Service vs survey statistics: an evaluation of contraceptive use in Pakistan ([dagger]).

Sultan, Mehboob ; Ali, Syed Mubashir

INTRODUCTION

Contraceptive use rates estimated from service statistics and contraceptive prevalence surveys often disagree particularly in larger populations. This disagreement is more pronounced in traditional societies where under-reporting of contraceptive use is a common feature not only because of methodological problems in sample surveys and deficient reporting and recording systems of family planning programmes but also because of socio-psychological reasons including social taboos, personal inhibition, shyness, lack of education and lack of openness to foreign ideas and mass disapproval by the society for cultural and religious reasons.

The disparity in contraceptive use rates based on these two sources of information is also acknowledged in countries with strong family planning programmes and efficient reporting and recording systems. Indonesia, which is often cited as a prime example of operating a successful family planning programme and is also quoted to be the one with "very strong reporting, recording and research component" [ESCAP (1989), p. 4] is not an exception. Comparative studies undertaken there reveal that service statistics estimates exceed that of surveys by 24 percent for IUD, 28 percent for pill and 110 percent for Condom [Streatfield (1985), p. 45]. Similar inconsistencies were found in India and Bangladesh [Koening et al. (1984) and Ahmed et al. (1987)].

In Pakistan wide differences have been observed in contraceptive use rates based on service statistics and those revealed by surveys [Syed (1981); Rukanuddin et al. (1985) and Sultan 0987)]. The deficient recording and reporting system of contraceptive use, the inappropriateness of the usage assumptions, continuation rates of different methods and the lack of necessary information on method effectiveness and use effectiveness cast doubt on the estimates of service statistics. Table 1 shows reported contraceptive use rates in the 1975 Pakistan Fertility Survey (PFS) and 1984-85 Pakistan Contraceptive Prevalence Survey (PCPS) together with estimated use rates based on service statistics for the corresponding reference periods. (See Table 1).

The above table indicates gross disparities between the estimates of the two sources both in 1975 and 1984. The difference is enormous for pill and IUD. While the 1975 service statistics estimates are higher for all methods, the 1984 PCPS results show higher rates for sterilization and injectables.

Service and survey statistics could both be doubted for accuracy. It is also misleading to accept either of the two sources as a standard while evaluating the other. In such a situation, the need is to explore some other method which may provide more reliable estimates. The basic aim of this paper is to examine the possibilities of an indirect technique for arriving at a more probable rate of contraceptive use independent of service and survey statistics.

DATA AND THEIR LIMITATIONS

For estimation of contraceptive use in 1975 and 1984, the Population Welfare Programme adjusted service statistics have been used. For 1975 the method of Couple Years Protection (CYP) was used whereas for 1984 Component Projection Approach Ill (CPA) was applied. Service statistics for conventional contraceptives and the pill represent sales reported by service outlets and not actual use. The denominators used for estimating use rates are based on the projected number of married women of reproductive age. Any under or over assessment of this number can lead to inaccurate estimation of use rates.

Potential fertility used in this analysis refers to the average of PGE longitudinal and Chandra and Deming method estimates. For indirect estimation of contraceptive use the PFS fertility schedule has been used. Other fertility schedules such as NIS (1968-69), the PLM (1979-80) and the PCPS (1984-85) were not up to the standard identified by Coale and Trussell (1978) and, hence, are excluded from the present analysis.

METHODOLOGY

Using the marital age-specific fertility rates (MASFER) of the PFS, indirect estimates of contraceptive use are calculated for Pakistan and compared with estimates from service statistics and surveys. These estimates are derived on the basis of the "m" parameter used in Coale's model of marital fertility (1971).

Coale's model of marital fertility can be explained by the following equation:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)

Where

r(a) = Observed marital fertility at age (a);

n(a) = Natural fertility at age (a);

M = Scale factor indicating the level of fertility with respect to average natural fertility;

e = Base of natural logarithm;

m = Degree of deliberate marital fertility control in a population; and

v(a) = Logarithmic departure from natural fertility.

The functions n(a) and v(a) given below were derived from empirical data. The values of n(a) were obtained by averaging 10 of the 13 fertility schedules designated by Henry (1961) as natural. The values of v(a) were obtained through employing 43 marital fertility schedules. (See Table 2).

