Skip Navigation Planning & Markets
Subscribe Submission Requirements Editorial Board Archive Links Search Home

III. Questions, assumptions and methodology

These findings lead to the empirical question of whether decentralizing taxing and spending powers to the local level is indeed enough to contain the local Leviathan, and whether this can positively influence growth rates and development. It is assumed here that the sum of the local government policies in a country influences to a large extent the outcome of the national economy, and that therefore extensive research on these issues on the local level are both timely and important for answers to growth and development questions.

The present research deals with the devolution of powers to villagers versus the exclusive or partial concentration of these powers in the hands of the local government council. When the talk is of decentralization of government, it is always understood to ultimately mean granting direct democratic powers to the villagers themselves, with the local government being the executioners of the peoples' will only. This is more or less the local democratic system in Switzerland, which is used as a benchmark in this research.

Adequate educational standards of the population and adequate tax bases for the financing of the local government are considered a prerequisite for a decentralized state with self-administering, empowered political units. Inadequacy with respect to these two factors is most often cited by central government as the basic hindrance to a decentralization process. The above suggests, that decentralization is often thought to be a concept more suitable for advanced, wealthier economies than for poor ones. The four localities mentioned above have been chosen to address this assertion. The four localities' basic structural data are in Table 4 below.

TABLE 4: Overview of the Localities

Huay Yai, Cholburi, Thailand Labac, Cavite, Philippines Zorita, Caceres, Spain Schwende, Appenzell I. Rh., Switzerland
Population 17,166 4,350 2,029 1,914
Number of Households 3,516 880 853 785
Average Household Income in 1998 US$ 2,432 3,000 10,976 40,000
Regular LG Income from Taxes and Charges in 1998 US$ ('000) 496.7 37.6 384.8 1,216.5
Regular LG Income in % Total Village Income 5.81 1.42 4.11 3.87
Upper Government Contribution in % Regular LG Income 30.0 258.0 153.0 4.8
LG Expenditure Structure in % of Total Budget
Administration, Debt Service 16.8 10.8 30.2 18.5
Education, Culture 7.0 7.4 8.3 14.7
Welfare ~ 0 ~ 0 10.3 13.06
Infrastructure 70.7 81.7 51.2 53.64
Savings 5.4 0.0 0.0 ~ 0

Analyses at the local level generally have to cope with poor data, both in quality and quantity. The research methodologies to be employed therefore had to allow the use of soft data and still produce valuable results. The objectives of the research are of the "what if" type, which means that answers must be found through simulations. A simulation methodology allowing the use of soft data is System Dynamics (SD), and this methodology was chosen for this research.

Although, the structure of political-administrative systems is generally considered to be static, the behavior of the systems is not. The state apparatus behaves dynamically within the "information/action/consequences" paradigm (R.G. Coyle, 1996). To analyze the dynamic behavior of highly complex systems such as politico-administrative systems, the research methodology to be applied must also enable dynamic modeling and simulation of systems with a large number of related sub-systems containing a great number of "non-observable variables" such as tastes, preferences and values. Due to the lack of hard data describing the very important non-observable variables in politico-administrative systems, econometric techniques become questionable as a major instrument of analysis, though a number of the non-observable variables could reasonably be treated as dummy variables.

We deal here with managed, self-steering, intentional systems of the "multiple-loop nonlinear feedback" type (Aulin, 1982, p. 68/69), as encountered in governmental reality. A system dynamic approach seems best suited for the simulation of decentralizing governments. Only in an iterative time simulation process can the many and highly complicated forces of influence be brought to work and made to reveal the long-term dynamic behavior of the system. This is exactly what can be accomplished in a system dynamic analysis (Forrester, 1971, Senge, 1990).

Unfortunately, a fruitless academic divide has developed between econometricians and system dynamicists. The econometricians rejected the system dynamic approach by claiming it to be unscientific, speculative, lacking factual analysis and missing a clearly defined methodological apparatus. The system dynamicists on the other hand rejected the purist econometric approach, arguing that econometric analysis is basically a static approach, as the past is merely and mostly linearly extrapolated into the future, whereas the system dynamic approach embodies multiple feedback learning systems in which everything changes and adapts over time to create a new future, different from the past status (Sommer, 1981).

Forrester (1971), the founder of system dynamics, argued that in analyzing socio-political systems, the very important "non-observable" variables mostly have to be excluded from econometric models due to the absolute need for hard statistical data. System dynamic analysis is based on the profound knowledge and experience of the researcher in the topic she analyzes. Rough estimates of necessary values are sufficient. The aim of system dynamics is not, as in econometrics, to determine functional relations based on theories and statistical data, nor to numerically predict short to medium range outcomes. Rather, the aim is to show in a strategic fashion development paths of complicated managed systems over the long to very long term. In system dynamics, actual or theoretical structures and processes are fundamentally important, and these can be derived only through expert knowledge of the problems and situations under review. Sommer (1981) declares that it is, therefore, futile to claim superiority for one or the other methodology, as they serve totally different aims. During the past twenty years, there have however, been developments, mainly in the field of policy consulting, to econometrically estimate parameters for system dynamic models, and to include them in the analysis. SD in this non-orthodox sense is used here.

