DEMOGRAPHY©
Volume 22, Number 3
August 1985
AN EVALUATION OF THE POPULATION AND DEVELOPMENT PROGRAM IN EGYPT J. Mayone Stycos International Population Program, Cornell University, Ithaca, New York 14853
Hussein Abdel-Aziz Sayed Department of Statistics, Faculty of Economics and Political Sciences, Cairo University and Population and Family Planning Board, Egypt
Roger Avery International Population Program, Cornell University, Ithaca, New York 14853
The Egyptian Population and Development Program (PDP) was one of the earliest to implement the ideology propounded at the Bucharest Population Conference concerning integration of family planning and development efforts. A major part of the national family planning effort, the PDP has made serious attempts to integrate grass-roots efforts at community development and family planning. Thus, the program stimulates family planning activity via a system of semi-volunteer outreach workers, and infuses the family planning message in other community projects. Local resources are mobilized and management capabilities upgraded in an effort to accelerate economic development, slow population growth, and improve community health and welfare. Small projects are stimulated by loans, while others are coordinated by improving leadership and organization. Initiated on an experimental basis in 1977, by 1978 the project was operating in over 800 villages, by 1979 in about 1,500, and by the end of 1980 in nearly 3,000 of Egypt's roughly 4,000 villages that include 56 percent of the country's population. By the end of 1983 over 1,100 village projects had been funded through loans and grants totalling nearly 2.5 million dollars. Of 584 projects carried out in 1983, the majority (72 percent) were aimed at improving such services as family planning, day care and youth centers, roads, and drainage. The remainder were largely devoted to stimulating small business projects in transportation, poultry raising, sewing, bee-keeping, carpentry, etc. (Population and Family Planning Board, 1984; Fattah, 1983). The program has attracted the attention of donor agencies, family planning professionals, and scholars, all of whom have been especially interested in its evaluation. The principal sponsoring agencies (initially, the United Nations Fund for Population Activity, and, more recently, USAID) have been concerned with evaluating the costs and benefits of a program that is more complex, and, in the beginning at least, more costly, than a traditional family planning program. In 1981, a detailed evaluation plan was designed by the Egyptian Population and Family Planning Board, in collaboration with the Cornell University International Population Program and with the financial assistance of the United Nations Fund for Population Activity. Although random assignment of villages to different treatment and control categories would have been desirable from a scientific point of view, the PDP is not unusual in setting up evaluation procedures after the fact. Indeed, two surveys conducted essentially for other purposes, a 1979rural fertility survey (RFS 1) and a 1980 Contraceptive Prevalence Survey (CPS) have already been used to evaluate program impact. Both surveys demonstrated a modest program impact but differed in their conclusions about its size, significance, and geographical location. I In 1982 a rural 431
432
DEMOGRAPHY, volume 22, number 3, August 1985
survey was carried out with program impact evaluation as its major goal (RFS2). In an effort to provide continuity with the 1979 survey, the same 85 villages were chosen, but 37 villages were added to improve representativeness, and a number of other improvements were introduced to assess better the program's impact. Had the program been overwhelmingly successful, these refinements might be of little substantive interest. In the early stages of a program, however, when impact may be small and incipient, more precise measurement and more exacting analysis may be required both to detect change and to chart its trajectory. With these general objectives in mind, the research reported here assesses improvements in: (1) the measurement of key variables, (2) sampling, (3) analytic techniques. MEASUREMENT
The earlier surveys used standard questions to elicit responses concerning knowledge of and attitudes toward fertility and contraception. To improve validity in these areas, the 1982RFS2 questionnaire was considerably expanded with respect to attitudes toward family size and contraception; and composite indices rather than single items were employed as dependent variables.i For example, traditional questions on the pill and IUD have been supplemented by questions of knowledge about how they are used; the Coombs scale and a set of other questions are used as measures of ideal family size; and various items reflecting attitudes toward additional children and toward family planning have been combined to form indices. The differences between the conventional measures commonly used in KAP analyses and those employed in the RFS2 are shown in the appendix. We will compare program and non-program areas using the traditional measures, and then using the indices. Comparing the first and third columns of Table 1, one is initially struck by the large differences between regions on all variables other than children ever born." The culture of Upper Egypt (the eight governorates south of Cairo) is generally