. Although this type of analysis is at best crude in view of the incompleteness and perhaps in~ccuracy of data on the cause of death, and in view of the failure to account for the interaction between the two sources of mortality decline due to specific diseases, it suggests two important tentative hypotheses requiring further systematic study. First, the proportion of deaths due to causes. easily amenable to reduction by inexpensive, narrowly-defined public health ~easures is quite 1ow to start with, and any mortality reduction due to these causes, however large and rapid, would have a relatively minimal impact on overall mortality decline. Secondly, the rate of decline in mortality due to the first set ofcau~es has considerably slowed down between 1960 and 1972, suggesting that further gains in mortality decline would have to be associated with the effects of a broad-based public health program and economic development. Just exactly what type of program and what type of development will have the greatest impact on mortality is still to be determined.by more systematic social science research. Detailed analysis of better cause of death staustics could be an important aspect of such policy-oriented research from which a specific strategy of mortality reduction could be based. For example, the percentages of all deaths due to pneumonia and respiratory tuberculosis were still the highest in 1972 as they were in 1946, both accounting for 27 per cent of all deaths. Clearly, significant mortality reduction can be achieved by strategies that bear upon such diseases. A set of studies which do not deal directly with mqrtality 'but which could provide a concrete hypothesis for the study of mortality determinants are those dealing with nutrition, morbidity, an~ the distribution and utilization of health services (Florencio 1977; Layo 1977; Paqueo 1977a, 1977b; Bat tad 1977, 1978; Adorna 1977). These studies collectively suggest large differentials in nutritional status and morbidity by type of diseases across regions a~d provinces as well as differentials in the distribution of health s~rvices. These differentials may well be directly related to the observed areal differences in mortality rates. At the household level, these studies suggest the importance of such variables .as income and education, as well as household size and composition, as determi-

298

Alejandro

N. Herrin

nants of the incidence of malnutrition among preschoolers (Bat tad 1977, 1978; Paqueo 1977a). On the other hand, such factors as: (a) household size and composition, (b) environmental variables such as quality of drainage and quality of ventilation, (c) traditional health beliefs and health knowledge, and (d) education beyond elementary school, have been found to be significant predictors of morbidity, with the fIrst two sets of variables being more significant than the latter two (Layo 1977). Together these factors indirectly affect mortalit~ via their impact on nutrition and morbidity. Much more detailed analyses are needed, however, before a set of definitive conclusions could be made. Consequences The most discussed effect of declinjng mortality at the macro level is the increase in population growth. At the household level, this is reflected in increased family size, as more babies survive to adulthood. While declining mortality may reduce fertility somewhat, the results of international studies show that the replacement effect of reduced mortality is less than compensatory, thus leaving a net effect of increased family size (Preston 1975a). In turn, the consequences of increased family size are often discussed in relatiob to fertility, and hence, these studies will be treated in the next section.
Impact of Public Policy

Quantitative studies on the impact on mortality of public interventions, even specific health interventions, are practically nonexistent. This is probably due to the strongly held, butunsystematic.ally documented, view that the public health programs would naturally have an impact on mortality. Another reason may be the difficulty of evaluating the impact of such programs in view of the limited data available and of the complex factors that affect mortality. A recent evaluation of a large-scale maternal and child healthbased family planning project implemented in Bohol, for example, did not show a decline in mortality as conventionally measured during the five-year duration of the project (1974-79). While many factors were probably responsible for this observed lack of mortality impact (e.g., short duration of observation period and relatively small sample size), the evaluators suggest that perhaps an important

Population

and Development

Research

299

contributory factor may be that the project efforts were not closely tailored to the major causes of death, which in turn was primarily due to the inadequate data on thi~ aspect (Parado, Williamson, and Maturan 1980). Nevertheless, evaluations of the mortality impact of public health programs are necessary in order to determine what specific types of programs will tend to have the largest impact on mortality per unit of resource used. The concern for the need to reallocate resources within the health sector to redress existing geographical and social group imbalances could be guided by the results of such studies.
Towards A Research Agenda

The main problem from a policy and program standpoint seems to be the determination of the type of health program and the type of development that will have the greatest,effect on mortality. The research strategies suggested by the International Review Group of Social Science Research on Population and Development seem especially worthy of consideration, namely: (a) to increase the awareness of both policy.makers' and the general public of the existing inequality in health status between the rich and the poor; (b) to establish more precisely the cost-effectiveness and likely mortality impact of reoriented health policies, and to make such results wellknown; and (c) to determine the feasibility of reorienting health policies in the absence of brgader changes in political and institutional structure (IRG 1979, p. 80). In the Philippine context, the above strategies could be operationalized more specifically in terms of studies dealing with determinants and consequences of, and the impact of public policy on, mortality and perhaps on two of its closest correlates;.morbidity and nutrition. Firstly, the earlier attempts to describe and analyze mortality differentials by geographical areas and by social groups need to be continued. New data sets that have become available since the 1973 NDS include several rounds of the Area Fertility Surveys (one round includes direct information on mortality), the 1978 RPFS, and the recently concluded 1980 census. The application of indirect estimation techniques such as the Brass methods, as earlier done, could be pursued. These methods, however, provide estimates of past mortality levels and patterns, and are more accurate for child mortality in the recent past than for adult mortality. The

