2.2.1. Indicators and their derivation
The available indicators collected and presented in Table 2-1 provide five interrelated dimensions of social and economic development: urbanization, economy, communications, education, and health2. The proportion of population living in urban areas (URBAN) as compiled by the National Census and Statistics Office (1983) for the 100 per cent census coverage for 1980 and 1970, and for 1960 (National Census and Statistics Office, 1973)3, is taken to indicate urbanization.
Average family income (INCOME), agricultural output per worker in pesos, adjusted for inflation (AGRICL) and percentage of households using electricity for light (ELECT) are proxies of economy4. For 1980, the average annual family income (INCOME) from the 1985 Family Income and Expenditure Survey (FTES) is considered as there was no similar figure for 1980. The immediately preceding FIES was conducted in 1975. However, Mangahas (1979) and Mangahas et al. (1977) found the 1975 FTES to be seriously defective compared with the 1971 FIES and regular reports from the National Income Accounts. Thus, the 1975 FEES data set has been disregarded (see also National Economic and Development Authority, 1987). The patterns of provincial variation in average annual income then are assumed to be similar in 1980 and 1985. This assumption may be valid as the country continued to experience unprecedented economic decline and political uncertainty in the early 1980s and this probably affected all provinces (Hill, 1986, 1988). As no comparable provincial FIES Figures for 1960 and 1970 are available, the estimated agricultural output per worker in pesos for 1971 and 1960, available in Pemia et al. (1983), was adopted to indicate the 1970 and 1960 economy. The percentage of households using electricity for light for 1980, 1970 and 1960 yielded by the censuses in the same years is considered. The manner of collecting this information for all census years is similar and therefore this measure is comparable over time.
Communications are also represented by road density or length of roads over total land area (ROAD) estimated by using Arriaga’s equation (1967:102):
ROAD-1,000(Kilometres of road/Area(Rm2))
and number of registered motor vehicles per 10,000 population (MOTOR). Again the available published data for these indicators in 1982 and 1964 do not refer to the reference date of analysis (1980 and 1960).
2The United Nations General Assembly in November 1979 adopted resolution 34/58 concerning health as an integral part of development (WHO, 1981a: 16).
3Definition of urban areas for 1960 is different from that followed for 1980 and 1970, see Appendix 2.2, Volume 2.
4If electricity is available in the area, there is at least one household with a television. In the raral areas, the household with a television is likely to be filled with neighbours during resting hours to watch television. In such instances. ELECT is also a proxy of commun i cations.
However, while the two and four year differences may have created some changes in the absolute values of the variables, they are unlikely to have changed much in relative standing, thus may still be useful as indices. Between 1980 and 1982, for the country as a whole the length of road network in kilometres increased, while the number of motor vehicles registered decreased, but by a mere two per cent (National Census and Statistics Office: Journal o f Population Statistics, 1984:ix-xix). It is possible that there was a similar magnitude of change in the early 1960s.
Three measures of education are examined from the censuses and the same years: percentage literate among persons 10 years old and over (LITERACY)5, percentage of persons with at least elementary education (ELEM), and percentage of persons with at least high school (HSCHOOL). Having all three education indicators at the provincial level, it may be helpful to examine the usefulness of each in the present analysis. In the Philippines, there is a sizable proportion proceeding to high school, around 20 per cent As shown in Chapter 9, Subsection 9.2.2, a four-level categorization of women’s educational attainment (primary and below, elementary, high school and college and over)6 revealed that child mortality rate associated with high school category is one-half of that referring to elementary category. It is worth pursuing whether HSCHOOL is as useful as LITERACY, or more useful, in provincial comparative analysis.
Several available health indicators are examined. For all reference years of analysis, four indicators are derived: percentage of households with a safe drinking water supply (WATER), percentage of households with sanitary toilets (TOILET), population per hospital bed (HOSPBED), and index of accessibility to hospitals (IAH), the derivation of which is presented below. The level of child malnutrition or percentage of children aged 0-6 years malnourished to the second and third degree (MAL) relates only to 1976. The population per rural health unit (RHU), population per physician (DOCTOR), population per nurse (NURSE), and population per midwife (MIDWIFE) refer to 1972.
The 1980, 1970 and 1960 censuses yield data on the number of households with safe drinking water supply and sanitary toilets. Calculations of pertinent percentages conform with the Ministry of Health 1978 and 1981 National Health Surveys’ definition of safe water supply which includes the waterworks system tap water, public well, public faucet, private deep well, improved dug well and improved spring; and of sanitary toilets which refer to all flush types with septic tank, direct to sewerage system, water seal, hand-flushed and pit only.
With available data on number of hospitals (public and private) per province, hospital density (HOSPD) could be calculated following Arriaga’s equation (1967: 102) as:
5As the 1980 and 1960 census figures refer to persons 10 years old and over, the 1970 figures are derived similarly for
comparability since the published values refer to persons 6 years old and over. 6Chapter 8, Subsection 8.5.1 presents a more detailed discussion of this issue.
HOSPD=l,0 0 0 (number o f h o s p it a ls /A r e a (Jan2))
When HOSPD is multiplied by ROAD, an index of accessibility to hospital services (1AH) is calculated in accordance with Arriaga’s model (1967: 102). The higher the index, the more accessible are the hospitals.
The percentage of children aged 0-6 years malnourished (second and third degree) is drawn from the individual regional development information for the period 1972-1982, one of the projects of the National Economic Development Authority to compile statistics for regional planning and policy-making. Unfortunately, the nutrition statistics down to the provincial level before and around 1980 are available for one year only, 1976. These data were based on the program called ‘Operation 77m6ang(weighing)’ to weigh every preschool child in the country, inaugurated in 1974 by the National Nutrition Council, which is the agency charged with co-ordinating the government’s nutrition programs and preparing a national nutrition plan.
