<p class="body-paragraph">不好意思，这个杂志没有pdf格式。</p><p class="body-paragraph">Comparison of Input-Output, Input-Output Econometric and Computable General Equilibrium Impact Models at the Regional Level</p><p class="body-paragraph">ABSTRACT Economic impact studies are a common practice--indeed, a necessary prerequisite in many cases of project development--in Australia. Although input-output (IO) is still regarded as the 'bread-and-butter' model for these types of study, attention in recent years has turned towards more sophisticated models, the main contenders being integrated IO +econometric and computable general equilibrium models. All these models, which are often promoted as substitutes, exhibit characteristics which are theoretically and empirically appealing, yet questions have been raised with respect to the different approaches, with apparently little awareness at the practitioner level as to the extent of these differences. This paper compares the three models and demonstrates that the differences can be quite substantial, even when the models rely on the same database and are subjected to the same impact scenario. </p><p class="body-paragraph">KEYWORDS: Input-output, CGE, integrated model, impact studies, Queensland </p><a name="AN9509104058-2"></a><span class="medium-bold"><a id="hd_toc_9" title="1. Introduction " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">1. Introduction </font></strong></a></span><p class="body-paragraph">There has been a plethora of economic impact studies undertaken in Australia over the last decade or so. Development projects are required by government legislation to have an environmental (including economic) impact statement (EIS) prepared and, in addition, many organizations undertake economic 'significance' studies to 'demonstrate' their contribution to the local economy. The majority of these studies have been carried out in an input-output framework but, while input-output is still regarded by many as the 'bread-and-butter' model, attention in recent years has turned towards more sophisticated regional models. Some of these have been one-off applications, such as input-output+shift share or input-output + projection models, but the main research interest, both in Australia and overseas, currently appears to be concentrated in the input-output+ econometric and the computable general equilibrium areas. These models are often promoted as substitutes for each other. As different research groups pursue their own research interests, and the different models become more widespread in use, potential (usually practitioner) users of these models appear to be becoming increasingly confused between the different methodologies and, more importantly, the different and sometimes conflicting results obtained from the various models in a regional impact situation. </p><p class="body-paragraph">This paper does not attempt to address all aspects associated with these differences. This would require a wide-ranging study involving the review, comparison and validation of the models from both conceptual and empirical viewpoints, under a wide range of contexts and applications. Instead, this paper has a more modest objective, i.e. to provide an overview of the different approaches and demonstrate the differences which may be obtained in the results. Validation of the results is not attempted; this, of course, is dependent on ex post analyses of the particular application. After briefly describing each model in the next section, they are compared first in a descriptive manner and then in terms of the multipliers and a case study. Importantly, an application has been chosen which has been analyzed at different times by all the models discussed in this paper. Finally, some additional points which need to be addressed by those seeking to choose an appropriate model are discussed. </p><a name="AN9509104058-3"></a><span class="medium-bold"><a id="hd_toc_15" title="2. Models " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">2. Models </font></strong></a></span><p class="body-paragraph">The models analyzed in this paper are the conventional input-output (IO) model, an integrated input-output+econometric model (IOE) and a simple computable general equilibrium (CGE) model. The models described below and used in the analysis in this paper are similar impact-type models, representative of those used in Queensland. They are single-region models; multi-region models, although more theoretically appealing, present a different and more complex modelling environment and are not available in all these model types in Queensland. Table 1 summarizes the main characteristics of each model. </p><a name="AN9509104058-4"></a><span class="medium-bold"><a id="hd_toc_19" title="2.1. IO Model " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">2.1. IO Model </font></strong></a></span><p class="body-paragraph">In the conventional IO model, the solution vector is obtained from the equation </p><p class="body-paragraph">delta x = (I-A)-1 delta f (1) </p><p class="body-paragraph">where Ax is the (direct and indirect) change in industry production levels as a result of an initial stimulus to final demand Af, and A represents the matrix of average input (purchase) coefficients by industry. In addition to output, it is common practice to calculate the value added, income and employment flow-ons from the initial expenditure, using average value added/output, household income/ output and employment/output coefficients respectively. </p><p class="body-paragraph">If only the productive industries are assumed to be endogenous, then the model is referred to as a 'Type I' model, as shown in Figure 1. However, if households are assumed to be an endogenous component of the economy selling labour and buying locally produced goods and services, the matrix A is extended (i.e. additional household rows and columns added) to include a 'household sector'. The most common model (type II) endogenizes 'extensive' income change. Here, it is assumed that all consuming households are homogeneous and employed. Therefore, if there is a stimulus to the economy, then the only place where industries can draw additional labour is from outside the region. </p><p class="body-paragraph">Additional household-induced effects which are sometimes added to the type II model include 'intensive' income changes from the marginal changes in income of employed persons (such as resulting from economies of scale, etc.) (type III) and 'redistributive' income changes as local unemployed people (on unemployment benefits) become employed (type IV). These are also shown in Figure 1. </p><p class="body-paragraph">The simplest method of solving these extended models is by the commodity-activity framework. Early work in this area of modelling households came from Miernyk et al. (1967), Tiebout (1969), Blackwell (1978) and Miyazawa (1976). More recently, the procedure is commonly associated with the work of Batey and Madden (see, for example, Batey & Madden, 1981; Batey et al., 1987; Batey & Weeks 1987). Another recent variation by Oosterhaven and Folmer (1985) and van Dijk and Oosterhaven (1986) makes use of vacancy chains, using matrices of transition probabilities to determine how vacancies are filled. The model of interest in this paper, however, is the familiar type II model. </p><a name="AN9509104058-5"></a><span class="medium-bold"><a id="hd_toc_33" title="2.2. IOE Model " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">2.2. IOE Model </font></strong></a></span><p class="body-paragraph">The concept of conjoining IO and econometric techniques, although not new, is becoming more common at the regional level; for example, the Hawaii demo-economic model (Joun & Conway, 1983) and the Washington (WPSM) (Conway, 1990) and Chicago (CREIM) models (Israilevich et al., 1994). National IOE models have been operational for many years, such as the INFORUM family of models (Almon, 1991). The aim is to retain the detailed sectoral disaggregation of the IO system and close it using a system of endogenous econometric relationships, generally expressed in elasticity form (Dewhurst & West, 1990). This closure forms the basis of the feedback mechanism between primary factors and final demand; also, because the econometric equations have a dynamic structure, the closure captures the response through time as the economy is subjected to external shocks. This dynamic structure' of the integrated model enables the temporal distribution of the flow-ons to be studied. More generally, the impact (short-term) multipliers, lag multipliers and dynamic (long-term) multipliers can be studied and the cumulative effects of dynamic impacts which occur over several years can be analyzed. </p><p class="body-paragraph">IOE models can and generally do have a wide range of specifications, depending on the region type (e.g. regional or national) and purpose of design (impact analysis, forecasting, demo-economic, income distributions, fiscal policy evaluations, etc.). The model type of interest in this paper is shown in Figure 2, which is an impact model with the emphasis on the household sector. This is the basic structure of the Queensland model used in this study, i.e. a single-region version of QUIP (West, 1994). It is not a forecasting model; instead, it retains the impact modelling structure of the IO model but this is extended to include medium-term dynamic impacts and projections. In addition, other variables of interest not included in the simple IO model are estimated, such as the total income, demographic indicators, labour force, taxation and government. </p><p class="body-paragraph">The model comprises a number of modules or blocks; in particular, the IO, demographic, labour, income, household consumption, price and government blocks. The relationships within and/or between each block are estimated econometrically using data from 1978-79 to 1989-90. All the sectoral relationships are derived from the value added, which determines the industry levels of wages and salaries and employment. The total gross state product also provides an input into the demographic block, which calculates various population-related statistics. </p><p class="body-paragraph">The total wages and salaries, employment and population statistics provide inputs into the labour block, which calculates the total labour force, unemployment level and the number receiving some form of government benefit, including social security and unemployment benefits. The income block calculates the total income, income taxes and other taxes and deductions; hence, it also calculates the total disposable income. Thus, it captures all household income, unlike the conventional type II IO model, which only captures extensive income changes. This is used to estimate the total household expenditure via a conventional dynamic Friedman-type consumption function, which in turn is used to estimate a set of demand functions by major commodity groups. These are transformed into household expenditure by industry category, to provide the private consumption expenditure component of the final demand. </p><p class="body-paragraph">Although integrated models normally endogenize other variables, such as government revenue and expenditure (including capital), these are set exogenously in this study, in line with the IO model. Here, other final demand is projected independently and this is normally where impacts are introduced as per the conventional IO model. </p><a name="AN9509104058-6"></a><span class="medium-bold"><a id="hd_toc_45" title="2.3. CGE Model " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">2.3. CGE Model </font></strong></a></span><p class="body-paragraph">The CGE model is another model which is receiving increased attention at the regional level. In contrast to the previous model, the CGE model is an optimization model, i.e. it provides the 'optimal' solution mix of endogenous variables in response to an exogenous shock. Also, unlike the IO model, which is demand driven, the CGE model contains explicit supply constraints, usually embedded in a neoclassical framework. A simple CGE structure is shown in Figure 3. </p><p class="body-paragraph">Essentially, unlike the IO and IOE models, each flow in the IO table is split into two components, i.e. quantity and price. These are often expressed in the form of composite goods and prices; for example, intermediate inputs are a composite of locally produced and imported goods (Armington, 1969), and primary inputs (such as labour) are composites of a number of more homogeneous primary factors. Prices are determined endogenously. Each intermediate column of the IO table is described by a multi-level or nested production function, usually Leontief, Cobb-Douglas or CES; and inputs into the production system are determined by a cost minimization or utility/output maximization procedure. Thus, for example, intermediate demands would generally be described by a Leontief production function (because of the structure of the IO table), while primary factor demands would be determined by a Cobb-Douglas or CES function. Unlike the IO model, which achieves equilibrium in supply and demand quantities only, the solution to the CGE model is given through both quantities and prices. </p><p class="body-paragraph">The model used in this study is a conventional Johansen-type model (see, for example, Dixon et al., 1992) with Keynesian-type closure. Cobb-Douglas-type functions are used, as a result more of lack of knowledge about substitution elasticities than theoretical considerations. While the import levels are endogenously determined, the import and export prices are assumed as given (the 'small country' assumption). Other final demand (excluding household consumption) quantities are exogenously set. The specification is designed to fit as closely as possible to the IO and IOE models, with the only real difference being the forced market clearing closure mechanism and the form of the production functions. </p><a name="AN9509104058-7"></a><span class="medium-bold"><a id="hd_toc_53" title="3. Model Comparison " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">3. Model Comparison </font></strong></a></span><p class="body-paragraph">Of the three models considered, the conventional IO model is the simplest, both in construction and implementation. It is also reasonable to say that it is the most naive in terms of assumptions and limitations. First, IO models are fixed input coefficient, static models. In practice, the assumption of constant coefficients may not be acceptable, bearing in mind that the economy is a changing entity. This can create problems if the time-frame between the base year and exogenous shock is excessive, exacerbated by the time-lag involved in the table construction process. Here, we are concerned with changes over time, mainly resulting from relative price changes, technological change and changing returns to scale. This can only be handled satisfactorily in a dynamic framework. </p><p class="body-paragraph">Secondly, the assumption of linearity in the production system is a valid criticism in many instances. It implies a strict proportional relationship between input coefficients and output. This may be regarded as being acceptable in some producing industries but it is more questionable in the household sector, where income coefficients are average propensities, employment coefficients reflect average labour productivity rates and household consumption is determined by average expenditure patterns. While this can be reduced through the use of type Ill or type IV models, the coefficients are still average propensities of marginal income groups. This probably explains in part the lack of widespread use of these type III and type IV extended models. Although there have been various attempts to model non-linearities in the input coefficients (see, for example, Evans, 1954; Lahiri, 1976; Hamilton & Pongtanakron, 1983), non-linear IO has failed to achieve widespread use. The preferred approach is to handle non-linearities in a CGE or IOE framework. </p><p class="body-paragraph">Thirdly, the IO model has no supply-side constraints, i.e. it assumes an infinitely elastic supply on inputs, including