One of the more fundamental critiques of econometricians was that system dynamics treated all realities as closed systems, thus not allowing exogenous or policy variables. The closed system approach was introduced because Forrester and most of his students had an engineering or natural science background. They, therefore, treated all systems as physical or biological systems that follow the mass conservation law of physics. This meant that no new mass can be created, or in more general terms, nothing can be created which was not already there. They thus followed closed systems. This is obviously wrong, particularly in the socio-political sphere where human inventiveness creates new structures, processes, products, and services. System dynamicists have, therefore, begun to work with open systems using exogenous variables and policy parameters, implying that there are things that penetrate and enrich existing systems from the outside. It is this non-purist system dynamic approach; also employed here, in which econometric tools are used together with system dynamic tools, that make this methodology interesting and rewarding.

One major problem in econometric analysis is the determination of causality. Through a system dynamics approach, this problem can be solved to a great extent: Total causality can be detected by departing from the smallest cause effect chain, and developing it into an ever larger interrelated system.

The basic system dynamics model elements are the stock and flow variables called levels and auxiliaries, respectively. Levels describe the status of the system, whereas auxiliaries cause change in the levels over time.

Though system dynamics simulations can be carried out based solely on assumed structures, processes and data, the present research was intended to be more empirical. To be able to fill the system models with empirical behavioral data, two questionnaire surveys on a confidence level of 90 percent were carried through in each locality, one with the local leaders and one with the general public. The questionnaires were a mixture of quantitative and semi-open qualitative questions. The objective of the surveys was to gain data on people's needs, perceptions and relations. The data were also used to estimate household incomes and expenditure patterns, because official data proved to be too contradictory to be confidently used. Relational data on the village leaders were collected to serve as the basis for a Social Network Analysis (SNA). The SNA was carried through to be able to include into the models bargaining powers of the relevant political village groups.

[The SD software used was Vensim PLE version 4.1 and the SNA software was Structure 4.2 (Burt, R., 1991). Table functions for the SD models were estimated statistically with SPSS software.]

Riggs' Sala Model (Riggs, 1966) is assumed to represent reality to some extent for developing countries or for small localities in industrialized countries. The Sala model deals with the full range of social phenomena and behavior, subsuming political and administrative aspects. It thus is a model pertaining to the ecology of administration in a society. Heady (1962, p.81) proclaims that "...the Sala is associated with unequal distribution of services, institutionalized corruption, inefficiency in rule application, nepotism in recruitment, bureaucratic enclaves dominated by motives of self-protection, and, in general, a pronounced gap between formal expectations and actual behavior. ..." It is assumed that the Sala behavior will be less prevalent, the more politically empowered the villagers are.

Structure and relations of social systems as outlined in the Sala Model can be described and analyzed with the tools of SNA. The intensity of relations, which resulted from the questionnaire surveys, and which were measured in terms of how often an individual meets with other individuals within a year, are the basis of the social network analysis. Cliques are the focal groups in SNA. Clique building depends very much on the researchers choice of relation measurement. Operationally, a clique is defined as an aggregate of actors clustered on a criterion of cohesion. Burt's (1998) Maximum Strength Relation measure is used in this research. This relation measures proximity of actors according to their strongest Euclidean Distances measured in terms of contacts per year.

In the clustering algorithm, the first step merges two most proximate or cohesive actors. See Figure 1 below. Proximity of an actor to this cluster is defined as the maximum cohesion between the actor and the actors in the cluster. The second step again merges the former cluster with the next most proximate actors and so on until the whole system is one cluster. This clustering algorithm ignores weak or absent ties. The higher up in the cluster hierarchy actors are, the more closely tied they are, and thus can be identified as a group entertaining more intensive relations among each other than with other actors of the system. Such a cluster can then be identified as a relation or interest clique.

D i e g o M a n u e l M a r i a i s a M a r i b e l M a n u e l a J u a n F r a n P i l a r F r a n c i s c I s i d o r o J u a n J o s e J u a n C a r l H e r m i n i a T
61.221 . . . . . X X X . . . . . .
84.942 . . . . X X X X X . . . . . .
86.603 . X X X . X X X X X . . . . . .
133.860 X X X X X . X X X X X . . . . . .
163.775 X X X X X X X X X X X X . . . . . .
338.305 X X X X X X X X X X X X X . . . . . .
363.456 X X X X X X X X X X X X X . . . . X X X
386.192 X X X X X X X X X X X X X . X X X . X X X
499.996 X X X X X X X X X X X X X . X X X X X X X X
801.862 X X X X X X X X X X X X X . X X X X X X X X X
968.466 X X X X X X X X X X X X X X X X X X X X X X X X
2412.394 X X X X X X X X X X X X X X X X X X X X X X X X X