The ratios of rural health units, physicians, nurses, and midwives to population are calculated by using available figures at the provincial level for such health services in 1972 and the provincial census population in 1970. Since the reference date of analysis is 1970, the 1972 health services figures are assumed to be the same as in 1970.
2.2.2. Data problems
In 1980, the nation comprised 73 provinces as noted above; in 1960, there were only 55 provinces. Between these years several provinces were subdivided by various laws and decrees (see Appendix 2.3, Volume 2) to create new provinces. In most cases, relevant data for the newly created provinces are not available, as the compiled statistics still refer to the old provinces; such changes have aggravated the problem of scarce data at the provincial level since 1960.
Moreover, published data based on censuses vary by base of estimation. Published socioeconomic and housing statistics for the 1960 and 1970 censuses are based on 100 per cent census coverage while those for 1980 are estimates based on a 20 per cent sample. This difference in the base of estimation may create some problems of comparability. Estimates based on the 20 per cent sample are more likely to be affected by sampling variability than are those based on the total coverage; the error may be more substantial in small provinces. In addition, there is the possibility of bias in sample selection. A clear example is Batanes which is the smallest province in land area (209.3 square kilometres) and population (10,000-12,000) for the 1960-80 period. The proportion of Batanes households using electricity for light in 1980 based on the 20 per cent sample is three per cent; in 1970, the corresponding percentage based on 100 per cent coverage is 10 per cent. This sudden decrease during the 1970-80 interval is very unlikely. Much, if not all of the
decline may be attributed to bias in the 20 per cent sample: it appears that households using electricity for light captured in the 20 per cent sample for this province are not representative of the total households with electricity in the whole province.
In addition, published raw data may contain errors of various types. One example is the possibility of overestimation of the level of child malnutrition. Overreporting of malnourished children may result from two causes: one is the deliberate selection of malnourished children through the very nature of Operation Timbang as a screening operation designed to identify the malnourished for direct individual intervention (Demographic Research and Development Foundation, 1985: 99); the other is that the local staff believed that the Operation Timbang results would be used to determine the allocation of food assistance, and thus for the weighing operation, they selected areas where malnourished children prevailed (Lim, 1984 as cited by Quisumbing, 1985:43.). Because of the possibility of errors, it is essential to treat published data with some reservation or to subject them to appropriate reliability tests.
Furthermore, the above-mentioned provincial compositional changes, differences in the estimation base of pertinent census data, and possibility of errors of many sorts prevent outright conclusions, particularly those connected with time trends and differential patterns. It is therefore important to evaluate the data quality further before they can be used meaningfully in the analysis of provincial variations by socioeconomic and health related conditions.
2.2.3. Methods of quantitative and qualitative appraisals of the indicators and of differential analysis by province
Quantitative appraisal. To determine whether the available socioeconomic and health statistics provide meaningful patterns of provincial variations quantitatively, two steps are followed. One is to examine whether the variables are nearly normally distributed. Recall that the derived indicators from such statistics provide five interrelated dimensions of social and economic development: urbanization, economy, communication, education and health. The first question to be asked then is ‘Are they really linearly interrelated?’. However, to test for their linear relationships requires that their observed distributions should at least approximate a normal distribution. Hence, a test for normal distribution is performed using histograms and summary statistics, such as the mean, standard deviation and skewness. If the observed distribution of a variable is approximately normal, its histogram follows a bell-shaped distribution, 95 per cent of all cases are within two standard deviations of the mean and skewness is close to zero. Observed distributions of the variables in question that are nearly normal indicate the usefulness of such variables in the analyses of linear relationships. Those considerably deviating from approximate normality will be log-transformed. If the distributions of such log-transformed variables still deviate greatly from nearly normal distribution, then such variables are not useful in the analyses of linear relationships and will be dropped.
Results of the first step are inputs to the other step, which is a zero-order correlation analysis, to relate each of the proxies of the five dimensions of social and economic development mentioned earlier for each of the reference dates of analysis: 1980, 1970 and I960.7 If social and economic development is defined as a combination of these five dimensions, then according to Cutright and Kelly (1981:147),
a high degree o f intercorrelation exists primarily because the process o f national development tends to be unified rather than discrete. For example, a nation is not likely to have a highly literate population and at the same time have a low average life expectancy or a low standard o f living.
This line of reasoning is followed in the assessment of applicability of the indicators listed in Table 2-1. Any indicator that has a significant correlation with another variable under consideration and substantively reflects a given situation (based on the qualitative analysis) is regarded as a useful indicator in the determinants of mortality analysis in Chapter 7 and in the establishment of a brief provincial profile in this chapter.
Qualitative appraisal. WHO (1981a, 1981b) provides a list of indicators (Table 2-2) to be used for global, regional, national and sub-national monitoring and evaluating progress towards the attainment of health for all by the year 2000 (through primary health care) as declared at the 1978 International Conference in Alma-Ata, USSR (WHO, 1978). In WHO’s (1981b:12; 1981c:18-19) guidelines for health program evaluation, indicators are defined as
variables which help to measure changes ... ideal indicators should be valid___ that is, they should actually
measure what they are supposed to measure; they should be objective___ the answer should be the same if
measured by different people in similar circumstances; they should be sensitive___ that is, they should be
sensitive to the changes in the situation; and they should be specific___ that is, they should reflect changes
only in the situation concerned.
Assessing further the socioeconomic and health-related indicators in Table 2-1 in accordance with the WHO’s health indicators and guidelines supplements the above quantitative appraisal.
Table 2-2: WHO list of indicators to be used for monitoring and evaluating progress towards the attainment of health for all