labour, into the production process. Thus, price has no role to play. There is no market feedback mechanism between primary factors (value added) and final demands in the simple IO model. However, the IOE and CGE models do include this feedback loop, through it is achieved in different ways: the IOE feedback loop is via a set of econometric equations, while the CGE model uses the price mechanism and market clearing assumptions. </p><p class="body-paragraph">It is reasonable to say that the IO model overestimates the static flow-on effects of an impact or shock to the economy, particularly with respect to the income and employment effects, as a result of the use of average fixed coefficients. However, it can also underestimate some flow-on effects because it ignores government and capital expenditure-induced effects, as well as demographic effects of population change. It also does not consider resource supply implications. </p><p class="body-paragraph">It is for these reasons that alternative and, by necessity, more complex models are being designed and built. Nevertheless, IO still provides a useful purpose as a descriptive tool of the economy in question, and is certainly indispensable as a base for many so-called extended models and more complex modelling procedures. </p><p class="body-paragraph">The main advantage of the integrated IOE model is that it uses econometric methods to track the relationships between key variables over time. This provides a dynamic structure in which the IO model, starting in the base year, is 'rolled over' on an annual basis, taking into account the marginal changes in the input coefficients, income, employment, household consumption and various other related (such as demo-economic) variables. This eliminates the static and 'out-of-date' criticisms often levelled at the IO model, in that the IO table (together with all the other variables) is updated to the relevant time period, prior to the shock being introduced. The updating procedure is performed in different ways in different models; for example, some use econometric procedures (as in the Washington model), while some use relative accounting price changes (as in the Queensland model). The latter comes about through the supply price equation </p><p class="body-paragraph">delta p= delta v(I-A)<sup>-1</sup>
</p><p class="body-paragraph">where p and v are vectors of industry accounting prices and value added per unit of output, respectively, in a similar manner to the CGE model. If the model is linked into a national forecasting model (as is the Washington model), it can also be used in a regional forecasting mode. </p><p class="body-paragraph">The closure mechanism generally comprises a system of dynamic econometric equations. Thus, the model is dynamic in the sense of an econometric model, not to be confused with Leontief's original dynamic IO model (Leontief, 1953). It attempts to plot the dynamic time path of the economy, which may be in a continuous state of disequilibrium, so capturing the short-term fluctuations in consumption, investment, labour, etc., in response to fluctuations in the underlying business cycle, or as a result of external shocks to the system. It does not necessarily assume perfect knowledge in the market-place. </p><p class="body-paragraph">Many of the relationships are non-linear and are often expressed in elasticity form. This is particularly true of the labour, income and final demand expenditure components. Thus, industries can increase output in the short-term without corresponding proportional increases in wage costs and employment, and not all the increased income is spent on consumption. However, intermediate industrial inputs are still usually depicted by the linear Leontief production function, as in the conventional IO and CGE models, as a result of the assumption of non-substitutability between these inputs. </p><p class="body-paragraph">IOE models can be either impact or forecasting models. In the case of them being impact models, they are really just a more sophisticated version of the IO model, albeit one with non-linear and dynamic properties. In the case of them being forecasting models, they can be integrated into a national forecasting framework with feedback links or, at a simpler level, utilize national forecasts of exogenous variables. </p><p class="body-paragraph">It has been suggested that the IOE model provides a stable platform for analysis (Israilevich et al., 1993), in a similar vein to the IO model. It appears to be relatively insensitive to data errors (at least in the IO table) but is still subject to the criticism that the model lacks adequate supply-side specification (Beaumont, 1990). </p><p class="body-paragraph">CGE models, similarly to IOE models, can have a wide range of specifications. Nevertheless, IO and CGE models are close relatives. IO can be regarded as a special case of the CGE model, particularly under the 'small country' assumption (Dervis et al., 1982; McGregor & Swales, 1994). The smaller the region is relative to the rest of the world, the more open will be the economy. For most commodities, local producers in a small region will not greatly affect the supply or prices outside the region. Importers and exporters will generally be price-takers and there will be a perfectly elastic supply of imports, including labour, in the long term. Thus, relative local prices will tend to gravitate to the relative foreign price levels. </p><p class="body-paragraph">It should also be noted that, in practice, the distinction between IOE and CGE models is becoming more blurred. For example, IOE models now often incorporate price responses into product and factor demands, as in the INFORUM models (Almon, 1991). However, CGE models assume full market clearing, whereas IOE models assume imperfect knowledge of product and factor markets. In some cases, IOE models are constrained to some long-term equilibrium relationship, but the emphasis is on tracking the short-term disequilibrium adjustments over time. </p><p class="body-paragraph">In practice, of course, we would expect some differences between the models. The reason for choosing a CGE model is the facility to impose capacity constraints, particularly in the short-term; in such cases, it has been demonstrated that IO and CGE can give quite different results (see, for example, Harrigan et al., 1991). The sensitivity of the model to the closure rules chosen (such as neoclassical, Keynesian, neo-Keynesian or Johansen) can also have a large effect, as demonstrated by Rattso (1982). Under neoclassical and Johansen closure, fixed employment is assumed, so the primary effect is to alter the composition of output, and increased demand tends to 'crowd out' private investment, which is savings driven. Keynesian closure can raise output, since employment is allowed to increase, so there is room for expansion. Neo-Keynesian closure tends to redistribute income from wage recipients to profit recipients. </p><p class="body-paragraph">CGE models tend to be highly non-linear, which, when combined with the size scale of the models, can result in the need for complex solution algorithms. In some cases, models can be transformed into linear forms through logarithms (such as the ORANI model of Dixon et al. (1982)), which makes solving easier, but the majority need to resort to numerical methods, such as the Newton-Naphson method. Similarly to IO models, most CGE models are static, yet solve for a long-term equilibrium solution. Capital stocks are normally assumed to be fixed in the short-term but, by selecting an appropriate stock adjustment function, can be updated from period to period. Thus, the adjustment path over time is obtained, until full equilibrium is restored, but the time path does not refer to 'real' time. </p><p class="body-paragraph">In summary, the CGE model also has pros and cons. The main benefit is that the supply side (as well as the demand side) is now explicitly determined with full price response. How generally applicable neoclassical general equilibrium theory is, particularly at the regional level, depends largely on the