300

Alejandro

N. Herrin

application of indirect estimation techniques from widowhood data has been shown to yield reasonable estimates of adult mortality in some international applications. This approach could be explored, and resulting estimates could be fitted with child mortality estima~es to generate a complete and consistent set of age-specific mortality rates and life expectancies. In addition, infant mortality can be estimated from pregnancy history .data collected in the abovementioned surveys, and from which analysis of differentials could also be made. Finally, even as vital registration is being improved, various techniques of estimation using vital registration data could still yield reasonable estimates of mortality trends. Secondly, a multivariate analysis of known or suspected determinants of mortality trends and differentials needs to be made. Since existing data sets do not often contain such rich data on possible explanatory variables, serious consideration should be made in collecting such information in future su1Vey rounds of on-going demographic projects, or in designing new surveys specifically meant to implement such types of analyses. Complementary to such multivariate statistical analysis would be the car.eful in-depth microlevel analysis of the mechanisms by which broad socioeconomic correlates, e.g., income and education, affect mortality change. The first and second research leads suggested above should likewise apply to morbidity and nutrition studies. Thirdly, cause of death statistics should regularly be compiled and analyzed for possible trends and differentials. While admittedly such data sets tend to be unreliable in view of the difficulty in most cases of. determining the precise cause of death, they may still provide useful indicators of trends in types of causes that could most effectively be affected by redesigned public health measures. Fourthly, there is a need to evaluate the mortality effects of a wide range of public inte1Ventioris, not only the more narrowly defined public health measures but also such programs as food and nutrition, water supply, housing, and environmental sanitation. The evaluation coUld also seek to what extent current inte1Ventions first affect the immediate correlates of mortality among thepopulatioil groups exhibiting highest mortality, namely, morbidity and malnutrition. While the evaluation of .such inte1Ventions is made difficult by the lack of adequate evaluative research methodology , the experience of the ESIA/WID Projects could hopefully provide useful research strategies for such evaluations.

Alejandro

N. Herrin

301

III. FERTILITY

Levels, Trends and Differentials Compared to mortality, we have more information regarding levels; trends and differentials in fertility. While vital registration data still remain inadequate as a major sourCe of reliable estimates, census data and, more recently, the national and regional demographic surveys have provided current estimates of fertility at both the national and regional levels, as well as by social groups. What we know about fertility levels, trends and differentials were summarized recently in several studies (e.g., Concepcion and Smith 1977, de Guzman 1978, Concepcion and Mijares 1979, WFS-RPFS 1979). National level and trends. Available estimates of national fertility levels reveal a fairly constant birth rate ranging from 50 to 56 births per thousanQ population in the fust half of the century , gradually declining beginning in the 1950's to reach around 40 to 43 births per thousand in 1970. A somewhat faster decline occurred in the mid-1970's, so that by 1971, the crude birth rate had been reduced to around 30-32. Estimates of total fertility rate and, total marital fertility rate for the mor~ recent period from 1965 to 1977 likewise reveal a downward trend. Total fertility rate declined from 6.3 births per woman in 1965 to 5.89 in 1970 and then to 5.01 in 1977. Total marital fertility rate, on the other hand, declined from its 1965 level of 9.67 births per ever married woman to 9.65 iI) 1970 and 9.10 in 1977 (WFS-RPFS 1979, Table 5.12). The evidence, however, indicates that the decline in fertility rates under age 25 has been due largel~ to deferment of marriage than to reduction in fertility within marriage. In fact, marital fertility rates for the three periods have slightly increased for women aged 15-19 and 20-24. Within marriage, the declines started among women by age 25 and over, with the largest absolute declines occurring among women aged 25-39. This pattern suggeststhat older and higher parity women take the lead.in reducing fertility through contraception. This pattern of fertility decline reflects the relative roles played by the two proximate determinants of fertility change, to be described later, namely, changes in nuptiality patterns and changes in contraceptive prevalence. Regional differentials. Data from three national demographic surveys (1968 and 1973 NDS, and 1978 RPFS) provide .further information on rural-urban and regional fertility differentials. Preliminary estimates of mean children ever-born to ever-married

302

Alejandro

N. Herrin

women who have been married 10-19 years at the time of the 1978 RPFS su1Veyreveal lower fertility in the urban than in the rural areas. -4.5 versus 5.3; and generally lower fertility in Metro Manila than in the rest of the country , with a gradient of higher fertility as one moves from Luzon to Visayas and to Mindanao -4.2,5.1, 5.2 and 5.4, respectively (WFS-RPFS 1979, Table 5.7). A similar regional pattern is observed for the period 1963-67 and 1968-72 from the 1968 and 1973 NDS data. In this latter set of data as reported by de Guzman (1978, Table 103), declining total fertility rates is evident in almost all regions with the consequence of slightly narrowing regional fertility differentials during the two periods. It is difficult at present, however, to pinpoint precisely the determinants of these more recen t regional trends. Some earlier studies, however, have attempted to relate regional or p~ovincial fertility differentials to several socioeconomic factors. Smith (1971) found that sociocultural, demographic and socioeconomic factors as indexed in 1938 by such factors as physical characteristics of dwellings, education, occupation, literacy, religion, sex ratio and density are negatively associated with the overall level of fertility in 1960 but that .this effect is felt almost exclusively via the association between these factors with the marriage pattern. The association with the level of marital fertility is negligible and in most cases positive. Regional fertility differentials could also have been due to differential migration patterns which affected nuptiality patterns (Smith 1975a). Relating regional fertility differentials obserVed in the 1960's to socioeconomic factorR, proxied by per capita incomes, and health conditions, proxied by infant mortality rates, Flieger (1975) found no definite association with respect to the former variable, and a sJight positive relationship for the latter. Finally, Pascual (1971) related regional fertility to an index of regional development constructed by combining such indicators as the percentage of the population that is urban, the percentage of occupied dwelling units with radios, and the percentage of male labor force in nonagricultural occupations. She found a nonlinear relationship between these two variables, with fertility ratios (children ever born per 1,000 ever married women aged 35-44 or 45-54 increasing from the least developed areas to some level and decreasing thereafter. For both 1960 and 1968 where such a relationship was tested, the lowest fertility was found in the most and in the least developed regional groupings.