considered to be more "traditional" than that of the delta region and the table provides striking evidence of how this affects the KAP variables. With the exception of children ever born (and, possibly, age at marriage) there are marked differences by region in the expected direction. Compared to these large differences, variation by PDP exposure within regions is small, though usually in the expected direction. Thus, in both regions PDP villagers are more likely to have heard of modern contraceptives, have more favorable attitudes toward birth control and small families, and, in Upper Egypt, are more likely to be practicing contraception (but no more likely to be practicing modern methods). Most of the attitude and behavioral differences fall short of statistical significance, however, and even where statistical significance is reached, the differences are small. One could conclude from the analysis of the conventional measures that the program has had a marked effect on knowledge, but has barely begun to affect attitudes or behavior. Table 2 substitutes indices for the simple variables, and introduces other variables less commonly used in KAP surveys. The use of scales to measure contraceptive use does not change the picture, nor does the refined measure offertility (DRAT) that takes age and age at marriage into consideration. The weighted knowledge score for five contraceptive methods does intensify the apparent program effect, with Fs of 47 in Lower and 65 in Upper Egypt, but the depth of knowledge scale for pill and IUD register little program effect. In the attitude area the situation is mixed, but the new measures here appear to be detecting the beginnings of program influence. On the negative side, the Coombs scale discriminates no better than preferred family size, and a scale measuring desire for no more children does no better than did the dichotomy "wants" or "does not
433
An Evaluation of the Population and Development Program in Egypt
want" more children. Further, the series of items reflecting advantages and disadvantages of small and large families (j) shows no differences between program and nonprogram areas, nor does the index concerning expected help from children (k). Two new measures, however, do produce selected differences: (1) a measure combining husband's desire (wife's report) for no more children with those of the wife (h) may be reflecting a modest program impact in Lower Egypt; and (2) a measure of husband-wife communication on family size and family planning (i) may be doing the same-though the sign is contrary to expectation in Upper Egypt. The various indices of attitudes toward family planning clearly strengthen the small differences seen initially in Lower Egypt, although the absolute differences remain small. Both the attitude and couple scales now show robust Fs, as do two scales reflecting the opinion of significant others (n) and the scale of potential users (0). Nearly all the other signs are in the direction that suggests the program is having its intended effect. In short, the use of more reliable scales and a more extended range of measurement has confirmed and thus strengthened the general conclusions derived from the simpler measures: knowledge of contraception has been most affected by the program and fertility least affected, with attitudes and contraceptive practice somewhere between. What has been added, moreover, are findings that the depth of knowledge about contraception has scarcely been affected by the program but that attitudes in Lower Egypt may be beginning to change in program areas, especially as regards contraception. The statistically significant relation between the current use index and program in Upper Egypt should not be overlooked, but, in light of the low absolute levels and the absence of attitude impact, should certainly be regarded circumspectly. We tum now to sampling variations, to see whether these conclusions will be maintained or altered after appropriate sampling weights are applied. Table I.-Selected conventional KAP variables: regional means and percentages and the absolute differences in means and percentages between women in PDP and non-PDP villages
5tr
Lower E
Behavior
Age at marrlage (x) Ever used blrth control (% ) Currently uSlng b i r t h control (% ) Currently usIng modern method (%)
Pregnant (% ) Live births (x and S. D. )
17.4 (3.9) 46 25 ,24 12 4.5 (3.1 )
Knowledge
pill (,) Know IUD (%) of eernods ltnown (x)d
Know
Number
Attitudes Wants no lIore children (%) Preferred number of children (xl
Approves of family planning
(%)
.05 4.1 _.2 c _.3 c -1.1
.l3 c
17.1 (3.5) 19 8 5 17 4.4 (3.3)
-.24 5.0** 3.8** 1.0 1.5 c -.11
98 86 2.2 (0.8)
6.6*** .15***
84 48 1.5 (1.0)
4.3** 12.7***
29 3.3 (1. 2) 85
.1 .11* 5.5**-
52 4.6 (2.1) 52
2.9 .28** .60
.1
.28***
a Meana are followed by standard deviations in parenth•• es. b Differences between PDP and non-PDP ••ana or percentage.. Chi-square testa were ueed for percentaged variables. F tests for others. Significance levels are indicated as follows: * •• 05 level: ** •. 01 level: **- •. 001 level. C
Difference is contrary to expectation.
d Regardless of probe. whether heard of pill, IUD, sterilization and condom.
Range· 0-4.
434
DEMOGRAPHY, volume 22, number 3, August 1985
Table 2.-Selected scales for KAP variables: regional means and absolute differences in means between women in PDP and non-PDP villages Lower
E',l~?t
Behavior a. b.
"'.