Coale and Trussell (1978) developed a standard technique for finding the two parameters "M" and "m" used in the above model. The logarithms of both sides of the Equation (1) yield.

ln [r(a)/n(a)] = ln M + m. V(a) ... ... ... (2)

Furthermore, if ln r(a)[n(a) = y

and

ln M = c and v(a) = x, then

the Equation 2 is found to be a linear equation in the form of

y = c + mx.

Least square regression was used to estimate "M" and "m" values for the following schedules and are presented in Appendix A.

(i) The marital age-specific fertility schedule observed in the 1963-65 Population Growth Experiment (PGE) Survey (average of longitudinal and Chandra and Deming method estimates), assumed to be potential fertility in this analysis;

(ii) The marital age-specific fertility schedule of the 1975 PFS; and

(iii) The marital age-specific fertility schedule estimated on the basis of contraceptive use in 1985 by taking the PGE schedule as potential fertility.

According to Coale and Trussell (1978) the reliability of the marital fertility schedule with respect to "m" values can be determined by examining the mean square error of the logarithmic regression of the optimal values of parameters. They proposed that a mean square error of zero will signify a perfect fit, a value of 0.005 will give a mediocre fit and that of a 0.01 will show a poor fit. As per this criterion, the fertility schedule of the 1975 PFS is very close to a perfect fit and the 1985 estimated marital schedule is better than mediocre and the PGE lies between mediocre and a poor fit (see Appendix A).

Conversion of "m" into Contraceptive Use

The estimated parameter "m" (0.254) for the marital fertility schedule of PFS indicates a degree of fertility control in Pakistan in 1975. We know that the value of "m" at 0.256 for the 1985 estimated fertility schedule is associated with 19.8 percent contraceptive use. For converting 'm' values into contraceptive use, two second degree polynomial regression equations were analysed, taking 'm' values as independent variable and contraceptive use being the dependent variable. The first equation included 26 countries while the second included 16 countries having comparatively more reliable data on contraceptive use and fertility.

RESULTS

Equation (1) explains almost 90 percent of the variance in the contraceptive use and the association is found to be highly significant. The standardized regression coefficient is 0.95.

The polynomial association from the origin is:

Y = 87.631(x) - 26.873 [(x).sup.2]

The estimated contraceptive use on the basis of this association is given in Table 3. The estimated use in Pakistan in 1975 is 20.5 percent against the "m" value of 0.254, whereas the reported use corresponding to this period is 5.3 percent and the estimated contraceptive use on the basis of service statistics is 16.8 (2) percent [Syed (1981)].

Since the value of "m" describes a degree of deliberate fertility control in a population, the predicted value on the basis of "m" value includes modern and traditional methods. Assuming that the fertility schedule for Pakistan is correct the reported contraceptive use in the 1975 PFS appears to be substantially underreported. Equation (2) explains 98 percent of variation in contraceptive use. The polynomial association found in this regression is:

Y = 86.776 (x)-27.286 [(x).sup.2]

The standardized regression coefficient is 0.99. The association is found to be highly significant. The predicted rates of contraceptive use are also close to the reported contraceptive use in these countries (Table 4). On the basis of Equation (2) the level of contraceptive use in Pakistan against the value of "'m" of 0.254 for 1975 is 20.3 percent. It is observed that with the increase in variance explained from 0.90 in Equation (1) to 0.98 in Equation (2), the estimated contraceptive use for Pakistan decreased marginally from 20.5 to 20.3 percent. Predicted contraceptive use includes both programme and non-programme methods of contraception and represents an overall degree of fertility control in Pakistan. Contraceptive use in 1975 estimated by Equations (I) and (2) was much higher than the reported contraceptive use in the 1975 PFS and was even higher than the estimated contraceptive use for 1975 based on service statistics. The estimated contraceptive use levels in this analysis can also be substantiated from the findings of Nortman (1982)and Lee and Lucas (1986). They observed that with a Crude Birth Rate (CBR) of 40 per thousand in Pakistan (PFS 1975) the corresponding contraceptive prevalence should have been around 20 percent.

Validity of Regression Results

We know that marital age specific fertility rates estimated for 1985 are based on 19.8 percent contraceptive use for which the estimated value of 'm' is 0.256. Against this value of 'm', Equation (1) and Equation (2) estimate a contraceptive use level of 20.7 and 20.4 percent respectively. The close agreement between the direct and indirect estimate of contraceptive use validate the application of the indirect method on any fertility schedule for arriving at reliable estimates of contraceptive use in Pakistan.