FIGURE 1: Stucture 4.2 Output of Positional Equivalence Analysis for the Zorita Municipal Council

One can identify two cliques in Figure 1 above. This grouping actually coincides with reality where, with the exception of Isidoro, that is the Municipal Secretary, the councilors Francisc(o) to Tomasa are all members of the Socialist Party. The group Diego to JuanFran(cisco) are from right-wing parties or independents. Pilar is the representative of the Provincial Government. The positional equivalence identified here are the socialist homogenous bloc and the heterogeneous, less powerful bloc of the other councilors. The heterogeneous bloc forms the opposition in the Municipal Council, who is equivalent in this function only.

Figure 1 also reveals that the socialist group is positionally less equivalent than the opposition. It appears that there is a female dominated group (Herminia, Tomasa; Juan Carlos), and a male group (Francisco, Isidoro, Juan Jose) within the socialist clique. Francisco is the least equivalent person within the municipal council. From this, one could infer that he is either the weakest or strongest or positionally most unequal person within the council. As he is the Municipal Mayor, this position clearly suggests that he is the most powerful person within the council, a fact that is confirmed through a subsequent power analysis, and acts as a bridge between the two political blocs. It can therefore be claimed that the methodology can identify cliques quite well.

Within the cliques, the question of which actor is most powerful, and who is most likely to be a follower, is of premier interest. It is assumed that personal influence patterns and power in the localities develops according to Burt's (1998) Power and Structural Hole models. Network power is defined as the relative connectedness of an actor that is reciprocated. The most powerful actor is the person with the highest relative score. In the context of the Sala model, the Structural Hole model describes opportunities for the trading of advantages within the local political context where persons at the crossroads of information flows gain information and connections across governmental levels and local social groups, enabling them to influence plans and budgets to their and their clientele's advantage through the trading of such information. The analysis of the Zorita Municipal Council indeed reveals Francisco as the most powerful person within the council, as was already suggested in the positional equivalence analysis in Figure 1.

To identify power and bargaining positions, cliques within the local government were identified, and the power of each village councilor in the four localities was thus calculated. The councilors were grouped in cliques on the basis of their answers to relevant survey questions, which allowed assigning them to one of two groups. Social concerns prevailed in one, and infrastructure and building concerns prevailed in the other.

On the basis of the relevant laws in each country, the survey, and the SNA results; a template "Politico-Economic System" was constructed where the ultimate drivers of the system are the "Development Frustrations" defined as the difference between the expected and actual development status of politically empowered villagers, and the local government "Self Interest" defined as project preferences as revealed by the field survey results.

Due to their large size, the full versions of the systems cannot be shown in this paper. Figure 2 below shows a simplified version only. The sytem process steps are numbered from 1 to 18.

Figure 2

FIGURE 2: Simplified Dynamic Politico-Economic System of a Locality

Some important sub-processes will be explained below. The models in this research contain one ultimate Level, the "Perceived Development Status", and four intermediary Levels: Education Status, Household Population, Total Local Government Debt and Total Village Income. These Levels are driven by all the auxiliaries in the system, which themselves are determined and driven by parameters, lookups, and time. Lookups are arbitrarily specified nonlinear relationships describing assumed or empirically derived relations between system variables. Lookups thus force variables to behave according to a theoretical or empirical model.

The system dynamic process is based on a system of first-order difference equations that drive the auxiliaries and levels from time step to time step.

It is important to note that the models of the four localities differ in two respects:

  • Structure: The researcher's present base is Thailand. More detailed information on the informal financing process in Huay Yai could be gained from field research. This allowed the construction of a detailed model of the irregular financing process for the Thai locality. This was not possible for the other localities, where a black box approach had to be applied with regard to irregular financing. Besides this small difference, the structures of the four localities are identical, and designed to accommodate various degrees of grassroots empowerment and political state organization ranging from dictatorship to a fully empowered federalist system, depending on the values given to the system's structural parameters.
  • Parameters and functions: apart from the Barro and Sala-I-Martin (1995) growth parameters (see explanation below), all other parameters and functions are different from locality to locality. They represent the legal basis and the present politico-economic reality of each country and locality found through the field research.

The full systems are very complicated and detailed representations of the villages' politico-economic reality. Their structure consist of the following elements:

  • Huay Yai, Thailand: 152 auxiliaries and constants plus 21 lookup functions, a maximum of 32,766 feedback loops; and
  • Labac, Phlippines; Zorita, Spain; Schwende, Switzerland: 134 variables and constants plus 21 lookup functions, a maximum of 32,766 feedback loops.

page 32


USC Seal

Main Page | Subscribe | Submission Requirements | Editorial Board | Archive | Links

ISSN 1548-6036

Copyright 1999-2000
University of Southern California
Los Angeles, California 90089-0626