Ever use Current use DRAT
.71
1.0 1.2 0.7
.06 .04 .04 c
1. 31 .32 .64
0.7 O.B 0.5
4.22 4.13
1.7 1.2
.47···
2.B3 3.18
1.9 1.2
5.16 2.45 2.43 1.49 3.39 1. 31
1.7 1.0 0.7 0.8 1.0 0.5
6.43 1. 78 1.88 1. 33 2.64 1. 24
1.6 1.3 0.8 0.7 1.4 0.4
1.04 2.54 2.37 3.45
0.6 1.1 0.7 1.3
.59 1.48 1.8B 2.63
0.6 1.3 0.7 L5
1.88 .95
.05
.10*· .02 c
Knowledge d. e.
Weighted methods Depth of knowledge
.06
.65*·· .10-
Attitudes about family size
f. g. h.
t, j.
k.
Coo.bs Wants no IlOre children Couple wants no more Talk/think size Size attitudes Expected child help
-.12 .05
.10·· .06 .03 -.04 c
-.03
.13.02
-.13***c .06
-.OS*c
Attitudes toward falli ly planning
.. 1-
n. o.
Birth control attitude Couple attitude Perceived attitudes Potential use b
.10**.21*·.14**.32**·
a F tests were used to determine level of significance.
.00 .Oll .04 .04
(See Table 1, Footnote b.)
b Calculated for non-users only. C Difference is contrary to expectation.
SAMPLING VARIATIONS
All three surveys were carried out under the same auspices (the Egyptian Population and Family Planning Board), with considerable overlap in professional staffs. All surveys covered the same 17 rural governorates, involved large numbers of interviews with ever married women" up to age 50 (3,845, 5,313 and 6,123 respectively), and experienced nonresponse rates of less than five percent. In all cases households were selected at random from prelists specially prepared before fieldwork. There were, however, some important differences. Husbands of currently married women were interviewed in RFS2, although not included in the present analysis. Further, whereas the CPS used a self-weighted sample, the RFSI selected an equal number of villages (5) from each of the 17 governorates and an equal number of households (50) from each village. The RFS2 included the same villages in its sample, but tried to improve representativeness by: a) adding 37 villages to the original 85; b) stratifying the 37 villages in six categories for each of the two regions, depending on their PDP status (non-PDP, PDP less than three years, and PDP for three or more years), and whether the village was classified as a "satellite" or "mother't;" and c) selecting more households (75) from each of the mother villages, as well as 50 from each satellite. Stratifying the village sample by program status guaranteed an adequate number of program and nonprogram areas; but in post-facto designs, where experimental and control areas have not been randomly assigned, there is always the danger that the experimental groups may have been biased on characteristics that could enhance the
An Evaluation of the Population and Development Program in Egypt
435
apparent experimental effect. In the present instance, where fertility and contraceptive prevalence are among the key target variables, if women in the experimental villages tend to be better educated or from less agricultural areas, any subsequently measured differences in the dependent variables could not be attributed to the program. Indeed, the RFSI has been criticized for having" ...concentrated on larger villages. Accordingly, the findings probably underestimate the extent of poverty, mortality, and illiteracy, and overestimate contraceptive use in rural Egypt" (Sirageldin, 1983: 173). In fact, the opposite seems to have occurred. First, since the same number of villages were selected from each governorate in the RFSI sample, the smaller governorates (Damietta, Aswan, and Ismalia) were overrepresented and larger ones (Dakahlia, Sharkia, and Beheira) were underrepresented. More to the point, since the number of households selected in each village was the same, the probability of an individual being selected varied inversely with village size; thus, women from smaller villages had a disproportionate chance of being selected. Although the RFS2 is subject to similar problems of disproportionate governorate representation, an analysis of village size as determined in the 1976 Egyptian census does not reveal a serious bias in village selection." For example, according to census data, among 122 villages one would expect by chance 47 villages with 5,000 or more population, 50 with populations between 2,000 and 5,000, and 25 with less than 2,000 population. In fact the sample has 47, 48 and 27 in these categories. Within Upper and Lower Egypt the match is equally good." While the number of larger villages seems correctly represented in the sample, the number of interviews in the larger villages is less than would be suggested by their population. In both RFSs, moreover, PDP villagers show lower literacy, higher proportions in agriculture and lower SES than non-PDP villagers-a difference not apparent in the self-weighted CPS. Thus, it may be that both RFSs understate the intrinsic difference in KAP variables between program and nonprogram areas. To test for the overall effect of the sampling procedures, the 1982 survey was weighted according to village population data obtained from the 1976 census." Although initial analysis indicated that the impact of the weighting was small when the total rural area was analyzed, within regions the implications are substantial (Table 3). Although fertility is not affected by the weighting procedures, most of the other variables are. This can be seen by the means, most of which show increased differences between program and nonprogram areas when the weighted sample is used. As a way of summarizing the changes, in Lower Egypt nine of the fifteen comparisons assessing program-effect originally fell short of statistical significance at the .05 level in the unweighted sample. In the weighted sample, four of these nine became significant at the .01 or .001 level. In Upper Egypt, only fertility and two family size attitudes show no differences after weighting. In ten instances there are increases that move up the level of significance by at least one step, in five of these from nonsignificance to significance. 9 The superiority of PDP women in contraceptive knowledge is further intensified, and now even the depth of knowledge scale shows significant differences between program and nonprogram villages in both regions. Attitudes toward family planning, which earlier showed no greater favorability among program women in Upper Egypt, now do so-especially regarding the attitudes of the couple-and the attitude differences are much strengthened in Lower Egypt. On the whole, attitudes toward family size are the least affected.