strength of the small country assumption. Unfortunately, although the CGE model is probably more theoretically satisfying, in that it conforms better with micro-economic theory, when considered empirically, it may be a different story. Its implementation necessitates the specification of a large number of parameters and coefficients, which are generally not available. Therefore, 'best guess' values must be used, which injects a large unknown element into the model. Evidence suggests that this unknown factor can have a significant effect on the empirical results. </p><p class="body-paragraph">Nevertheless, it would seem that a new dimension is emerging for the CGE model. In the same way that national and regional IO has followed different research and development paths, CGE appears to be following a similar pattern. Certainly, the current research in the US on small multi-region CGE models is vastly different to that undergone at the national level. Given the correct conditions, regional CGE models will be a strong contender in the future. </p><a name="AN9509104058-8"></a><span class="medium-bold"><a id="hd_toc_91" title="4. Empirical Comparison " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">4. Empirical Comparison </font></strong></a></span><p class="body-paragraph">The IO table forms the basis of all the models discussed in this paper. The table used is the 1985-86 15-sector table for the State of Queensland, based on that published by the Queensland Government Statistician's Office (1990). The 15 sectors in the table are given in Table 2. The choice of sectors is determined by the availability of a consistent set of time series data for a number of variables, including value added, wages and salaries and employment, at the sectoral level. All the models are calibrated to the 1985-86 IO data and all the variables are measured in constant 1985-86 values. </p><p class="body-paragraph">In this section, an empirical comparison of the models is undertaken. First, the multipliers are derived, then a case study of the impact of tourist expenditures on the Queensland economy is described, which further highlights the differences between the models. It must be stressed that the following results should not be regarded as showing definitive differences between the three models but, instead, are indicative of the general differences which may be observed. Of course, different model structures and assumptions, and different applications would produce a different set of results. </p><a name="AN9509104058-9"></a><span class="medium-bold"><a id="hd_toc_97" title="4.1. Multipliers " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">4.1. Multipliers </font></strong></a></span><p class="body-paragraph">The value-added and employment multipliers are given in Tables 3 and 4 respectively. These multipliers represent the dollar change in value added per unit dollar increase in final demand expenditure of the sector in question, or the change in employment per million-dollar increase in final demand, in the case of the employment multipliers. </p><p class="body-paragraph">The tables provide five sets of multipliers derived from three models: (a) the conventional IO (type II) model, which is the most common IO model used at the regional level; (b) the integrated IOE model, with impact or short-term multipliers (i.e. the change in value added or employment which occurs in the base period 1985-86) and dynamic or long-term multipliers (the cumulative change over the period 1985-86 to 1990-91); (c) the CGE model under two closure scenarios. The first closure scenario holds capital supply fixed. This is the usual assumption in the short term, so is referred to here as the short-term model. The second CGE model allows capital supply to vary, as would occur over time, and is called the long-term model. </p><p class="body-paragraph">The value-added multipliers are given in Table 3. As expected, the long-term IOE model generally produces the largest multipliers, with an average value of 0.906 (or 109.4% of the average type II IO multiplier value), because of the additional induced demographic effects over time. Similarly, the short-term CGE model produces the smallest multipliers, with an average value of 0.601 (or 72.6% of the average IO multiplier), as a result of the constraints on capital supply. We would also expect the short-term IOE model to produce smaller multipliers than the IO model, because of the marginal rather than average household-induced relationships, and similarly with the long-term CGE model, except that the multipliers should get closer to those of the IO model as supply restrictions are relaxed. </p><p class="body-paragraph">However, there are also some significant differences in the distributions of the multipliers for each model. For example, public administration has the largest IO multiplier at 1.086, while the largest short-term IOE multiplier is 0.857 in mining. The largest CGE multiplier also occurs in public administration, with a value of 0.963 in the short-term and 1.071 in the long-term. It is also interesting to note that the short-term and long-term multipliers from the IOE model have different distributions, indicating that the temporal effects are significant and that the flow-on effects wash out at different rates over time in different sectors. A similar and probably more pronounced conclusion could be reached between the short-term and long-term CGE models, although obviously the absolute differences are not as great. The overall spread of values from the CGE model is also greater, as a result of the additional limited resource factor. Sectors which have limited access to capital will experience additional dampening effects, while sectors which can easily draw capital from other sectors will show relatively larger multiplier effects. </p><p class="body-paragraph">The employment multipliers are given in Table 4. In terms of relative sizes, the same general conclusions can be reached as for the value-added multipliers, except for greater variation in the IOE short-term and long-terms models, resulting from greater scope for labour productivity gains in the short term but greater growth in the long term caused by population growth over time. The CGE models give marginally greater relative multiplier values, as a result of the Keynesian-type closure. </p><p class="body-paragraph">While the multipliers form one basis for a comparison between models, they can be misleading in some ways, if taken as a general guide to the relative differences in any given application. The reason is that impact situations are usually more complex, involving multiple changes across a range of sectors. In the following section, a case study is used to highlight further the differences between the models. The analysis here uses a common data set to aid in the comparison. However, it must be stressed that this case study is only one of a number that could have been used and we again warn the reader of the danger of generalizing these results out of context. </p><a name="AN9509104058-10"></a><span class="medium-bold"><a id="hd_toc_111" title="4.2. Case Study " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">4.2. Case Study </font></strong></a></span><p class="body-paragraph">The impact scenario chosen for this study is the impact of expenditures by visitors (tourists) who stayed in commercial accommodation in Queensland in 1990-91 on the Queensland economy. This application has previously been studied in all the modelling frameworks considered in this paper (Bureau of Industry Economics, 1984; West & Bayne, 1990; West, 1993; Adams & Parmeter, 1993) and presents as near as possible a valid comparison of the three models, since visitor expenditures can be classified as final demand (final consumption expenditure of tourists) in all the models. While it is possible to analyze other types of impact, such as new or existing industries in the state or changes in taxation policy, a less direct comparison between the models would be obtained, because the analyses would, by necessity, be handled differently in each of the models. Obviously, some models are better suited to handling some applications than others. </p><p class="body-paragraph">The expenditure data by visitors were obtained from the Queensland Tourist and Travel Corporation's Queensland Visitor Survey and the National Centre for Studies in Travel and Tourism. They have been deflated to 1985-86 values, allocated to industry sectors and converted to producers' values, as shown in Table 5, to be compatible with the IO table. West and Bayne (1990) provided a detailed description of the allocation procedure. All the results are expressed in 1985-86 values. The implementation of the impact analyses in all the models is similar, in that the visitor expenditures are incorporated into the models as final demand shocks. The IOE long-term scenario is taken over the period 1990-91 to 1999-2000, by which time the incremental flow-on effects are negligible. </p><p class="body-paragraph">The results pertaining to the impact scenario on value added and employment are given in Tables 6 and 7 respectively. They show the percentage breakdowns over the industrial sectors of the total impact on the Queensland economy. For convenience, the IO model is used as a reference point, and the 'Index' row in each table refers to the indexed aggregate impacts relative to those obtained from this model. In line with a priori expectations, the total impacts derived from the conventional IO model are greater than those from the other static models, while those obtained from the short-term CGE model are the smallest. Obviously, the dynamic impacts from the IOE model will be the largest, because of the additional induced demographic effects (population growth over time). </p><p class="body-paragraph">In terms of the aggregate impacts, the estimated value added (Table 6) from the IO model is Aus$1899.31 million. The short-term IOE model's estimate is slightly less at $1824.49 million (or 96.1% of the IO model), while the short-term CGE model produces the lowest estimate of $1178.47 million, or only 62.1% of the IO model's value. The corresponding multipliers are 0.899, 0.864 and 0.558 respectively. Although not shown in the table, the proportion of the impact on the value added taken up by wages and salaries decreases from 57.6% with the IO model to 52.5% and 55.7% for the short-term and long-term IOE models, respectively, and to 29.1% and 39.1% for the short-term and long-term CGE models, reflecting these models' capabilities to redistribute resources and increase production without necessarily corresponding proportional increases in labour costs and employment. </p><p class="body-paragraph">This last point is brought out in Table 7, which shows that the employment flow-ons are much smaller relative to the IO model than those for value added, indicating the greater role played by marginal labour productivity changes. For example, the short-term IOE model produces only 72.4% of the employment impact of the IO model, while the short-term CGE model falls to 47.5% of the IO value. </p><p class="body-paragraph">The general distributions of the impacts across the industrial sectors agree more or less with expectations. The largest effects occur in those sectors directly affected by the visitor expenditure, i.e. recreation, trade, transport and food processing. In the long-term IOE model, the effects on recreation, transport and food processing are reduced as the flow-on effects disseminate throughout the economy over time. In the longer term, manufacturing and the other service sectors increase their shares of the proceeds. With employment, the rankings differ marginally but, overall, the distributions are much the same. Obviously, labour has a greater impact on labour-intensive industries (such as service industries) and less impact on manufacturing and other more capital-intensive industries. </p><p class="body-paragraph">Generally speaking, the IOE model produces relatively larger impacts in the manufacturing sectors and smaller impacts in the service sectors, particularly with respect to wages and employment. In other words, the service-type industries are better able to support the increase in tourist activity largely within existing resources, whereas manufacturing-type industries, which have more rigid production structures, respond in a manner closer to that of the Leontief production system. </p><p class="body-paragraph">However, the CGE model results in a much larger redistribution of resources among all the sectors in the economy; in particular (in this case), from machinery, appliances and equipment, metal products, non-metallic mineral products, construction and public administration (which all experience negative flow-on effects) to the sectors most affected by the boost in tourist activity, i.e. recreation activity, trade, transport, food processing and agriculture, and other manufacturing. This occurs as capital is drawn away from those sectors with more abundant and less efficient usage, going to those sectors in greater need in the short term. The variation across the sectors is less in the long-term model. In the longer term, as capital stocks adjust to the increased demand, manufacturing, construction and public administration regain some of their initial losses. </p><p class="body-paragraph">So far, the discussion has concentrated on the static aspect of the impacts. In the IOE model, in addition to the short-term effects, the temporal distribution of flow-on effects can be analyzed to give a greater understanding of the real long-term impacts on the economy. Table 8 shows that the cumulative flow-on effects resulting from continued (and lagged) population growth after the initial impacting period can more than double the flow-on effect recorded by the short-term model. The cumulative impact on the Queensland economy amounts to $4016.13 million to total value added and 167850 employed over the period 1990-91 to 1999-2000. The dynamic multipliers are now 1.901 for value added and 0.079 for employment. In this case, it can be seen that the additional flow-through impacts in later periods are not inconsequential. It should be noted that there is a larger difference between the short-term and long-term multipliers than that shown in Section 4.1. This is because of the non-linearities in the model, i.e. large initial shocks will produce relatively larger multiplier values than will small shocks, and will also take longer to wash out, resulting in larger cumulative multiplier values. </p><a name="AN9509104058-11"></a><span class="medium-bold"><a id="hd_toc_131" title="5. Further Considerations " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">5. Further Considerations </font></strong></a></span><p class="body-paragraph">The discussion so far points out obvious differences between the models. However, the comparison is not as simple as comparing structure, multiplier values or impacts. The environment in which the models operate also plays an important part. It is important to explore the theoretical foundations and empirical possibilities of these models in a wider framework, and to develop the capacity to evaluate and compare the models and their applications in different situations. Some questions that prospective users and developers need to address include the following. </p><a name="AN9509104058-12"></a><span class="medium-bold"><a id="hd_toc_135" title="5.1. Level of Complexity " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">5.1. Level of Complexity </font></strong></a></span><p class="body-paragraph">Generally speaking, the models cover a wide range of levels of complexity, even within model types. Integrated models can have relatively simple closure mechanisms or can contain extremely complex and extensive modular blocks or submodels, such as some of the INFORUM national models. They can be topdown, such as REMI (Treyz, 1993), or bottom-up models (QUIP). Similarly, CGE models can range from small city-based models (such as for Melbourne (Horridge, 1991), which contains about 1700 equations) to large national models, such as the Australian ORANI model (Dixon et al., 1982), which