Population

and Development

Research

303

Fertility differentials by social groups. Studies on fertility differentials by social groups are many, and among the more intensive of these are those by Concepcion (1963, 1-964) and Pascual (1971). A recent summary of these studies is found in de Guzman ( 1978) incorporating data found in the 1973 NDS, while preliminary estimates from the 1978 RPFS are presented in WFS-RPFS ( 1979). In most general terms, fertility differentials have been more commonly noted, mostly in the hypothesized direction including possible nonlinearities, in social groups differing in education of the wife, occupation of the husband, income of the family or of the husband, work status of the woman, religion, type of households, and place of residence (rural-urban). Such information is suggestive oJ the factors affecting fertility. However, the lack of controls characterizing most descriptive analyses limits the usefulness of such analyses in assessing the effect of socioeconomic factors on fertility. Statistical analysis which controls for several explanatory variables from which further inferences can be made is described below. Determinants Some statistical associations. Several attempts have been made to distinguish several socioeconomic factors that are closely related to fertility by applying multivariate statistical analyses. Among the more clear-cut associations so far noted are those. between fertility (measured in terms of the number of children ever born), on the one hand, and education and income, on the other. Controlling for the timing and duration of marriage, age of woman and residence, Encarnacion (1973, 1975) and Canlas and Encarnacion (1977) have found that there is a threshold level of education of the wife and family income such that the effect of each respective variable on fertility is positive below' some threshold level and negative above it. Such results have been obtained using data from both the 1968 and 1973 national demographic survey. Among the less clear-cut statistical associations are those between fertility and female labor force participation. Earlier studies that controlled for some possible confounding variables (Concepcion 1973, Feranil and de Guzman 1977) have suggested that female labor force partiCipation per se may not greatly affect fertility; the type of work don.e and the place of work, to a large extent, are what really matter .Specifically, the fmdings show that working women in

304

Alejandro

N. Herrin

highly urbanized areas, as well as those specifically engaged in economic activities away from their homes, in nonfamily enterprises or on an employee status, generally exhibited lower fertility. Women working in home or family arrangements exhibited fertility levels comparable with those of nonworking women. " More recent econometric studies suggest that there is very little direct effect on fertility of female labor participation independent of other possible detenninants notably duration of marriage (Herrin 1980), and likewise a very weak sequential relationship is observed between past female employment and current or expected fertility (Herrin 1980, Ros~nzweig 1976). The observed variations bet~een female employment and fertility can in fact be explained in large part by their dependence on a common set of economic and social factors (Herrin 1980). With respect to the effect "of mortality on fertility, Hannan ( 1970) found that infant and child mortality had a significant positive impact on fertility especially among older women. The same relation was 'oberved when the perception of the risk of losing a child as measured by the community infant safety factor was used. Fernandez ( 1979) and Paqueo and F ernandez ( 1979) found that life expectancy appears to have an insignificant effect on the fertility of wives with family income below threshold values, while the effect is significantly negative above the threshold. While the studies cited above .(and others not cited) singly and in combination offer insights into the possible interaction between socioeconomic factors and demographiC: variables, the specific mechanisms involved in their interactions have yet to be fmnly established. In most cases,the socioeconomic variables that are being related to fertility are merely proxies for the major detenninants hypothesized. For example, in threshold-type studies, education and family income below some threshold are taken to represent more basically the level of health and nutrition of the mother which, in turn, directly affects her fecundity or capacity to bear live births. Social science research dealing directly with the effect of health and nutrition of mothers on fertility has yet to be conducted systematically in the Philippines. On the other hand, above some threshold values, both education and family income could represent a host of intervening factors which ultimately affect fertility. A review" of the international literature on the relation between education and fertility (for example, Cochrilne 1978), suggests that education

Population

an-d Development

Research

305

could affect fertility indirectly through its effect on the age at marriage, contraceptive knowledge, alternativ~ satisfactions, infant mortality, etc. Following a framework developed by Davis and Blake ( 1956), understanding the determinants of fertility would require more intensive investigations on it~ more proximate determinants, i.e., those factors related to entry into unions, contraception, and gestation. Among the proximate determinants of fertility, nuptiality patterns and contraception have received attention in the recent past, notably in studies by Smith and Laing. Only recently have data and analyses been made on such other proximate determinants as breastfeeding, separation of spouses, age at menarche, etc. (WFS-RPFS 1979, Laing 1979). Nuptiality patterns. It was mentioned earlier that fertility trends observed in the more recent period especially among younger women appear to be much more a result of changing nuptiality patterns than of declining marital fertility. Thus, continuing information on nuptiality patterns is important in understanding the processes behind fertility changes. A recent synthesis of available analyses on trends and differentials in nuptiality as well as inferences on the social processes involved was made by Smith ( 1978). Marital status distributions by age obtained from census data from 1903 to 1970 reveal a significant long-terms trend which is more pronounced for females than for males. The percentage of those never married increased steadily especially among younger women aged 15-19 and 20-24 over the seven decades, with the shift among women 15-19 years of age occurring before 1939, while for women aged 20-24 the shift occurred later (Smith 1978, Table 113). The singulate mean age at marriage has risen from 20.9 years in 1903 to 22.8 years in 1970. Data from the same sources likewise reveal areal variations in marriage patterns both at the regional and provincial levels. Some .of the most sizable nuptiality differentials have been traced to the effects of selective migration with respect to age, sex and marital status. The overall nuptiality patterns have also been related to three important social processes as gleaned from the 1973 National Demographic Survey by Smith (1978). These are: (a) urbanization and the expanded role of females in rural-to-urban transfers as they seek~

