| Summary: | This thesis takes the form of three essays in which I use disaggregate and
aggregate information to examine Canadian economic growth.
In the first essay, I present evidence that the process of economic growth
differs for low income per capita provinces and industries. This contrasts with
results from traditional studies of economic convergence. In those papers, estimates
of a rate of convergence suggest that poor provinces eventually "catch
up" to rich provinces by growing faster. Unfortunately, this approach ignores
the pattern of economic growth within the cross-section distribution. Explicitly
modelling the evolving distribution, I find little mobility in the cross-sectional
ordering and some evidence of divergence. In the long run, the poor stay (relatively)
poor and the rich remain (relatively) rich.
In the second essay, I examine the dynamic effects of aggregate and disaggregate
disturbances on both economic growth and the interaction between
disaggregates. The approach is motivated by the class of models which predict
two-way interaction between aggregate and disaggregate behaviour, such
as Durlauf [28]. The disaggregate disturbance is identified as having no long-run
impact on aggregate economic growth. I find that the aggregate shock has a
large impact on aggregate income in both the short and long run; and accounts
for most of its variation. The disaggregate shock contains some information for
aggregate activity at business cycle horizons. Most interaction is explained by
the disaggregate disturbance; the aggregate shock contributes little.
In the third essay, I present results from a variety of unit root tests on
provincial and manufacturing industry panel income data. Standard Dickey-
Fuller unit root tests applied to panels require averaging of data across the
cross-section. More powerful tests allow pooling of cross-section and time-series
information. Using these methods, I find that the null hypothesis of a unit
root is rejected—strongly contrasting with results obtained using the standard
Dickey-Fuller methodology.
|
|---|