contains several million equations. </p><p class="body-paragraph">At what point is it wise, from a viewpoint of model (and application) validity, to move from a model at one level of complexity to another level, either higher or lower? This is a question of the marginal value (positive or negative) of additional model components and additional (possibly less reliable) data. There are obvious trade-offs involved. For instance, at what stage can we expect price effects to be significant, or demographic and socio-economic factors, or supply restrictions? Furthermore, how do we know that they are actually significant? Do we actually need to build the model to find out, or can we rely on the opinion of so-called experts? </p><a name="AN9509104058-13"></a><span class="medium-bold"><a id="hd_toc_141" title="5.2. Size and Type of Region " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">5.2. Size and Type of Region </font></strong></a></span><p class="body-paragraph">Do smaller regions require less complex models, or vice versa? In Australia, the move is towards community-based IO models. One would expect price and supply effects to be minimal for many commodities in small area economies (the so-called 'small country' assumption). As the size of a region increases, where does it become appropriate to consider the use of integrated IOE or CGE models to increase the comprehensiveness of the studies? Large rural areas presumably require modelling strategies different from those encompassing large metropolitan areas. At what stage does it become desirable to move from single-region models to multi-region models? What is the significance of the national-regional relationship, both in the use of national drivers for the regional economy and in terms of national data usage? Small-area models are invariably constructed from the bottom up, while large regions are more likely to make use of top-down construction techniques. </p><a name="AN9509104058-14"></a><span class="medium-bold"><a id="hd_toc_145" title="5.3. Type of Problem " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">5.3. Type of Problem </font></strong></a></span><p class="body-paragraph">The type of problem under consideration must be an important consideration in the choice of model. There is an unfortunate tendency by some modellers to push a particular type of model as a general or all-purpose model which encompasses all other models and applications. We live in an era where competition between models, particularly those based on large investment, encourages conflict rather than objectivity. Each model contains particular characteristics which make it more suitable for a given application than other models. Ideally, models should be tailor-made to the problem, whether it be studies of industry significance, fiscal policies, interregional and international trade, or demographic-economic and socio-economic relationships. Unfortunately, the time and cost associated with constructing models make this prohibitive, so each model involves a compromise. Nevertheless, broad rules can be determined. Is the study a forecasting or impact analysis? Should it be couched in an optimization framework or do we want to model the economy 'as it is'? Are the temporal effects important? What about spatial effects? Aggregation may be an important factor. </p><a name="AN9509104058-15"></a><span class="medium-bold"><a id="hd_toc_149" title="5.4. Assumptions " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">5.4. Assumptions </font></strong></a></span><p class="body-paragraph">Last, but not least, attention needs to be drawn to the analytical significance of the assumptions and characteristics of the model used. This can vary widely, even within model types. This basic statement of the strengths and weaknesses of the model, of the expected level of validity in its specific use, and of a 'consumer beware' nature would seem to be required as a professional responsibility. All too often, the degree to which this occurs would seem to vary inversely with the model size. </p><a name="AN9509104058-16"></a><span class="medium-bold"><a id="hd_toc_153" title="6. Conclusions " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">6. Conclusions </font></strong></a></span><p class="body-paragraph">The shift from the simple IO model to more comprehensive models is an inevitable movement, reflecting the trend in technology and technological development. These more comprehensive models are intended to encompass the IO model. At the same time, the results derived in this paper demonstrate that there are substantial differences between the models. Therefore, which set of results is the more reasonable? In some ways, it often comes down to personal prejudices. It becomes a question of balance between choosing a simple, easy to understand model versus a more complex, often difficult to understand and apply, model which is more theoretically appealing in some ways. It depends on the application. </p><p class="body-paragraph">While many have forecast the end of the simple IO model, saying that it is inadequate, it has proved surprisingly resilient. One area where IO still has the advantage over more sophisticated models is at the very small-region level (such as small towns), where IO really provides the only option to planners, despite its known limitations. </p><p class="body-paragraph">From a personal point of view, it would be advantageous to see the IOE and CGE frameworks move closer together and, at the same time, make the models more flexible and adaptable to local conditions without resorting to making each model application specific. This may mean making the models smaller in some ways rather than bigger (this is not necessarily a contradiction); large models tend to become a law unto themselves, and complexity is not necessarily a good guide to performance. </p><a name="AN9509104058-17"></a><span class="medium-bold"><a id="hd_toc_161" title="Table 1. Model characteristics " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">Table 1. Model characteristics </font></strong></a></span><a name="AN9509104058-18"></a><h4>IO (type II) </h4><pre class="ct">* Static
* Linear functions
* Fixed coefficients (fixed technology)
* No supply constraints (demand driven)
* No price effects
* Partial equilibrium (quantities only)
* Partial optimization
* Impact model
* Full employment (in region) but infinite elastic labour supply
* Wage income only
* Household expenditure determined by average expenditure patterns
* Intermediate and primary factor demands determined by Leontief
function
</pre><a name="AN9509104058-20"></a><h4>IOE </h4><pre class="ct">* Dynamic
* Non-linear functions
* Variable coefficients (technological change)
* Primarily demand driven, some supply constraints
* Some price effects
* Long-term equilibrium short-term disequilibrium
* Describes economy 'as is' (non-optimization model)
* Impact or forecasting
* Unemployment permitted
* Total (wage and non-wage) income
* Household expenditure determined by dynamic consumption function
* Intermediate demands determined by Leontief function
* Primary factor demands determined by econometric functions
</pre><a name="AN9509104058-22"></a><h4>CGE </h4><pre class="ct">* Static (some dynamics, e.g. capital stocks)
* Non-linear functions
* Usually fixed coefficients
* Demand and supply
* Full response price effects
* General equilibrium (prices and quantities)
* Optimization model
* Impact model
* Either full employment or unemployment
* Total (wage and non-wage) income
* Household expenditure determined by utility maximization
* Intermediate demands determined by Leontief function
* Primary factor demands determined by CES or Cobb-Douglas
function (cost minimization)
</pre><p class="body-paragraph">Characteristics are not necessarily mutually exclusive or exhaustive </p><a name="AN9509104058-24"></a><span class="medium-bold"><a id="hd_toc_177" title="Table 2. Queensland industrial sectors " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">Table 2. Queensland industrial sectors </font></strong></a></span><pre class="ct">Sector title Sector name
