27.12.2012 Views

The 1995/1996 Household Income, Expenditure - (PDF, 101 mb ...

The 1995/1996 Household Income, Expenditure - (PDF, 101 mb ...

The 1995/1996 Household Income, Expenditure - (PDF, 101 mb ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

11.15<br />

involves the difficulty in defining the extent of investment (saving) inherent<br />

in many "consumed" goods. For example, in some sense the purchase of<br />

schoolbooks is consumption, but it is partly investment as well, as indirect<br />

learning will eventually (it is hoped!) generate income. Housing purchases are<br />

another example: it is difficult to separate the investment aspects of the<br />

purchase of a house from its consumption (use and depreciation) aspects. So<br />

the estimation of savings rates often hinges on categorization of goods as<br />

either consumption or savings goods, when in truth many commodities retain<br />

continuous properties of both.<br />

A better relationship, therefore, focusses on the residuals of the<br />

consumption function. With consumption on the left hand side, and<br />

expenditure on the right hand side, parameter estimates for the regions<br />

appear in Table II.13. Here 92.0568 piasters of every Egyptian pound of<br />

expenditures is consumed. <strong>The</strong> unexplained variation in consumption (less<br />

than 3%) goes to transfers and premiums. To refine this equation for a better<br />

savings estimate one would subtract from the left hand side those goods (or<br />

components of goods) which clearly are investments. Heteroskedasticity<br />

appears to be a feature of the model; the residuals vary with the magnitude<br />

of the total expenditures, the means turning positive somewhere above 6000<br />

£E, and the variances growing ever larger:<br />

Evidence of Heteroskedasticity in the Crude Consumption/<strong>Expenditure</strong> Function<br />

<strong>Expenditure</strong> Group Nu<strong>mb</strong>er of Std Dev of the<br />

Observations Residuals<br />

o - 1500 £E 304 51.761<br />

1501 - 2000 £E 393 63.614<br />

2001 - 2500 £E 478 80.361<br />

2501 - 3500 £E 1563 87.289<br />

3501 - 4500 £E 2059 139.390<br />

4501 - 6000 £E 3361 133.156<br />

6001 - 8000 £E 3041 237.981<br />

8001 + £E 3606 1500.35<br />

<strong>The</strong> F Statistic computed for the Goldfeld-Quandt test produces a value of<br />

204.19 with {14805 - 4935 - 2 (9)}/2 degrees of freedom for both the numerator<br />

and the denominator, indicating easy rejection of the null of homoskedasticity.<br />

Although the least squares model still produces unbiased coefficients, the<br />

model is not efficient in the sense that it will have a higher variance than a<br />

maximum likelihood estimator, regardless of the sample size.<br />

ILL Back to the Engel Relationship<br />

This chapter concludes by returning to the Engel relationship, to place<br />

some empirical bounds around the expenditure level/food share relationship.<br />

It will be extended to group expenditures below using Pearson correlation

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!