A Top-down Framework for Regional Historical Analysis

Abstract

Bottom-up regional computable general equilibrium (CGE) models have clear theoretical advantages over their top-down counterparts. However bottom-up models are data intensive. Hence they face practical difficulties in applications requiring high levels of regional and sectoral disaggregation, such as explaining regional economic outcomes, and regional forecasting and policy analysis. This paper develops a top-down framework for explaining recent economic history for many regions. This requires estimation of variables describing regional structural change. These variables have a further use in generating plausible regional forecasts. Such forecasts are a prerequisite for convincing regional policy analysis.

Abstract

Un cadre descendant pour une analyse régionale historiqueLes modèles CGE régionaux ascendants ont un clair avantage théorique sur leurs contreparties descendantes. Pourtant, les modèles ascendants sont à données intensives. Ils doivent par conséquent faire face à des difficultés d'ordre pratique pour les applications nécessitant de hauts niveaux de désagrégation régionale et sectorielle, comme l'explication des résultats économiques régionaux ainsi que les prévisions régionales et l'analyse des politiques. Cette étude développe un cadre descendant pour expliquer l'histoire économique récente de nombreuses régions. Ceci nécessite l'estimation de variables décrivant un changement structurel régional. Ces variables ont un usage supplémentaire dans la création de prévisions régionales plausibles. De telles prévisions sont une condition sine qua non pour une analyse convaincante des politiques régionales.

Abstract

Un marco descendente para un análisis histórico regionalLos modelos regionales ascendentes de equilibrio general computable (EGC) tienen ventajas teóricas evidentes frente a sus homólogos descendentes. No obstante, los modelos ascendentes contienen mayor cantidad de datos. Por consiguiente, tienen dificultades prácticas en aplicaciones que requieren altos niveles de desagregación sectorial y regional, como por ejemplo, la explicación de resultados económicos regionales, y previsiones regionales y análisis de políticas. Este estudio desarrolla un marco descendente para explicar la historia económica reciente de diversas regiones. Esto requiere una estimación de las variables que describen el cambio estructural regional. Estas variables también se utilizan para generar previsiones regionales verosímiles. Estas previsiones son un prerrequisito necesario para desarrollar análisis de políticas regionales convincentes.

Keywords:

• Regional general equilibrium
• regional forecasting
• top-down model

Keywords:

• D58
• R11
• R13

Acknowledgements

The research reported in this paper was supported by an Australian Research Council Discovery Grant.

Notes

1. For example, Horridge et al. (2005).

2. Thirty-seven sectors and two regions in Giesecke (2002) and 27 sectors and eight regions in Giesecke & Madden (2006).

3. The model is fully documented in Dixon & Rimmer (2002). Subsections 2.1 and 2.2 of this paper provide an overview of the model.

4. Based on the ‘sketch model’ of Dixon & Rimmer (2002), p. 243).

5. The origin of equation (5) is the identity NGNP=GDP×P_GDP – NNFL×R, where NGNP is nominal GNP, GDP is real GDP, P_GDP is the price of GDP, NNFL is nominal net foreign liabilities, and R is the interest rate on net foreign liabilities. Dividing through by P_C, the price of consumption, gives equation (5), where Q(TOT)=P_GDP/P_C, and NFL=NNFL/P_C. The ratio P_GDP/P_C is a positive function of the terms of trade, because P_GDP contains export prices, while P_C does not.

6. Equation (9) is based on the first-order condition that the value of the marginal product of capital equals the rental price of capital, noting that the production function is constant returns to scale. See Dixon & Rimmer (2002), p. 244).

7. Equation (10) is based on the first-order condition that the value of the marginal product of labour equals the wage, noting that the production function is constant returns to scale.

8. A fuller (although still simplified) description of the relevant MONASH equation, of which equation (11) is a stylized representation, is:where t is the initial year, t is the solution year, D is the rate of depreciation and all other variables are as defined in the main text. See Dixon & Rimmer (2002, pp. 43–47) for details.

9. The reader is referred to Dixon & Rimmer (2002, pp. 10–13) for a more detailed discussion.

10. The equations describing these links can be found in Dixon et al. (2007).

11. There is no multi-production in the version of MONASH and MONASH-TDR used in this paper. Hence the commodity (COM) and industry (IND) sets are identical. See Appendix B.

12. With equation (A7) we recognize that Australia has regionally dispersed beneficial interests in capital income via high levels of asset securitization, a long-standing compulsory superannuation scheme, and responsibility for the bulk of capital income taxation resting with the federal government. Nevertheless, by using employment, rather than, say, gross regional product, as the variable determining changes in regional disposable income shares, equation (A7) may somewhat underestimate the responsiveness of regional disposable income to regional economic conditions by ignoring changes in income from locally owned regional capital. This could be rectified in future development of MTDR through explicit representation of regional ownership shares in regional industry capital and land.

13. The models are solved with the GEMPACK economic modelling software (Harrison & Pearson, 1996).

14. Such a closure is described by column 2 of Table A1 in Appendix C.

15. Exogenous determination of the other demand-side components of state GSP follows the same steps.

16. The purpose of this last closure swap is not immediately obvious, until one notes that, via equation (A5), the exogenous determination of is sufficient to determine national investment. But so too is the exogenous determination of x2state_obs _s (and thus x2state _s ). Any contradiction is reconciled by endogenizing the uniform scalar f2obs. In Step 3, movements in f2obs (and the other right-hand-side scalars in equations (A35)–(A39)) are small, reflecting the small differences between the MTDR and ABS regional shares in national macro variables.

17. Equation (13) is derived from the stylized NFL equation in note 8 above, after assuming linear growth in the right-hand-side variables of the stylized equation.

18. The finding of adverse technical change in construction reflects the study period ending during a period of very low interest rates (the Australian standard home loan rate fell to a 30-year low in December 2001) and thus high dwellings construction. Difficulties in finding skilled workers, quality building products and reliable contractors were common complaints among participants in the construction industry at this time. For basic metal products, the issue was lower global growth in 2001/2002. This caused prices for metal products to fall sharply, and Australian exports of metal products contracted in response to lower global demand. The finding of lower productivity for basic metal products reflects labour hoarding in the presence of a temporary reduction in demand.

19. Private consumption is approximately 77% of total consumption. Hence, in column 10, total consumption rises by 0.5% (=0.77×1.7+0.23×−3.4).

20. Since much of the national activity of industries producing goods for public consumption (such as Public administration, Defence, Education and Health) is concentrated in the Australian Capital Territory.

21. Table 3, column 10: compare outcomes for sectors 18 and 19 (Education, and Health) with those for sectors 24 and 25 (Public administration, and Defence).

22. The merits of including population as an argument in equation (A7) aside, if the effect of retirees on the Queensland economy was large but missing from MTDR, we would expect a positive contribution to Queensland real GSP by the consumption shift. However this effect is negative (–3.2 percentage points).

23. As reflected in 1996/1997 values for regional sourcing shares, consumption shares, export shares, investment shares and public consumption shares.

24. 100×(17.8 – 15.0)/15.0=16%.

25. Being the sum of Western Australia's entries in columns 3–6 of Table 5 (i.e. –6.5=–2.1 – 1.5 – 3.8+0.9).

26. The sum of columns 3–6, row 57, Table 5.

27. Row 57, column (7) less column (1).

28. A good recent example is the Productivity Commission (2006, pp. 331–335). Dixon & Rimmer (2002) describe in detail how results from a historical simulation can provide inputs to forecasting simulations.