1 Agric. Agriculture, forestry and fishing
2 Mining Mining
3 Food Food processing
4 Wood Wood and paper manufacturing
5 Equip. Machinery, appliances and equipment
6 Metals Metal products
7 NonMet. Non-metallic mineral products
8 O'Mfg Other manufacturing n.e.i.
9 Elect. Electricity, gas and water
10 Const. Building and construction
11 Trade Wholesale and retail trade
12 Transp. Transport and communication
13 Finance Finance and business services
14 P. Admin. Public administration, defence and
community services
15 Rec. Recreation, personal and other services
</pre><a name="AN9509104058-26"></a><span class="medium-bold"><a id="hd_toc_181" title="Table 3. Value-added multipliers, 1985-86 " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">Table 3. Value-added multipliers, 1985-86 </font></strong></a></span><pre class="ct"> IO IOE CGE
Sector type II Short-term Long-term Short-term Long-term
Agric. 0.870(6) 0.809(4) 1.004(4) 0.531(12) 0.837(2)
Mining 0.928(3) 0.857(1) 1.125(2) 0.573(9) 0.80(5)
Food 0.882(5) 0.756(7) 0.924(8) 0.612(5) 0.803(6)
Wood 0.737(12) 0.588(13) 0.701(13) 0.599(7) 0.622(13)
Equip. 0.675(13) 0.551(14) 0.664(14) 0.53(11) 0.647(12)
Metals 0.861(7) 0.744(8) 0.918(9) 0.601(6) 0.745(8)
NonMet. 0.793(10) 0.699(10) 0.893(10) 0.542(10) 0.701(10)
O'Mfg 0.660(14) 0.670(12) 0.877(11) 0.444(14) 0.590(14)
Elect. 0.846(8) 0.766(6) 0.945(7) 0.529(13) 0.833(3)
Const. 0.834(9) 0.711(9) 0.959(6) 0.632(4) 0.741(9)
Trade 1.015(2) 0.855(2) 1.127(1) 0.770(2) 0.825(4)
Transp. 0.926(4) 0.801(5) 0.986(5) 0.677(3) 0.781(7)
Finance 0.526(15) 0.434(15) 0.547(15) 0.421(15) 0.528(15)
P. Admin. 1.086(1) 0.830(3) 1.094(3) 0.963(1) 1.071(1)
Rec. 0.779(11) 0.672(11) 0.818(12) 0.578(8) 0.650(11)
Mean 0.828 0.716 0.906 0.601 0.743
Index 1.000 0.865 1.094 0.726 0.898
Coeff. 0.166 0.161 0.178 0.212 0.174
variation
</pre><p class="body-paragraph">Numbers in parentheses denote the rank. </p><a name="AN9509104058-28"></a><span class="medium-bold"><a id="hd_toc_187" title="Table 4. Employment multipliers, 1985-86 (per million Australian dollars) " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">Table 4. Employment multipliers, 1985-86 (per million Australian dollars) </font></strong></a></span><pre class="ct"> IO IOE CGE
Sector type II Short-term Long-term Short-term Long-term
Agric. 40(4) 21(5) 39(4) 21(5) 39(3)
Mining 16(15) 22(3) 43(2) 1(15) 11(15)
Food 35(5) 20(6) 36(7) 25(6) 33(5)
Wood 34(6) 16(13) 27(13) 28(4) 29(6)
Equip. 26(9) 15(14) 26(14) 21(9) 25(9)
Metals 24(10) 19(8) 35(9) 13(11) 19(12)
NonMet. 24(11) 18(12) 34(10) 13(12) 20(11)
O'Mfg 21(13) 18(11) 33(11) 12(13) 18(13)
Elect. 18(14) 20(7) 36(8) 5(14) 18(14)
Const. 30(8) 19(9) 37(6) 21(8) 26(8)
Trade 51(1) 22(1) 43(1) 41(2) 43(2)
Transp. 33(7) 21(4) 38(5) 23(7) 28(7)
Finance 24(12) 11(15) 21(15) 20(10) 24(10)
P.