306

Alejandro

N. Herrin

jobs and education; (b) the rise of mass education, and the increased participation of females therein; and (c) the growth of the nonagricultural labor force, in which females have also had an increasingly important role~ The first factor tends to lower the sex ratio as migration to urban areas becomes female-dominated. This reduces the probability of early marriage. Data provided by Smith (1978), as well as the most recent preliminary estimates by WFS-RPFS ( 1979), show. an increasing age at marriage as educational level increases. Finally, female employment especially in the modern sector and in high level jobs is associated with delayed marriage, partly due to the effect of education. In summary , the source of nuptiality change over the course of seven decades can be traced, on the one hand, to environmental pressure on the traditional systems of landholding in the rural sectors, and on the other, to several interrelated processes of modernization including urbanization, educational expapsion and the shifting composition of the labor force. The first of these forces had )ed to differential migration patterns which in turn affected marriage patterns in both receiving and sending areas in the earlier part of the century; while the second set of forces played an increasingly larger role during the postwar period especially beginning in the 1960's. If nuptiality is a detenninant of fertility, how much of fertility change can be attributed to changes in the marriage pattern? Smith (1975b), analyzing data from the 1900 census and the 1973 NDS, found that between 1960 and 1970, 15 per cent of the decline in overall fertility in the Philippines could be attributed to nuptiality, while for the earlier period from 1903 to 1960, 63 per cent of the decline in overall fertility was due to nuptiality. Furthennore, large regional variations characterize the role of nuptiality in fertility change. Nuptiality accounts for all the changes in overall fertility in three regions, and for between 40 .3;lld 73 per cent in five other regions. In Metro Manila, nuptiality accounted for only less than five per cent of the change in overall fertility from 1960 to 1970. Metro Manila already had a much lower overall fertility by 1960 compared with the other regions: the more recent decline in overall fertility, therefore, is due mostly to changes in marital fertility through contraception. Contraception and other determinants of marital fertility. The evidence on fertility levels and trends presented earlier suggeststhat

lation

and De VI

~n l' R

307

the decline in total fertility rate obselVed in the period ~--m 1965 frnJ to 1977 has been due to changes in nuptiality patterns especially among younger women, and to changes in marital fertility for women aged 25 years and over. Within marriage, however, fertility is proximately determined by severalintennedi~te variables including contraception, involuntary infecundity; voluntary and involuntary abstinence, and induced and spontaneous abortions (Davis and Blake 1956). While several studies have been made on contraceptive prevalence Nhich has increased from 16 per cent in 1968 to 42 per cent in 1978 :Laing 1979), very little data have been available until recently regarding the other proximate determinants of fertility. Laing (1979) tells of an analysis of 1974 National Acceptor SulVey (NAS) data which indicated that family planning acceptors who breastfed their children were protected for over four months longer. on the average, than acceptors who did not breastfeed. The effect of breastfeeding was equivalent to 0.15 births averted, comparable to the averageprotection provided by condoms following an acceptance of that metht"\~
In ,the recently on that factors have a direct completed other effect postpartum of analysis amenorrhea, of that on the the ovulation, of 85 for in an rural the per spouses, length is and of the sexual has been postpartum age at relations, made of sexual menarche. because that Data 1549 months. to women. to breastfeed breastfeed The the length length or w! of . exerts for the the years Older their of 1978 than on age WFS-RPFS at marriage These su1Vey, and use information of contrainformaof mensetc.

was s ~vy.~:Jn npnt;, tion truation, A

-;ought

fertility. amenorrhea,

included regularity abstinence,

on

breastfeeding, frequency

postpartum the data on

preliminary

breastfeeding, temporary results show suppressing influence inte1Val breastfed and those slightly postpartum in in

postpartum separation general effect on

abstinence, The of

breastfeeding, factor inte1Val. of of tended of related women 11.2

its

greatest last of women children the closed age

length that child reside than

birth cent

revealed their who longer amenorrhea months.

average areas groups

other is

positively who did

breastfeeding

Women

not

up while period mean

to

two women

months who

had breastfed

a

mean for

amenorrheic a fun year

amenorrheic of 8.6

months. period

Women of

12.5

who months. breastfed

110 breastted Only tor of 3.5 months, I: leriod hac 1 a mean amenorrheic fnr 30 months had a prolonged, can lengttL--l

Postpartum

abstinence,

if

sufficiently

len

308

Alejandro

N. Herrin

the pregnancy intervals. The duration of postpartum abstinence is usually related to medical reasons, cultural norms, sociai pressure, and individual inclination. The data from the survey- suggest, however, that postpar.tum abstinence does not constitute an important factor in determining the length of either the open or the last closed pregnancy interval. The majority of the women (56 per cent) had resumed sexual relations two months after the pregnancy had ended. Higher postpartum abstinence is directly but slightly related to age of the woman and indirectly to level of education. Temporary separation of spouses due to sickness, work or family obligations can also lengthen the pregnancy intervals. The available data, however, suggest that temporary separation is not widespread in the Philippines (only 3 per cent reported a temporary separation of 3 months or more), nor is it of long duration (the mean length of temporary separation of all women is a mere 0.3 months). None of the background variables shows any relationship with duration of separation of spouses. Finally, the onset of menstruation, which is a biological factor influenced by women's general health and nutritional state, ranged from 12 to 19 years of a woman's age, averaging 13.9 years. The data further reveal that younger women tended to have an earlier age at menarche. This suggests that the trend in health and nutritional levels in the country over the years may have affected this trend of increasing age at menarche, which in turn could have some influence on overall fertility. The trend towards later age at marriage, however, would tend to minimize the impact of this factor on overall fertility. In view of the deficiencies of the quality of the data noted by the study, more detailed analyses are required to assessthe implications of these proximate determinants of marital fertility. Value of children. One of the more recent social science contributions to the understanding of fertility is the literature on the value of children (VOC) conducted in several countries including the Philippines (e.g., Bulatao 1975, 1978, 1979a, 1979b; and Bulatao and Arnold 1977). The main purpose of the sutdies was to identify various domains in which, the Filipino child is perceived to provide some utility, and to determine whether the degree of a person's concern with these domains relates to childbearing preferences. One advantage of such an approach is that it puts the conventional economic cost-benefit calculus commonly suggested by

Population

and Development

Research

309

economists into the broader social-cultural and psychological context of childbearing decisions. One disadvantage, however, is that the values are related more to fertility preferences rather than to actual fertility. Nevertheless, the results are interesting and, in most cases, reassuring for those who have confmed their investigations solely to the economic determinants of fertility. In the more detailed analysis of the data obtained from a national sample of 1,691 wives and 382 of their husbands in 1975, Bulatao (1978) reports that the value domain within which children are considered are several: (a) the domain of instrumental assistance or practical help, (b) the domain of interactions that are socioemotionally rewarding, (c) the domain of psychological appreciation, (d) the domain of coping with social pressures; (e) the domain of marital security and closeness, (f) and the domain of family and kin preservation. The instrumental-assistance value, fmancial help expected of children, old age security , help with household chores, and caring for other children were found to be highly salient and most central values. However, the second domain, the one on interactions that are socioemotionally rewarding, such as companionship which the children provide, the opportunity to give and receive love and affection, happiness from being with childreI}, etc., were about equal in salience to instrumental-assistance values and perhaps marginally ahead in centrality. The third and fourth domains were found to be both low in salience and in centrality while the fifth and the sixth were low in salience but high in centrality. The four major disvalues include: (a) emotional strain of having children, including worries about childbearing; (b) financial problems caused or aggravated by children; (c) restriction on parents' activities and limitations on time for one's self or for one's spouse; and (d) concern about overpopulation. The results of the analysis indicate that fmancial costs were less salient than worries of childrearing, but appeared more central and ranked first in importance among the disvalues. The other two sets of disvalues appeared less frequently and received less importance, indicating that few respondents see a child in the context of opportunity costs or the social costs of overpopulation. The analysis also revealed that the contribution of a child in each value domain depends on at least two factors -its sex and its birth ranking. Childbearing intentions are influe.nced by differen.t considerations as a family grows. The analysis also revealed that Filipino

310

Alejandro

N. Herrin

parents appear to value children for largely individualistic reasons. The respondents tend to be lessinfluenced by pressuresfrom relatives, from the community, or from religion as reasons for childbearing. "Such pressures may operate nevertheless through the value themselves, through couples internalizing. and identifying as their own personal motives the dominant cultural patterns to valuing children" (p. 171). Finally, changes in the perceived values and disvalues of children may occur as modernization proceeds. In summary , the VOC studies, by providing a better understanding of the values of children, should provide significant insights into fertility behavior as well as guide efforts toward fertility control that consider and, possibly compensate for, the values that may be lost as familes reduce their size. Consequences of Fertility Trends

Relative to the studies dealing with the correlates of fertility, very few quantitative studies have been done on the actual consequences of fertility change either at the macro or the micro level. Part of the reason is the lack of adequate and reliable historical data that would allow such quantitative analysis. Another reason is perhaps the fact that the persuasive rhetoric of the 1960's (both at the international and local levels) regarding the adverse consequences of rapid population growth convinced government authorities of the need to launch public pr,?grams to reduce the birth rate. Once such a commitment was obtained, there was little need for more detailed studies showing the .generally held but often simplistic view that rapid population growth will have adverse consequences on the national well-being. A review of available Philippine studies dealing with implications of high fertility is made below. Macro studies. Largely influenced by the earlier work of Coale and Hoover ( 1958) on the impact of alternative fertility trends, several studies have been conducted in the Philippines to examine the impact of alternative fertility trends on such macro variables as per capita income, saving and investments, and employment. In general, these macrolevel studies are simulation exercises to determine the economic implication of alternative fertility trends (or alternative population growth paths). They do not, therefore, reflect actual consequences of past demographic trends. Notable among these

Population

and Development

Research

311

earlier studies are those by Lampman (1967) and Ruprecht (1967, 1969). In examining some interactions between economic growth and population change in the Philippines, Lampman ( 1967) asked two questions: (a) what is the economic price or cost associated with the postwar acceleration of population growth? and what economic benefits would flow from a gradual return to a lower rate of population growth?; and (b) given the present rate ofpopulation growth, what cost must be paid to accelerate the rate of economic growth? The answer to the first question is based on a simple mechanical exercise of assuming a given GNP , then calculating the per capita GNP under constant fertility and a sharp decline in fertility. Predictably, the result would be a higher per capita under the latter than in the former fertility regime. Similarly, given the level of GNP, since population will be larger under constant fertility, and a larger population would require larger investments and social expenditures, the amount available for consumption will necessarily be lower under a -higher fertility regime than under a declining one. -Hence, consumption per capita will be lower under constant fertility. With respect to the second question, the price of growth under the high rate of population growth situation will tend to be higher at both the aggregative level (more capital, more labor, more technical advance, and so forth) and the intersectoral and intergroup level (more risk, loss of preferred and secured status for some, and considerable change in the way of life for all). Ruprecht ( 1967), on the other hand, developed an econometric model which allows the projection of GNP as a function of land, capital, labor and time, and two submodels which allow for the projection of three alternative population growth, and of savings and investment. Having projected GNP, population growth rates, investment rates, etc., under different conditions of fertility control, the advantage in terms of per capita GNP of immediate fertility control over no control or postponed control is demonstrated. One important implication of the exercise is that the initiation of fertility decline cannot wait until its- adverse economic effects have become apparent. By that time it may be too late as the population momenturn at that point would be overpowering. Ruprecht ( 1979), in a subsequent study, attempted to analyze the impact Qf alternative population trends and the consequent growth of income on the structure of the economy (in the inputoutput sense), on the assumption that different rates of population

312

Alejandro

N. Herrin

growth would have somewhat different effects on different sectors of the economy, With the growth of some sectors being retarded, and the growth of others perhaps accelerated. He also examined the employment implications of the resulting structural effect and, at a more disaggregated level, the structural implications on several manufacturing sectors. The results of the exercise suggest that, for the Philippines, "a reduction in fertility would contribute to an economic structure which emphasized non-agricultural activities to a greater extent, was more capable of providing full employment, and which produced in the manufacturing sector a greater inducement for subsequent growtht' (p. 11). Among other limitations of studies of the type represented above is the omission of the cost of a birth control program necessary to effect the fertility reduction in the first place. Such cost could in fact be sizable; hence there is a need to explicitly consider this cost in analyzing the macroeconomic benefits to be derived from reduced fertility . The recent work by Paqueo (1974, 1977) specifically addressed this issue of the cost of the public intervention to assessthe implications of fertility reduction on the economy. In general, Paqueo constructed a family planning submodel that allows a translation of the number of family planning acceptors into births averted. This submodel is then grafted into a larger econometric model which is a modification of an earlier economic-demographic model developed by Encarnacion, Mangahas, Paqueo and Smith (1974) to analyze the economic effects of the birth control program. In general, the results of the ~mulation exercise by Paqueo suggest that the effect of birth control on per capita income and real wage rate is significant. Family incomes, however, appear largely unaffected, and the effect on the traditional investment-to-output ratio seems minimal. Of considerable significance is the fmding that, while per capita incomes tend to increase, aggregate output due to a relatively smaller labor force is actually reduced. The payoff of the fertility reductiont therefore, is essentially due to the decrease in the number of persons sharing national output and not from increased production and saving. As Paqueo ( 1977) concludes: .'This observation would suggest that population control does not necessarily lead to more rapid growth defmed as sustained increase in total output. This interpretation should, of course, be qualified by the fact that many causal processes whereby family planning

Population and Development

Research

313

could enhance productivity and capital (human and material) formation are not included in the model. Nevertheless, in the light of the debate regarding development versus population control, it would seem appropriate to end this study by noting that family planning is not a substitute for effective development policies" (p. 215). In addition to the above studies looking at the probable consequences of alternative fertility (and, therefore, population growth) trends on per capita income and other aspects related to it, several studies have looked at specific sectors of the national economy that will be affected by pQpulation growth. These sectors include health, education, food, housing, natural resources and environment (Calzado et al. 1978, de la Paz 1978, Intengan 1978, Luna 1978, etc.). These studies generally suggest that, while rapid population growth exerts pressure on the demand for basic services, the problems associated with their provision are more directly due to a set of factors more complex than what a rapid increase in population would imply. Micro-level studies. Turning now to the microlevel impact of fertility change, several studies have focused on the impact of family size on faroily savings and expenditures (Peek 1974, Power 1971, Mangahas 1974), on morbidity rates (Layo 1977), and on the nutritional status of household members (Bat tad 1977). Using data from the PSSH of the BCS for 1961, 1965 and 1971, Peek (1974) examined the effects of family size among others on household savings, defined alternatively as per capita savings, per househoJd savings, and per adult equivalent savings. Among the tentative conclusions are as follows: (a) gross dependency burden (not taking into account the contribution of the dependents to family income) has a negative impact on the. savings rate which is only to a small extent offset by economies of scale in consumption; and (b) there is a negative life cycle effect on savings in terms of expected dependency burden, which more than offsets the positive economies of scale effect of savings. One limitation of Pe.ek's study is that the effect of family size on family income is not explicitly considered in the model. Such effect, however, is considered by a study conducted by Mangahas (19?4). This study attempted to quantify two processes by which family size may affect family expenditure (and, as a residual, savings). Firstly, he" consider~d the relationship between family size and the number of working!amily members. The size of the family's working force then affects family income and, consequen tly , famiJy expend-