436
DEMOGRAPHY, volume 22, number 3, August 1985 ANALYTIC METHODS
We have already noted that PDP villagers tend to be of lower socioeconomic status and education. This is only partly a product ofthe size of the village. When the weighted sample is used, differences by education are eliminated in Upper but are intensified in Lower Egypt. Economic differentials are reduced in both regions but remain highly significant (Table 4). The fact that PDP villages are not only smaller, but, independently ofthis, poorer and (in Lower Egypt) less educated has obvious implications for receptiveness to the program. A priori one would anticipate greater difficulties in the poorer and less educated communities-although the need is greatest there. To correct for this disparity, we carried out a series of multiple regressions progressively holding constant three sets of variables: (1) Demographic: age and number of living children. Fertility has not yet been responsive to PDP efforts, and, since higher parity women are more likely to initiate family planning, the relation between current use of contraception and fertility is positive. Since we view parity as an important potential determinant
Table 3.-Difference in means" between women in PDP and non-PDP villages, by region, weighted and unweighted samples. Lower Egypt
Upper Egypt
x'S
Differences in (PDP - non-PDP) Unwelghted Welghted
Dlfferences in x's (PDP - non- PDP) Unwelghted Welghted
Behavior a. b. c.
Ever use Current use DRAT
.06 .04 .04 c
.14** .17** .03 c
.05 .10** .02 c
.11*** .15*** .04 c
.47*** .06
.56*** .19***
.65*** .10*
.86*** .26
-.03 .13* .02 -.13***c .06 -.05*c
-.17* .23*** .06 -.ll***c .15* -.03 c
Knowledge d. e.
Weighted methods Depth of knowledge
Attitudes about family size
e, g. h. i.
j. k.
Coombs Wants more children Couple wants more Talk/thlnk size Size attitudes Expected child help
-.12 .05 .10** .Ob .03 -.04 c
-.14 .03 .09** .16*** .04 -.Ol c
Attitudes toward fallily planning 1. II.
n. o.
Birth control attitude Couple attitude Perceived attitudes Potential use b
.10*** .21 *** .14*** .32***
.18*** .39*** .28*** .46***
.00 .08 .04 .04
.06* .20*** .06* .14*
a hsterisks refer to the statistical significance (Ileasured by F-tests) between PDP and non-PDP, within each category of weighting and region. Levels of significance are as noted in Table 1. b Calculated for non-users only. c Difference is contrary to expectation.