CONCLUSION

Both service statistics and survey statistics are found to be inadequate for estimating contraceptive use in Pakistan. Service statistics, if taken at face value, will highly overestimate contraceptive use, whereas, survey reports are underestimating contraceptive use in Pakistan. The indirect estimates using 'm' values in the polynomial regression equation appears to be fairly reliable as they provide estimates closer to the one derived from adjusted service statistics.

Appendix

Appendix Table A
Marital Age-specific Fertility Rates

 Estimated
 Schedule
Age PGE PFS 1985

20-24 0.362 0.355 0.336
25-29 0.349 0.362 0.282
30-34 0.331 0.286 0.265
35-39 0.233 0.221 0.165
40-44 0.110 0.104 0.087
45-49 0.066 0.009 0.056

M 0.782 0.887 0.728
m 0.074 0.254 0.256
Percent Error 10.6624 5.5568 8.0996
Mean Square Error 0.0067 0.0003 0.0036

REFERENCES

Ahmed, G., W. P. Schellstede and Nancy E. Williamson (1987) Underreporting of Contraceptive Use in Bangladesh. International Family Planning Perspective 13 : 4 136-140.

Coale, Ansley J. (1971) Age Pattern of Marriage. Population Studies 25 : 2 193-214.

Coale, A. J., and T. J. Trussell (1978) Technical Note: Finding the Two Parameters that Specify a Model Schedule of Marital Fertility. Population Index 44 : 2 203-213.

ESCAP (1989) Population Headliners No. 173. Bangkok : ESCAP.

Henry, Louis (1961) Some Data on Natural Fertility. Eugenices 8 : 2 81-91.

Koening, M. A., G. B. Simons and B. D. Misra (1984) Husband-Wife Inconsistencies in Contraceptive Use. Population Studies 38 : 281.

Lee, Sun-hee, and D. Lucas (1986) Estimating Required Contraceptive Prevalence. Research Note No. 51. International Population Dynamics Programme, Department of Demography, Canberra. The Australian National University.

Nortman, D. L (1982) Population and Family Planning Programmes: A Compendium of Data Through 1981 (11th Edition). New York: The Population Council.

Nortman, D. L. (1985) Population and Family Planning Programmes: A Compendium of Data Through 1983 (12th Edition). New York: The Population Council.

Pakistan, Government of (1976) Pakistan Fertility Survey: First Report (World Fertility Survey). Islamabad: Population Planning Council.

Pakistan, Government of (1986) Pakistan Contraceptive Prevalence Survey 1984-85. Islamabad: Population Welfare Division.

Rukanuddin, A. R., G. Y. Somro and N. I. Farooqui (1985) Evaluation of the Impact of Family Planning Programmes: The Case of Pakistan. Studies to Enhance the Evaluation of Family Planning Programme. New York: United Nations. (Population Studies No. 87)

Streatfield, Kim (1985) A Comparison of Census and Family Planning Programme Data on Contraceptive Prevalence, Indonesia. Studies in Family Planning (Part 1) 16 : 6 342-349.

Sultan, Mehboob (1987) The Family Planning Programme and its Effects on Fertility in Pakistan. Unpublished Master's Thesis. Canberra: The Australian National University.

Syed, S. Hasan (1981) Acceptors of Population Programme in Pakistan. The Pakistan Development Review 20 : 1 81-93.

Vaessen, M. (1981) Knowledge of Contraceptive Use; An Assessment of World Fertility Survey Data Collection Procedures. Population Studies 35 : 3 357-373.

([dagger]) Comments on this paper have not been received.

(1) For details of these methods see U. N. Manual IX.

(2) Service statistics estimates for 1975 were derived using wastage factor of 0.5 to conventional contraceptives and 0.25 to pills.

MEHBOOB SULTAN and SYED MUBASHIR ALI *

* The authors are Research Officer, Population Welfare Division, Islamabad and Research Demographer, Pakistan Institute of Development Economics, Islamabad respectively.