314

Alejandro

N .Herrin

iture. Secondly, the study considered the relationship between family size and the number of adult-equivalent consumers in the family. The latter variable, jointly with family income, determines family expenditure. The data used were the Family Income and Expenditure Surveys (PIES) of 1957; 1961, 1965 and 1971 of the Bureau of the Census and Statistics, and the National Demographic Survey (NDS) of 1968. The interrelationships. suggested are as follows: increases in family size lead to increases in the family labor force and, in turn, to increases the number in the number of working members, which, in combination with the age of the household head, the education of the wife, and (in urban areas) the labor force participation of the wife, then determine family income. In the second process, family size determines the number of equivalent adult members in the family. In combination with family income, this in turn detel111ines the consumption level of the family. With respect to the potential effect of family size on morbidity, Layo ( 1977) found that the most important determinants of total illness in the household as measured by the magnitude and significance of the regression coefficients are the number of household members 0-5 years old and the number of older members aged 45 and over controlling for such factors aseducation, per capita income, rural-urban residence, quality of drainage, of ventilation and of water, health beliefs and knowledge. For acute illness, the best predictors continue to be the demographic .variables especially the number of household members 0-5 years old. Bat tad ( 1977), in a study of the detenninants of the nutritional status of Laguna preschoolers of 1975, found a negative effect of the number of children less than six years on a measure of child nutritional status, controlling for such factors as income per capita, education of the mother, age and sex of the child, mother.s nutritional status, and incidence of chronic illness. The measure of child nutritional status is the ratio of the child.s actual weight to the standard weight for age and sex. The main explanation for such negative relationship is as follows: the more young children, the greater the time inputs into child care by the mother and other household members. and the harder it gets to meet each child's nutrition needs. Bat tad also states that the negative marginal effect of the number of children was larger for 2-3 years olds than for 4-6 year olds. This is probably explained by

Population

and Development

Research

315

the fact that children aged 2-3 years are still evolving feeding habits, are just starting to exercise self reliance, and still need much supervision from the mother. By the age of 4-6, a child becomes more capable of handling the shift in the mother's attention but the net effect may still be a lowering of nutritional status. Popkin (1976) found that an addition to the number of children aged zero to 6 years increased the Laguna mother's time for child care. That the nutritional status of children aged 0-2 was not significantly affected by an additional preschooler probably indicates that the Lagunamothers concentrate more attention on the newborn or younger preschoolers than on the older ones. Boulier ( 1976) found that an additional infant increased a mother's time for child care mote than an additional older preschooler and that the increase moved according to family size. In larger families, more older children substituted for the mother's time; hence, the increase was not as large as in small families. This substitution of mother's care may contribute to the decline in older preschoolers' nutrition status. Impact of Public Policy Studies dealing with the impact of public interventions on fertility, other than that of the family planning program, are virtually nonexistent. The studies thus far available (e.g., Herrin 1979, on the impact of rural infrastructure, and Paqueo 1978, on the impact of public health and education) are as yet too tentative to provide fIrm guideljnes for possible public. policy redirection or program redesignning. Nevertheless, there is growing interest in evaluating the impact on demographic trends, notably fertility , of public programs primarily designed with nondemographic objectives in mind. While such evaluation activities currently suffer methodological and research design problems, such studies could have a significant contribution to policy-making and program design in the near future as studies (e.g., the ESIA/WID Project) begin to cumulate. The past and current planning and design of public programs, therefore, can still be characterized as essentially population-responsive. Estimates of age-sex composition and geographical distribution of the population are virtually the only demographic inputs in desi~~g the scale and coverage of public programs. The designing of public programs with a view to also affecting demographic variables

316

Alejandro

N .Herrin

has yet to wait for research on the topic.

more defmitive results from social science

Towards a Research Agenda

On the basis of the foregoing discussion of social science research bearing upon the relationship between fertility and economic development, several areas for further research seem worth exploring. Levels, trends and differentials. First, we have noted that more and more demographic data of reasonably good quality are becoming available. These data sets include the censuses, the regular na~ional demographic surveys, and the several rounds of area fertility surveys. While determining the levels and trends at the national context will be a continuing task, there is a greater need now to eRtt'asize the estimation of regional and areal differentials in fertility levels and trends, and to pinpoint which areas are lagging behind in the overall trend in fertility decline. In addition to regional or areal differentials, we need also to continually monitor fertility differentials by social groups to pinpoint which subgroups of the population are still exhibiting high fertility .Such information is essential for designing policies and programs that will effectively make a difference in reducing national fertility .This information is also important from the welfare point of view, to the extent that high fertility among some groups is shown to be associated with.negative effects on their socioeconomic and health welfare. Determinants. With respect to the determinants of fertility , several types of research can be suggested. F:irst, we have noted significant regional fertility differentials both in terms of levels and trends. Nevertheless, we have practically no studies in the more recent years examining the determinants of s~ch differ~ntials. Demographic surveys often collect only very limited information on the socioeconomic characteristics of the. respondents and practically none on the communities where the respondents reside. It will be extremely helpful in understandil1cg the current fertility declines observed in many regions to relate such demographic information to corresponding area-level information on socioeconomic factors likely to have affected such observed fertility trends, including the role of family planning inputs. One viable research approach might be to relate fertility trends to an analytical description of the soci~ economic changes that have occurred in the area combined with the~