437
An Evaluation of the Population and Development Program in Egypt Table 4.-Education and economic status: regional means and differences" between women, in PDP and non-PDP villages, by region, weighted and unweighted Upper Egypt
Lower Egypt
D~fferences
D~fferences
Mean
(PDP - non-PDP)
Mean
(PDP
- non-PDP)
Education scale b Unweighted Weighted
1.83 1. 78
.22*** .33***
1. 51 1.51
7.84 7.47
1.11 *** .63***
5.34 5.54
.11* .02
Economic status C Unweighted Weighted
1.17*** .70***
a Asterisks refer to the statistical significance (measured by F-tests) between PDP and non-PDP. within each category of weighting and region. Levels of significance are as noted in Table 1. b A six point index ranging from illiterate and unschooled (1) to seven or more years of schooling (6). Two-thirds of the women are in the lowest category. c The sum of 26 items scored as 0 or 1 (possession or non-possession) of furniture (such as tables and beds). equipment (such as sewing machines and clocks), and services (such as water and electricity) •
of the other KAP variables, it is included with age as a control when assessing PDP impact. 10 (2) Socioeconomic: scales of education and economic status (see Table 4, notes a and b). (3) Community modernization: measured by computing mean values of the two previous scales for each village.'! Since there was little correlation between demographic variables and the PDP, little effect resulted from holding the first set of variables constant. With each of the next two sets of controls, however, the correlation between the PDP and the criterion variables increased markedly, and weighting further compounded the effect. 12 For example, in Lower Egypt the unstandardized regression coefficient (.04) for the PDP dummy variable and current use of contraception is both weak and statistically insignificant in the unweighted sample. Controlling for the socioeconomic and demographic variables raised the coefficient to .10, and adding community modernization raised it to .18. Weighting for population size further increased all of these coefficients, so that the final regression coefficient for the weighted sample with full controls is .32. Coefficients for knowledge in depth rose from .06 to .42 in Lower and from .10 to .37 in Upper Egypt over the same range of weights and controls. A number of variables for which the PDP initially appeared to have no effect now show substantial relations and high significance levels. In order to summarize this analysis Table 5 highlights the extremes, that is, the coefficients for program effect when the data are unweighted and uncontrolled, when they are unweighted but with all three sets of controls, and when they are both weighted and controlled. With few exceptions (fertility and one or two indices of family size attitudes) all signs are in the expected direction and coefficients become
438
DEMOGRAPHY, volume 22, number 3, August 1985
Table 5.-Regression coefficients for PDP effect" with and without weights and controls", by region Lower Egypt Unweighted No Weighted controls Controls controls
No controls Controls
Weighted controls
Behavior
a. b. c.
Ever use Current use DRATd
.06 .04 .04 c
.16**.18**· .06· c
.27*·· .32**.0Sc
.05 .10*'*' .02 c
.47**.06
.70**.22**-
.79**.42**'*
.65**.10·
-.27*·.05 .08*-
-.33**.07 .10·· .24**.12· -.Olc
.13**.19**.U3 c
.13**· .19**· .04 c
1.02**.35**-
1.05**.37**·
-.03 -.20·· .13· .20**'*' .07· .02 -.13*·*c -.06· c .24**.06 -.OS·c -.03 c
-.23***
Knowledge d. e.
Weighted methods Depth of knowledge
Attitudes about falli ly size f. g. h. i.
j. k.
Coombs Wants no more children Couple wants lIIore Talk/think size Size attitudes Expected child help
- .12 .05 .10*· .06 .03 -.04 c
.13···
.08" .01
.24**.07· -.09·· c
.22**-
-.02 c
Attitudes toward family planning
.. 1-
n.
o.
Birth control attitude Couple attitude Perceived attitudes Potential usee
.10···
.21*·.14**.32**-
.14**-
.28···
.20*·.]8**·
.24··· .54*·.35··· .60*·-
.00 .08 .04 .04
.09*'*'
.24*·· .08·· .14·
.09··
.24··· .06 .15·
a Asterisks refer to statistical significance, as measured by T-tests, of the regression coefficient for PDP effect. Levels of significance are as noted in Table 1. b Controls include all demographic, individual socio-economic, and community level variables. C Relation
is contrary to expectation.
d Does not include living children among controls. e Regression estimated for non-users only.
