11.11.2014 Views

"Life Cycle" Hypothesis of Saving: Aggregate ... - Arabictrader.com

"Life Cycle" Hypothesis of Saving: Aggregate ... - Arabictrader.com

"Life Cycle" Hypothesis of Saving: Aggregate ... - Arabictrader.com

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.

Recent Declines in the <strong>Saving</strong>s Rate 127<br />

<strong>of</strong> l <strong>of</strong> just about 0.3, and that one dollar <strong>of</strong> government deficit reduces national<br />

saving by 77 cents, when according to Barro [4] the effect should be no more<br />

than some 15 cents. The coefficient <strong>of</strong> the other variables are somewhat reduced,<br />

especially that <strong>of</strong> labor force participation, which is nearly halved.<br />

The next battery <strong>of</strong> tests presents results for cross-section tests for each <strong>of</strong> the<br />

decades. In these tests, the results are broadly consistent with those reported<br />

above, but, presumably because <strong>of</strong> the much smaller samples, and apparent multicollinearity,<br />

the estimates are less stable and tend to be a good deal less significant.<br />

We report one single equation corresponding to line 5, i.e. with all the<br />

variables and population weighted. Line 7 <strong>of</strong> table 4.3 is based on the first decade.<br />

The problem <strong>of</strong> instability, and multicollinearity, is clearly visible in the results.<br />

The point estimates <strong>of</strong> the coefficients other than that <strong>of</strong> GS are not very different<br />

from those <strong>of</strong> line 5 but the last three variables are all insignificant, despite<br />

the high R 2 , and insignificantly different from the coefficients in line 5. The coefficient<br />

<strong>of</strong> GS is the one that is appreciably more negative and thus less consistent<br />

with LCH. It implies that a one dollar increase in government deficit would be<br />

<strong>of</strong>fset to the extent <strong>of</strong> 77 cents by higher private saving. However, this result is<br />

not a serious challenge to LCH. In the first place it is unique in the battery <strong>of</strong><br />

tests, and in the second place the estimate is not very reliable: it has a standard<br />

error <strong>of</strong> 0.33. Thus while it is not significantly different from the value <strong>of</strong> 0.85<br />

called for by Barro, it is also not significantly different (even at the 10 percent<br />

level) from the value implied by LCH <strong>of</strong> around 0,25. The coefficient estimate<br />

is just too unreliable to permit a sound conclusion.<br />

In line 8, covering the ’70s, most coefficients, including that <strong>of</strong> GS, are back<br />

at a level close to those <strong>of</strong> line 5, expect possibly fior that <strong>of</strong> Dependency. The<br />

remaining regression for the ’80s is the least satisfactory. All the coefficients are<br />

generally different from (5), but the estimates are subject to such large error that<br />

the differences from the estimates <strong>of</strong> line 5 are not significant.<br />

2.3 First Difference Tests<br />

The purpose <strong>of</strong> the tests reported in the last four lines <strong>of</strong> table 4.3 is to inquire<br />

how well the LCH can account for the large changes in saving that occurred<br />

between the ’60s and ’80s. In this case there are no solid grounds for expecting<br />

that the coefficients will be very close to those obtained from the level cross<br />

section like lines 5 and 6 because <strong>of</strong> the different structure <strong>of</strong> errors; in particular<br />

one can expect that the difference in a given variable between countries will<br />

tend to correspond more nearly to a permanent difference than might be the case<br />

with differences for one country, between two relatively close points in time.<br />

We found in fact systematic differences between the coefficients <strong>of</strong> a magnitude<br />

that was rather surprising. In particular the coefficients <strong>of</strong> the two

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

Saved successfully!

Ooh no, something went wrong!