Table 1
Reported and Estimated Contraceptive Use Rates by
Source, Pakistan 1975 and 1984

Source (1) (2) (3) (4) (5) (6) (7)

PFS (1975) 0.6 0.9 0.9 1.1 -- 3.5 5.3

Services
 Statistics
 Estimates
 (1975) 3.0 1.1 3.2 9.5 -- 16.8 --

PCPS (1984) 0.8 2.6 1.4 2.2 0.6 7.6 9.1

Services
 Statistics
 Estimates
 (1984) 2.0 1.6 5.8 10.2 0.2 19.8 --

(1 = IUD, 2 = Sterilization, 3 = Pill, 4 = Conventional
Contraceptives, 5 = Injectable, 6 = All programme Methods,
7 = All Methods).

* For service statistics estimates of 1975 and 1984.
[See Syed (1981) and Sultan (1987)] respectively.

Table 2
Values of n(a) and v(a) by 5 Year of Age Groups

 Age Groups

 15-19 20-24 25-29 30-34

n (a) 0.411 0.460 0.431 0.395
v(a) 0.0 0.00 -0.279 -0.667

 Age Groups

 35-39 40-44 45-49

n (a) 0.322 0.167 0.214
v(a) -1.04 -1.41 1.671

Source: [Coale and Trussell
(1978), p. 205].

Table 3
Reported Contraceptive Use and Estimated Use
on the Basis of Regression I
(26 Countries)

 Reported Estimated
 'm' Contraceptive Contraceptive
Country values Use Use

1. Bangladesh 0.326 07.9 25.7
2. Colombia 0.420 42.4 32.1
3. Costa Rica 0.512 64.4 44.9
4. Dominican Republic 0.437 31.8 33.2
5. Egypt 0.698 24.0 48.1
6. Fiji 0.795 41.5 52.7
7. Guyana 0.878 30.8 56.2
8. Hong Kong 1.569 72.4 71.3
9. Infonrdis 0.482 26.3 36.0
10. Jamaica 0.538 38.4 39.4
11. Jordon 0.453 25.2 34.2
12. Kenya 0.042 9.0 3.6
13. Korea 1.077 54.5 63.2
14. Malaysia 0.817 33.0 53.7
15. Mexico 0.375 30.4 29.1
16. Nepal 0.133 2.4 11.2
17. Pakistan 0.254 5.3 20.5
18. Panama 0.768 54.0 51.5
19. Peru 0.318 31.4 25.1
20. Philippines 0.365 36.2 28.4
21. Singapore 2.010 71.0 67.6
22. Sri Lanka 0.681 32.0 47.2
23. Syria 0.103 20.0 8.7
24. Thailand 0.363 33.3 28.3
25. Trinidad and Tobago 0.669 52.0 46.6
26. Turkey 0.789 50.0 52.0
27. Pakistan * 0.256 19.8 20.7

Source: For reported contraceptive use see [Nortman
(1985), pp. 56-57].

Note: The value of 'm' is based on estimated MASFR for 1985
which is the result of 19.8 percent contraceptive use on
the basis of service statistics.

* Not included in the regression.

Table 4
Reported Contraceptive Use and Estimated
Use on the Basis of Regression II (16 Countries)

 Reported Estimated
 'm' Contraceptive Contraceptive
Country Values Use Use

1. Dominican 0.437 31.8 33.2
 Republic
2. Fiji 0.795 41.5 52.7
3. Hong Kong 1.569 72.4 71.3
4. Indonesia 0.482 26.3 36.0
5. Jamaica 0.538 38.4 39.4
6. Jordon 0.453 25.2 34.2
7. Malaysia 0.817 33.0 53.7
8. Korea 1.077 54.5 61.8
9. Mexico 0.375 30.4 29.1
10. Peru 0.318 31.4 25.1
11. Philippines 0.365 36.2 28.4
12. Singapore 2.010 71.0 67.6
13. Sri Lanka 0.681 32.0 47.2
14. Panama 0.768 54.0 51.5
15. Thailand 0.363 33.3 28.3
16. Turkey 0.789 50.0 52.0
17. Pakistan * 0.256 19.8 20.7

Source: For reported contraceptive use see
[Nortman (1985), pp. 56-57].

Note: The value of 'm' is based on estimated MASFR for 1985
which is the result of 19.8 percent contraceptive
use on the basis of service statistics.

* Not included in the regression.