highly significant. With proper weighting and controls, the apparent program performance is substantially increased. In order better to compare the degree of impact among the various sets of dependent variables and between regions, we calculated the ratio of each coefficient to the standard deviation of the variable, and averaged these ratios for each block of variables. Figure 1 shows that all the substantive areas appear to register a program effect; that the effect on knowledge was strong in both areas; that the effect on contraceptive attitudes was equally strong in Lower but relatively weak in Upper Egypt; that impact on family size attitudes was weak but in the expected direction in both regions; and that an effect on contraceptive practice occurred in both regions. Thus far, the results have been presented in terms of composite indices of dependent variables that may not be easy to interpret for program purposes. The various scales may reveal differences that are judged to be "substantial" or "significant" or "more valid" to the analyst, but these may be too abstract to be meaningful to the planner. For this reason we have estimated the impact of the PDP program on contraceptive prevalence using conventional percentage expressions, maintaining the same weights and control variables as used earlier. Utilizing Multiple Classification Analysis 13 to hold constant the same set of six demographic, socioeconomic, and community variables, we find that in Lower Egypt, prevalence differs by
439
An Evaluation of the Population and Development Program in Egypt
over 8 percentage points between PDP and non-PDP areas when weights and controls are applied (26.7 versus 18.4 percent) but only -0.2 (i.e., not at all) when weights and controls are not used. In Upper Egypt, the controlled and weighted prevalence data show a difference of 6 percentage points between PDP and non-PDP areas (l0.4 and 4.4 percent), contrasted with 3.8 percentage points for the uncontrolled and unweighted data (9.2 versus 5.4 percent). In short, with appropriate weights and controls, prevalence rates in Lower Egypt are estimated to be 45 percent higher in PDP than in non-PDP areas; and in Upper Egypt are over twice those in non-PDP areas, albeit at a very low absolute level. With appropriate controls the program increased knowledge by about one third of a method in Lower and one half a method in Upper Egypt. Since the women in the former knew only about two methods and in the latter only about one, the PDP can be said to have increased women's options about family planning in a very concrete way. On the other hand, even current measures of fertility such as pregnancy status and open interval (in years) show no program impact. 14 CONCLUSIONS
Using conventional measures without controls and without weighting for size of village suggested clear and substantial program effects only in the area of contraceptive knowledge. The use of supplementary and more refined measures suggested
Figure 1. Ratios of Regression Coefficients for Program Effect to Standard Deviations, for Groups of Dependent Variables, by Region Mean Ratio* .5 Upper Lower
c--
Lower
.4
.3
Lower Upper
.2
Upper Lower
[lo
.1
.0
Contraceptive Knowledge (variables d,e)
Birth Control Attitudes (variables 1-0)
Fertility Attitudes, f-k (variables f-k)
Contraceptive Behavior (variables a.b)
*The columns represent the mean value of the ratios for those variables within the knowledge, attitude, and behavioral sets.
440
DEMOGRAPHY, volume 22, number 3, August 1985
positive program impacts on contraceptive attitudes in Lower Egypt and contraceptive use in Upper Egypt, but in general weakened the suggestion of any impact on attitudes toward family size. Weighting the sample, however, produced apparent impacts in all areas short offertility itself, and, with the exception of attitudes toward family size, differences between program and non-program women were highly significant and reasonably large. Finally, adding various groups of individual and community socioeconomic controls further intensified the effects of the program, to the extent that all areas in both regions show significant program effects. Table 6 summarizes these findings. In short, the impression received from the initial findings using crude measures, an unweighted sample, and gross comparisons without socioeconomic controls was highly misleading, in part because PDP villages tended to be smaller and less developed, and partly because batteries of indices are more reliable indicators than single items. Since it is often the case that statistical controls weaken the apparent relations between two variables of interest, it is instructive to note an instance in which the relation is greatly strengthened. Beyond the methodological lessons that can be learned from this exercise, the substantive conclusions are of some importance. (a) The Egyptian Population and Development Program does appear to be having an impact on most of the cognitive, attitudinal, and behavioral variables believed crucial for fertility change, and this impact is larger than has been suggested by previous studies. (b) The pattern of prefertility changes seems quite reasonable, after analysis. Knowledge differences between program and nonprogram areas are large and constitute a plausible first step in the chain of changes leading to fertility reduction. Further, change in attitudes toward fertility control has occurred most where it was initially most positive, in Lower Egypt, while Upper Egypt with a more traditional culture and initially more negative attitudes, has lagged, though it also registers program effects after appropriate controls. Attitudes toward family size, not surprisingly, have been most resistant to change or constitute the most difficult set to measure adequately. Such attitudes may be more basic than the others and may be expected to respond more to the efforts at community development than to direct family planning education and services. The unresponsiveness of fertility to these cognitive, attitudinal, and behavioral changes is puzzling. It might be explained by inefficient or erratic use of contraception that could be characteristic of early efforts at fertility control among rural Egyptians, or by a disproportionate tendency on the part of high fertility women to respond to the program. These issues will be subsequently investigated. In the
