Geoscience Reference
In-Depth Information
16.5
Simulation Results
Our explanation of the economic modeling results is based around a series of
decomposition figures explaining outcomes for regional macroeconomic variables
in terms of the individual contributions made by the eight sets of shocks. The
decomposition figures are created by running the CGE model nine times: one full
simulation in which all eight sets of shocks are implemented simultaneously, and a
further eight simulations in which each of the eight sets of shocks is implemented
individually. This allows us to explain total impacts in terms of the individual
contributions made by each of the behavioral and resource shocks.
18
We explain the
figures in a logical sequence, relying on references to economic mechanisms within
the LA-DYN model to support our narrative.
19
We begin our discussion with the
event year (2013). There are two main points of entry to understanding the 2013
results: negative deviations in the regional terms of trade (Fig.
16.3
), and invest-
ment (Fig.
16.4
). In Fig.
16.3
we see that the dominant contributor to the 2013 terms
of trade deviation is the decline in willingness to pay for LA County goods. This
exerts a direct effect on the regional terms of trade, depressing prices of LA County
goods relative to competing imports. In Fig.
16.4
we see three shocks exert a strong
negative influence on 2013 investment: investor risk premium, willingness to pay
and BI.
The rise in required rates of return on LA County capital directly affects regional
investment, depressing capital formation relative to baseline for any given rate of
return (e.g. by about 0.2 % in the event year).
18
The sum, for any variable, of results from the eight individual simulations is close to, but not
exactly equal to, the results from the full simulation. This is because the model is non-linear, and
interactions between the individual shocks that are captured by the full simulation are missed when
the shocks are implemented individually. The difference between the sum of the eight individual
simulations and the full simulation is reported as “Residual”. The value for this is small for all
variables in all years.
19
Dixon and Rimmer (
2013
) describe eight ways in which CGE model results can be benchmarked
or validated. Not all the methods they outline are required for every application. Rather, they
advocate tailoring the validation procedure to the purpose at hand. In our discussion of results, we
use the third of Dixon and Rimmer's procedures: qualitative validation via a narrative relying on
economic mechanisms within the model (pp. 1297-1298). At the same time, while not reported in
this paper, we have also relied on the first two of their validation methods (test simulations for
which the results are known a-priori, and within-simulation cross-checks of national accounts
identities). Their remaining methods (particularly vi-viii, p. 1272) are well beyond the scope of the
present paper, representing independent CGE validation modeling exercises in their own right. For
example, Dixon and Rimmer discuss validation of CGE results through out-of-sample forecasting.
Examining the question with a 500 sector model of the U.S. economy, they find their CGE model
forecasts over a seven year period are more accurate than trend extrapolation. They go on to argue
that CGE forecasts can be improved further with better forecasts for macro and trade variables, and
greater use of publicly available information on plausible future paths for commodity-specific and
industry-specific variables relating to tastes, technologies and policy (Dixon and Rimmer
2013
,
pp. 1314-1324).
