Trade and Employment From Myths to Facts - International Labour ...
Trade and Employment From Myths to Facts - International Labour ...
Trade and Employment From Myths to Facts - International Labour ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Chapter 3: Assessing the impact of trade on employment: Methods of analysis<br />
for workers in penetrated industries versus those in others. Return <strong>to</strong> figure 3.6 <strong>and</strong><br />
consider that the interaction of the skilled premium with import penetration must<br />
there change the slope of each of the separate regressions in the right-h<strong>and</strong> panel.<br />
Galiani <strong>and</strong> Sanguinetti (2003) find that the interaction terms are indeed significant<br />
but not large enough <strong>to</strong> explain the more than about 16 per cent of the<br />
skilled labour premium. This is certainly an interesting result <strong>and</strong> this methodology<br />
could easily be applied in a number of different countries. One important point <strong>to</strong><br />
mention, however, is that since import penetration does not explain much of the<br />
rising skilled-labour premia, something else must. When the unknown effects are<br />
fully accounted for then perhaps it will be seen that import penetration suffers from<br />
omitted variable bias. Further research would be needed <strong>to</strong> answer this question definitively.<br />
It should not come as a surprise that trade liberalization leads <strong>to</strong> an increase<br />
in the skilled-labour wage rate relative <strong>to</strong> the unskilled wage rate. As old sec<strong>to</strong>rs contract<br />
<strong>and</strong> new sec<strong>to</strong>rs exp<strong>and</strong>, workers with greater mobility, non-specific creative<br />
abilities <strong>and</strong> generally higher levels of education will be in short supply, <strong>and</strong> even<br />
more so than in times of more balanced growth. The wage premium therefore reflects<br />
the reality that change requires adaptive talent, <strong>and</strong> the signal sent <strong>to</strong> those without<br />
that adequate human capital is that all the incentives are pointed in the direction<br />
of taking greater advantage of the educational system. Indeed, if skill bias in wage<br />
growth did not appear, one could be appropriately sceptical that any change was<br />
actually taking place in the economy.<br />
3.4.3 <strong>Trade</strong>, productivity <strong>and</strong> employment<br />
The previous section conveyed a s<strong>to</strong>ry of rising dem<strong>and</strong> for labour, but with a lag,<br />
as liberalization proceeds. Wage inequality sends a powerful signal that new opportunities<br />
for significantly better living st<strong>and</strong>ards are present if one takes the proper<br />
steps <strong>to</strong> prepare. This is an optimistic scenario <strong>and</strong> a reality that not all countries<br />
have experienced. The transition <strong>to</strong> openness might begin with significant import<br />
penetration, job loss, but then lack the investment necessary <strong>to</strong> open new branches<br />
of production with an export orientation.<br />
Might it be possible <strong>to</strong> determine empirically the slope of the implied relationship<br />
between employment <strong>and</strong> productivity? The problems with a project of<br />
this nature are legion, however, in that one must deal with simultaneity (the correlation<br />
between the independent variable <strong>and</strong> the error term) as well as the usual<br />
problems of data reliability <strong>and</strong> comparability. Above all, there is the problem of<br />
lack of data on productivity <strong>and</strong> employment. Both have <strong>to</strong> be manufactured from<br />
existing data sources before any attempt at estimation can be made.<br />
A second major estimation problem comes in the weighting of the data. Is it<br />
really possible <strong>to</strong> take one observation of a country the size of China <strong>and</strong> India<br />
along with states a micro-fraction of their size <strong>and</strong> deduce anything of scientific<br />
value? Finally, there is the problem faced by most time-series studies that spurious<br />
correlation must be removed by way of co-integration or other techniques. Despite<br />
109