Table 6.-Has the program affected KAP variables in Upper Egypt, Lower Egypt, both, or neither? Knowledge Extensl ve IntenSIve
Conventional quest i ons and measures
Attitude toward ContraceptIon FamIly
x
Both
Lower
(Both)
Use of contraception
Fertility
Upper
Neither
Reflned measures
Both
Nei ther
Upper
Neither
Weighted sample
Both
Both
Both
(Both)
Both
Neither
StatIstical controls
Both
Both
Lower Upper
(Both)
Both
Neither
Note:
x
~
nOt tested:
( )
= weak
(Upper)
SIz.e
relation:
= strong relation
441
An Evaluation of the Population and Development Program in Egypt
meantime, it is evident that certain preconditions offertility decline-knowledge and acceptance of modern means of control-have been responsive to program efforts. Fertility decline should not be far behind. Appendix Conventional KAP measures and those used in RFS2. Concept Behavior
Conventional Measures
RFS2 Indices
Ever use of contraception Current use of contraception Live births
b. Current use scale
Contraceptive Knowledge
Methods volunteered or recognized
d. Weighted number known e. Depth of knowledge index f. Coombs scale
Attitudes toward family size
Ideal number of children Desire for more children
g. Wants no more index h. Couple wants no more index i. Talk/think size j. Size attitudes k. Expected child help
Attitudes toward family planning
a. Ever use scale
c. DRAT
I. Birth control attitude m. Couple attitude n. Perceived attitudes of significant others o. Potential use of birth control
Definition of Indices: a. Three point scale in which never use of any method = I, ever use of a traditional method = 2, and ever use of a modern method = 3. b. Four point scale from never use of any method (0) to current use of modern methods (3), with past use scored as I and current use of a traditional method as 2. c. This measure has been adapted from B. Boulier and M. Rosenzweig (1978). DRAT as used here is the ratio of actual to expected fertility between marriage and time of interview, in the absence offamily planning (m = 0), as determined from the Coale and Trussell (1974) model offertility. The measure differs from that employed by Boulier and Rosenzweig, since we did not include date of initiation of contraception in the model. d. Ten point scale summing knowledge of five contraceptive methods: pill, IUD, condom, sterilization, and breastfeeding. Each method was scored as 0 for no knowledge, I for recognition after probe, and 2 for spontaneous mention. e. Sum of correct responses to several questions about the pill and IUD; e.g., how often pill should be taken and how to check if an IUD is in place. Range = 2-6.
442
DEMOGRAPHY, volume 22, number 3, August 1985
f. Scale devised by Coombs and Coombs to measure famly size preferences by means of forced choices between pairs. (Coombs, Coombs, and McClelland, 1975). g. A four point scale ranging from wants no more (3) to wants 3 or more additional children (0). h. A three point scale ranging from 1 = both husband and wife want more children (according to wife's report) to 3 = neither wants more. L A three point scale ranging from (1) have never thought about the number of children wanted to (3) have both thought about and discussed it with husband. j. The sum of four questions requiring a positive or negative response: "Who is more fortunate, a child raised in a small or a large family? Some say the woman with a small number of children is not happy at all, others say she is fortunatewhat do you think? Of two men whose living conditions are the same, one has eight and the other three children. Who will be respected more by the people? Do you think that many children means more security to the parents at old age, or that 2 or 3 could provide the same security?" Range = 0-4 with the higher value favoring the small family. NOTES After analyzing 1979 survey data, Kelley et al. (1982) concluded that "current and ever use of contraceptive methods are more prevalent in the PDP villages" but that "the impacts of the PDP have been largely in Lower Egypt." Based on the CPS conducted only one year later, Khalifa et al. (1982: 159) conclude that the PDP "has had some influence on family planning behavior and attitudes in Upper and Lower Egypt;" and that impact on current and ever use was "substantial" in Upper and "significant" in Lower Egypt. Using the same survey data, Stycos et al. (1982) termed the PDP impact "noticeable;" but claimed that in Lower Egypt any differences' between program and non-program areas were "smaller and often not large enough to reach statistical significance." 2 One analysis of the CPS also employed scales as dependent variables, though less extensively than the present study. See Stycos et al. (1982). 3 Although the number of live births is the same in both regions, Lower Egypt shows more living children (3.5) than Upper Egypt (3.1) because of lower infant mortality. 4 Only currently married women are included in the present analysis. 5 A mother village is usually a larger and centrally located village around which approximately five satellite villages are located. The mother village usually has more facilities and acts as an administrative center. The distribution of villages and currently married women in the various strata are as follows: I
Villages Mother Villages No PDP PDP 1-2 yrs. PDP 3+ yrs. Satellite Villages No PDP PDP 1-2 yrs. PDP 3+ yrs.
