In this paper we describe several regional development-assessment methods and subsequently apply them in a comparative development level analysis of the Slovenian and Croatian municipalities. The aim is to compare performance and suitability of several parametric and non-parametric ranking methods and to develop a suitable multivariate methodological framework for distinguishing development level of particular territorial units. However, the usefulness and appropriateness of various multivariate techniques for regional development assessment is generally questionable and there is no clear consensus about how to carry out such analysis. Two main methodological approaches are based on parametric and non-parametric methods, where in the former an explicit econometric model containing theory-implied causal and possibly simultaneous relationships is estimated using likelihood-based methods and formally assessed in terms of the goodness of fit and other test statistics, subsequently allowing for estimation of the development level on a metric scale, while in the later, territorial units or regions are essentially classified into clusters or groups differing in the development level, but no formal inferential methods are applied to confirm the validity of the model, or to establish the difference in the development level on a metric scale. The possible advantages of the first approach are in the existence of formal testing and evaluation procedures, as well as in producing interval ranks of the analysed units, while its disadvantages are in the lack of robustness; often unrealistic distributional assumptions; and possible invalidity of the theoretically implied causal relationships. In this paper we consider a parametric, inferential approach based on maximum likelihood estimation of the linear structural equation model with latent variables for metric-scale development ranking, and a non-parametric approach based on cluster analysis for development grouping. Our analysis is based on ten regional development variables such as income per capita, population density, age index, etc. which are similarly collected and generally compatible for both analysed countries. Within the parametric approach, a simultaneous equation econometric model is estimated and latent scores are computed for each underlying latent development variable, where three latent constructs are postulated corresponding to economic, structural and demographic development dimensions. In the non-parametric approach, a combination of Ward?s hierarchical method and K-means clustering procedure is applied to classify the territorial units. We apply both methodological frameworks to Slovenian and Croatian municipality data and assess their regional development level. We further compare the performance of both methods and show to which degree their results are compatible. Finally, we propose a unified framework based on both parametric and non-parametric methods, where clustering techniques are performed both on the original development indicators and on the computed latent scores from the structural equation model, and compare these results with the results from each of the two methods applied separately. We show that a combined parametric/non-parametric approach is superior to each approach applied individually and propose a methodological framework capable of estimating the development level of territorial units or regions on a metric scale, while in the same time preserving the robustness of the non-parametric techniques.
Download Info
To download:
If you experience problems downloading a file, check if you have the
proper application to
view it first. Information about this may be contained
in the File-Format links below. In case of further problems read
the IDEAS help
page. Note that these files are not on the IDEAS
site. Please be patient as the files may be large.
Publisher Info
Paper provided by European Regional Science Association in its series ERSA conference papers with number
ersa03p350.
References listed on IDEAS Please report citation or reference errors to , or , if you are the registered author of the cited work, log in to your RePEc Author Service profile, click on "citations" and make appropriate adjustments.: