Measurement error in recall surveys and the relationship between household size and food demand.

has earthen floor Average floor area of dwelling 42.3 41.0 0.63 of the household who are male 47.2 47.6 0.70 of the households whose head 70.8 73.2 1.19 is literate Sample size 3,000 3,000 (a) Corrected for cluster structure of the samples. (b) The exchange rate in Round 1 was Ra.780 per US$, and in Round 2 R3,840 per US$. ** = significant at 1% level. Table 4. Comparison of Proxies for Interviewer Performance in the Two Rounds of the 1999 Cambodia Socio-Economic Survey

Round 1 Round 2 of re-interviews due to 40.0% 28.0%

incomplete, incorrect or

doubtful entries Average % of households 48.1% 42.8%

recording zero purchases

of foods Average % of households 69.8% 57.8%

recording zero

own-produce/gifts of

foods Source: Author's calculations. Table 5. Regression Estimates of Food Demand Models with Explanatory Variables Interacted with Indicators of Survey Recall Type, SUSENAS Urban Java

Equation (8)

Wage Income > 0

OLS OLS IV (a) Engel curve coefficient estimates ln(x/fz)([??]) -0.1332 ** -0.1308 ** -0.1322 **

(0.0029) (0.0033) (0.0056) [ln(x/n)] x D ... ... ... ln n([??]) -0.0625 ** -0.0667 ** -0.0672 **

(0.0029) (0.0037) (0.0042) [ln n] x D 0.0158 ** 0.0120 ** 0.0122 **

(0.0031) (0.0041) (0.0042) [R.sup.2] 0.379 0.388 0.388 Sample size 32,037 19,594 19,594 Elasticity of per capita food expenditure w.r.t household size [partial derivative]ln([g.sub.f]/n)/ [partial derivative]ln n = [gamma]/[[bar.w].sub.f] Short questionnaire -0.1045 ** -0.1119 ** -0.1127 **

(0.0048) (0.0063) (0.0070) Long questionnaire -0.0835 ** -0.0977 ** -0.0983 **

(0.0049) (0.0059) (0.0062) [H.sub.0]: equal F = 15.24 ** F = 4.01 * F = 4.02 * elasticities Engel estimate of household scale economies: [delta] = [gamma]/[beta] Short questionnaire 0.469 ** 0.510 ** 0.508 **

(0.018) (0.026) (0.026) Long questionnaire 0.350 ** 0.418 ** 0.416 **

(0.018) (0.023) (0.024) [H.sub.0] [[sigma].sub. F = 26.05 ** F = 8.57 ** F = 8.74 ** short] = [[sigma]. sub.long]

Equation (9)

Wage Income > 0

OLS OLS IV (a) Engel curve coefficient estimates ln(x/fz)([??]) -0.1290 ** -0.1259 ** -0.1355 **

(0.0041) (0.0046) (0.0083) [ln(x/n)] x D -0.0092*** -0.0104*** 0.0081

(0.0052) (0.0060) (0.0107) ln n([??]) -0.0609 ** -0.0652 ** -0.0689 **

(0.0034) (0.0042) (0.0052) [ln n] x D 0.0131 ** 0.0100 *** 0.0156 *

(0.0044) (0.0055) (0.0064) [R.sup.2] 0.382 0.390 0.389 Sample size 32,037 19,594 19,594 Elasticity of per capita food expenditure w.r.t household size [partial derivative]ln([g.sub.f]/n)/[partial derivative]ln n = [gamma]/[[bar.w].sub.f] Short questionnaire -0.1018 ** -0.1093 ** -0.1155 **

(0.0056) (0.0071) (0.0088) Long questionnaire -0.0855 ** -0.0986 ** -0.0952 **

(0.0053) (0.0065) (0.0067) [H.sub.0]: equal F = 4.68 * F = 1.28 F = 3.46 *** elasticities Engel estimate of household scale economies: [delta] = [gamma]/[beta] Short questionnaire 0.472 ** 0.518 ** 0.508 **

(0.020) (0.029) (0.027) Long questionnaire 0.346 ** 0.405 ** 0.418 **

(0.019) (0.024) (0.027) [H.sub.0] [[sigma].sub. F = 22.21 ** F = 8.98 ** F = 5.44 * short] = [[sigma]. sub.long] Note: Standard error estimates and hypothesis tests correct for cluster structure of the samples. Standard errors in (); ** = significant at 1% level, * = significant at 5% level, *** = significant at 10% level. The degrees of freedom for the F-tests are I and 1.431. All regressions contain the logarithm of per capita total expenditure, the logarithm of household size- the ratios to household size of the number of males and females in the 11-4, 5-9, 10-14, 15-29, 3054, and (males only) 55+ age groups, the fraction of adults whose main occupation is agricultural self-employment, agricultural employment and non-agricultural work, a dummy variable for farm households. and dummy variables for province. (a) The log of per capita wage and salary income is used as the instrument. The partial [R.sup.2] for the instrument in the first stage regression is 0.175 and the F-test for excluding the instrument is 4,154. Table 6. Regression Estimates of Food Demand Models with Explanatory Variables Interacted with Indicators for Survey Round, CSES Cambodia

Equation (8)

OLS IV (a) Engel curve coefficient estimates ln(x/n)(R) -0.1162 ** -0.1523 **

-0.0069 -0.0094 [ln(x/n)] x D ... ...

-0.0153 ln n([??]) -0.0434 ** -0.0566 **

-0.0082 -0.0088 [ln n] x D 0.0386 ** 0.0447 **

-0.0085 -0.0088 [R.sup.2] 0.418 0.405 Sample size 6,000 5,997 Elasticity of per capita food expenditure w.r.t household size [partial derivative]ln([q.sub.f]/n)/ [partial derivative]ln n = [gamma]/[[bar.w].sub.f] Round 1 (less probing) -0.0657 ** -0.0858 **

-0.0125 -0.0134 Round 2 (more probing) -0.0072 -0.0178

-0.0123 -0.0125 [H.sub.0]: equal F = 20.88 ** F = 26.72 ** elasticities Engel estimate of household scale economies: [sigma] = [gamma]/[beta] Round 1 (less probing) 0.373 ** 0.372 **

-0.067 -0.052 Round 2 (more probing) 0.041 0.079

-0.071 -0.054 [H.sub.0]: [[sigma]. F =19.43 ** F = 26.64 ** sub.Round1] = [[sigma]. sub.Round2]

Equation (9)

OLS IV (a) Engel curve coefficient estimates ln(x/n)(R) -0.1020 ** -0.1630 **

-0.0082 -0.0117 [ln(x/n)] x D -0.0519 ** 0.0180

-0.0195 ln n([??]) -0.0409 ** -0.0607 **

-0.0102 -0.0112 [ln n] x D 0.0308 * 0.0527 **

-0.0141 -0.0148 [R.sup.2] 0.431 0.407 Sample size 6,000 5,997 Elasticity of per capita food expenditure w.r.t household size [partial derivative]ln([q.sub.f]/n)/ [partial derivative]ln n = [gamma]/[[bar.w].sub.f] Round 1 (less probing) -0.0620 ** -0.0920 **

-0.0155 -0.0169 Round 2 (more probing) -0.0151 -0.0119

-0.0144 -0.0144 [H.sub.0]: equal F = 4.91 * F = 12.99 ** elasticities Engel estimate of household scale economies: [sigma] = [gamma]/[beta] Round 1 (less probing) 0.401 ** 0.373 **

-0.095 -0.061 Round 2 (more probing) 0.066 0.055