Currently married women
Lower
Upper
Lower
Upper
6 7 5
4 2 3
347 412 290
190 106 181
9 22 19
16 15 14
385 980 726
715 612 591
6 In RFS2 the general problem was mitigated by the stratification procedures that improved representation of mother versus satellite villages, and PDP versus non-PDP villages. 7 Village size correlates no higher than .10 with such variables as knowledge and use of contraception as contrasted with correlations of over .30 with socioeconomic status. Moreover, in the regressions shown in Table 7, its relation was entirely explained by socioeconomic status and it was consequently dropped from the model. S To obtain the appropriate weight the population of each mother village (according to the 1976census) was divided by 75 and each satellite village by 50. Then, in order to correct for governorate, the weights were multiplied by the inverse of the probability of the villages being selected in each governorate (i.e., number of villages in the governorate divided by the number of villages sampled). These final weights
An Evaluation of the Population and Development Program in Egypt
443
were normalized to obtain a mean value of approximately one. To test for the effect of differential population growth between 1976 and the time of the survey, 1964-1976 village growth rates were used to estimate village population sizes in 1982. This procedure led to weights virtually identical with those derived from using the 1976 data. 9 Although the significance levels used assume a simple random sample, the two stage clustering used in RFS2 violates the assumption. This could have a substantial effect in reducing the significance levels of regression coefficients and differences between means. The effect of clustering is suggested by the impact on coefficients after weighting and controlling for community level variables. Thus the significance levels reported here should be used only as guides to order of magnitude changes. Nevertheless, the coefficients are not affected by this problem and in most cases their confidence intervals could be doubled and still not include zero (Kish et al., 1976). 10 Although highly correlated with parity, age was included since it could (and often does) have an independent effect, e.g., through generational differences. 11 Strictly speaking, since the sample was drawn from selected strata within the village, the mean refers only to those areas of the village sampled. 12 This was true for Lower Egypt. In Upper Egypt the controls were sufficient and weighting did not increase the coefficients. Among both weighted and unweighted samples the coefficients first rise with the introduction of the individual level socioeconomic controls, and again as the community measures are added. 13 Although MCA should not, strictly speaking, be used for binary variables such as prevalence, the results were very similar to those obtained using the current use scale. MCA has the advantage of being directly comparable with our multiple regression results and of providing coefficients that can be readily interpreted for planning. 14 It may take time for contraceptive practice to become efficient enough to counteract the tendency for higher fertility women to be the first to adopt it. Efforts to relate duration of the PDP to KAP variables however, suggest that the relation is not linear-at least within the narrow time scope of the PDP at the time of the survey. (Duration categories ranged only from one year to a maximum of three or more, with four years the mode for the oldest programs. In Upper Egypt, mean program duration was 2.8 years, in Lower Egypt 2.7.) Moreover, preliminary work on this question suggests that in Upper Egypt the impact is sluggish but accelerates rapidly as time passes; while in Lower Egypt the impact seems very rapid but with little subsequent gain, a fact which could help to account for differences in original impact measured by earlier and later surveys. Regularity of use was not covered in our survey, but in the CPS at the end of two years only about one-half of pill acceptors would still be users, with strong regional differences in the usual direction (Population and Family Planning Board and Westinghouse Health Systems, 1982). REFERENCES Boulier, B. and M. Rosenzweig. 1978. Age, biological factors, and socioeconomic determinants of fertility: A new measure of cumulative fertility for use in the socioeconomic analysis of family size. Demography 15:487-498. Coale, A. J. and J. Trussell. 1974. Model fertility schedules: variations in the age structure of childbearing in human populations. Population Index 40:185-258. Coombs, C. H., L. C. Coombs, and G. H. McClelland. 1975. Preference scales for number and sex of children. Population Studies 29:587-611. Fattah, A.A. 1983. Communication for Population and Development Programmes: An Egyptian Experience. Paris: UNESCO Series in Population Communication No.9. Kelley A. C., A. M. Khalifa, and M. N. EI-Khorazaty. 1982. Population and Development in Rural Egypt. Durham, N.C.:Duke University Press. Khalifa, A. M., H. A. A. H. Sayed, M. N. EI-Khorazaty and A. Way. 1982. Family Planning in Rural Egypt 1980. Population and Family Planning Board and Westinghouse Health Systems: Egypt 1980 Summary Report. Contraceptive Prevalence Survey. Kish, L., R. M. Groves, and K. P. Krotki. 1976. Sampling Errors for Fertility Surveys. World Fertility Survey Occasional Papers, #17. Population and Family Planning Board. 1984. 1983 Annual Report of Activities. Cairo. Population and Family Planning Board and Westinghouse Health Systems. 1982. Contraceptive Prevalence Survey Egypt 1980. Sirageldin, I. 1983. Book review of Population and Development in Rural Egypt. Population and Development Review 19:173. Stycos, J. Mayone, A. Bindary, R. C. Avery, A. M. Khalifa, H. A. A. Sayed and A. Way. 1982. Contraception and community in Egypt: A preliminary evaluation of the population development mix. Studies in Family Planning 13:12.