Population

and Development

Research

317

knowledge of the timing of the availability and use of family planning inputs. At a more disaggregated level than regions, there seems to be a greater need for understanding fertility changes as they occur at the community or village level. At this level there seems to be a greater potential for combining institutional analysis with the usual microlevel statistical studies characteristic of past approaches ~ the Philippines. At this level, on needs to delineate the patterns qf social or~anization in the community and examine how these patterns influence individual economic and demographic decisions. Microlevel decisions are to be viewed not only as responses to opportunities and constraints prevailing not only at the household level but also at the community level. Several issues can be investigated within such a research framework, including the economic roles of children, the distributional consequences of reproductive behavior, the impact of demographic factors on the economic and social structure of the community , as well as the influence of such structure on fertility behavior . Consequences of fertility trends. What has often been a neglected area in Philippine social science research is the distributional impact of alternative fertility trends. Studies that shed light on this issue could provide indicators as to where (among what groups) efforts to implement fertility policy would redound to the greatest national advantage (taking distributional goals into consideration). These studies might also suggest the advisability of policies of other kinds that would attempt to compensate for the distributive outcomes of current reproductive patterns. Demographic impact of public interventions. As stated earlier , studies on the demographic impact of public interventions are just beginning. As such, theoretically and methodologically sound research designs have yet to be developed. Cumulative experience in this area, however, could not only advance the state of the arts but also begin to provide some guidelines for the possible restructuring of development strategies and programs to maXimize the attainment of traditional development goals and demographic objectives as well. Emphasis may be placed on evaluating those programs geared towards raising the levels of educational opportunities for both males and females, improving health and reducing mortality , promoting greater female participation, etc., that is, programs already desirable on traditional grounds and which already expend large amounts of

318

Alejandro

N. Herrin

the national budget. Knowledge of their possible additional demographic impact could lead to a possible modification programs to maximiZe their overall developmental impact, lead to policies and programs that will minimize their adverse demographic consequences.
IV. Patterns of Migration INTERNAL MIGRATION

indirect of such or could possible

The cumulation of migration studies sinc~ the 1960ts based upon census information andt more recently, upon the 1973 NDS has provided a broad picture of the size and pattern of population movements over the century . That substantial population movements have occurred in the Philippines is evidenced by the fact that in 1960 12.6 per cent of the population of all ages (representing some 3.4 mil~on persons) were living in a region different from the one in which they were born, and in 1970 this lifetime migration figure was 13.2 per. cent or 4.8 million persons. Even in the much shorter period between 1960 and 1970, 8.3 pet cent of the population 10 years old and over had moved to another region within the decade. This does not reflect population movements occurring within regions, and the multiple moves occurring between birth and the period of enumeration. Information on lifetime migration up to 1960 points to the predominance of long distanc.e movements involving either interprovincial or interregional transfers. This is characterized by relatively unidirectional, frontierward, male-dommated streams from rural origins to rural destinations. During the period from 1960 to 1970, while long distance flows continued to be a major component of the total migration pattern, the directions and social compositions have changed. Counterstreams to the dominant ones have appeared, and a centripetal pull to the metropolitan region has steadily grown, replacing to a degree the centrifugal drive to the frontiers. These patterns contributed to the predominance of female migrants so much in evidence in data for the recent past (Smith 1977). Data on in-migration and out-migration rates from the 1960 and 1970 censuses compiled by Perez (1978) reveal which regions gained and lost populations through such movements. Prior to the .1960ts, Cagayan Valley, Southern Tagalog including Manila, and

Population and Development

Research

319

Mindanao were the major receiving areas, while the llocos, Central Luzon, and .the Visayas were the major sending areas. Between 1960 and 1970, Southern Tagalog and two regions in Mindanao were the main receivers while two regions in the Visayas and Manila were the major out-migration areas. The out-migration from Manila represented in most part the suburban movement into what is now known as Metro Manila and the adjacent areas of Southern Tagalog and C~ntral Luzon. Determinants of Internal Migration Studies on the determinants of migration in the Philippines focused both on the determinants at the macro level (interprovincial or interregional migration) and on the detenninants of household migration. Examples of these types of studies will be briefly described here. Interprovincial de tennin an ts. Using the 1960 census data that cross-tabulated the population by province of birth and by province of residence, losa (1973) attempted to examine the relative effects on interprovincial migration rate (number of people born in province i but enumerated in province j) of selected provincial socioeconomic characteristics and of selected indicators of "movement-modifying" variables between origin and destination provinces. The lagged socioeconomic variables basically representing standards of living and employment opportunities were indexed by such factors as: (a) the difference in percentage of families using radios in 1948; (b) the difference in percentage of males employed respectively in extractive activities, in manufacturing industries, and in supportive services in 1939; and (c) the difference of unemployed males in 1948. On the other hand, the "movement-modifier" variables which are assumed to facilitate migration are measures of physical and social distance between provinces; the latter for example being indexed by major ethnic affiliation and by frontier staius of the province. Severe multicollinearity among the above variables limit the results of the study even after prior selection with a larger set of variables has been made. Nevertheless, one set of regression results (when all variables were included), indicated that among the significant variables the socioeconomic pull factors (radio usage, extractive. activities, manufacturing activities) were positively related to interprovincial flows. On the other hand, of the facilitator variables~ ethnic factor and frontier factor were positively related, while distance and regional similarity were negatively related to interprovincial flows. All var-

320

Alejandro

N .Herrin

iables together accounted for only 15 per cent of total variations. These results, interpreted broadly, are n