Admin. 51(2) 22(2) 42(3) 46(1) 50(1)
Rec. 42(3) 18(10) 32(12) 34(3) 37(4)
Mean 31.267 18.800 34.800 22.000 28.000
Index 1.000 0.601 1.113 0.704 0.896
Coeff. 0.340 0.156 0.176 0.544 0.369
variation
</pre><p class="body-paragraph">Numbers in parentheses denote the rank. </p><a name="AN9509104058-30"></a><span class="medium-bold"><a id="hd_toc_193" title="Table 5. Visitor expenditures in Queensland[a] " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">Table 5. Visitor expenditures in Queensland[a] </font></strong></a></span><pre class="ct">Sector 1985-86 1986-87 1987-88 1988-89 1989-90 1990-91
Food 195.96 224.04 276.86 312.48 299.22 340.55
O'Mfg 133.09 152.91 199.20 233.55 222.13 247.53
Trade 156.58 180.08 230.96 268.25 254.58 285.93
Transp. 70.56 85.72 114.70 139.88 122.17 140.45
Rec. 603.38 700.42 891.64 1034.96 966.72 1098.14
Total 1149.56 1343.17 1713.18 1989.12 1864.82 2112.60
</pre><p class="body-paragraph">a In millions of Australian dollars at 1985-86 dollar values. </p><a name="AN9509104058-32"></a><span class="medium-bold"><a id="hd_toc_199" title="Table 6. Percentage distribution of value-added impacts, 1990-91 " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">Table 6. Percentage distribution of value-added impacts, 1990-91 </font></strong></a></span><pre class="ct"> IO IOE CGE
Sector type II Short-term Long-term Short-term Long-term
Agric. 7.51(5) 7.40(5) 6.61(6) 11.05(4) 9.98(4)
Mining 2.62(10) 3.13(10) 2.57(10) 2.77(9) 2.70(9)
Food 7.57(4) 7.44(4) 6.14(7) 10.32(5) 9.45(5)
Wood 1.70(11) 1.87(11) 2.51(11) 1.38(10) 1.51(10)
Equip. 0.62(13) 0.66(13) 0.90(12) -0.37(12) -0.06(11)
Metals 0.45(14) 0.47(14) 0.48(14) -0.78(13) -0.43(13)
NonMet. 0.29(15) 0.29(15) 0.30(15) -0.25(11) -0.10(12)
O'Mfg 5.02(7) 4.94(6) 3.60(9) 7.23(6) 6.51(7)
Elect. 3.48(9) 3.29(9) 4.80(8) 4.26(8) 4.12(8)
Const. 0.81(12) 0.73(12) 0.81(13) -5.62(14) -3.76(14)
Trade 21.80(2) 21.79(2) 25.19(1) 27.7(2) 26.27(2)
Transp. 10.42(3) 10.13(3) 9.50(4) 11.21(3) 10.98(3)
Finance 6.06(6) 4.53(7) 8.09(5) 6.69(7) 6.71(6)
P. Admin. 3.93(8) 4.41(8) 10.04(3) -17.04(15) -10.64(15)
Rec. 27.71(1) 28.93(1) 18.45(2) 41.13(1) 36.76(1)
Total
impac[a] 1899.31 1824.49 4016.13 1178.47 1356.85
Multip-
lier
0.899 0.864 1.901 0.558 0.642
Index 1.000 0.961 2.115 0.621 0.714
</pre><p class="body-paragraph">a In millions of Australian dollars at 1985-86 dollar values. </p><p class="body-paragraph">b Multiplier values per unit change in final demand (visitor expenditure). </p><p class="body-paragraph">Numbers in parentheses denote the rank. </p><a name="AN9509104058-34"></a><span class="medium-bold"><a id="hd_toc_209" title="Table 7. Percentage distribution of employment impacts, 1990-91 " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">Table 7. Percentage distribution of employment impacts, 1990-91 </font></strong></a></span><pre class="ct"> IO IOE CGE
Sector type II Short-term Long-term Short-term Long-term
Agric. 7.77(3) 7.65(3) 6.88(5) 10.40(3) 9.47(3)
Mining 0.41(13) 0.49(12) 0.40(13) -0.12(10) 0.13(10)
Food 5.06(6) 4.98(5) 4.09(7) 7.29(4) 6.41(6)
Wood 1.76(9) 1.93(9) 2.75(8) 1.10(8) 1.40(8)
Equip. 0.48(12) 0.51(11) 0.73(11) -0.86(12) -0.33(12)
Metals 0.23(14) 0.25(14) 0.26(14) -1.12(13) -0.56(13)
NonMet. 0.16(15) 0.15(15) 0.16(15) -0.42(11) -0.20(11)
O'Mfg 3.06(8) 3.00(8) 2.13(9) 4.63(7) 3.99(7)
Elect. 1.05(10) 1.00(10) 1.55(10) 0.93(9) 1.04(9)
Const. 0.54(11) 0.49(13) 0.56(12) -6.89(14) -3.99(14)
Trade 24.61(2) 24.56(2) 29.78(1) 31.89(2) 29.59(2)
Transp. 6.97(4) 6.76(4) 6.45(6) 6.43(6) 6.70(5)
Finance 6.66(5) 4.96(6) 9.45(3) 6.73(5) 7.07(4)
P.
Admin. 4.03(7) 4.52(7) 11.40(4) -25.16(15) -14.02(15)
Rec. 37.22(1) 38.78(1) 23.42(2) 65.19(1) 53.31(1)
Total
impact
[a] 90.12 65.27 167.85 42.80 54.91
Multip-
lier 0.043 0.031 0.079 0.020 0.026
Index 1.000 0.724 1.863 0.475 0.604
</pre><p class="body-paragraph">a In thousands. </p><p class="body-paragraph">b Multiplier values per unit change in final demand (visitor expenditure). </p><p class="body-paragraph">Numbers in parentheses denote the rank. </p><a name="AN9509104058-36"></a><span class="medium-bold"><a id="hd_toc_219" title="Table 8. IOE dynamic impacts, 1990-91 to 1999-2000 " href="http://web.ebscohost.com.ezproxy.lib.ucalgary.ca/ehost/detail?vid=4&hid=3&sid=b872c3d4-384b-49c2-889f-08c5e88a0e3a%40SRCSM1#toc"><strong><font color="#0033ff" size="2">Table 8. IOE dynamic impacts, 1990-91 to 1999-2000 </font></strong></a></span><pre class="ct">Year Value added[a] Employment
1990-91 1824.49 65.27
1991-92 391.93 31.66
1992-93 369.19 14.03
1993-94 349.28 13.23
1994-95 314.75 12.10
1995-96 267.43 10.47
1996-97 211.45 8.48
1997-98 151.83 6.31
1998-99 93.82 4.14
1999-2000 41.97 2.15
Total 4016.13 167.85
</pre><p class="body-paragraph">a In millions of Australian dollars at 1985-86 dollar values. </p><p class="body-paragraph">b In thousands. </p><p class="body-paragraph">CHART: Figure 1. Type I-IV IO models. </p><p class="body-paragraph">CHART: Figure 2. Simplified IOE model. </p><p class="body-paragraph">CHART: Figure 3. Simplified CGE model. </p><a name="AN9509104058-38"></a><h4>References </h4><p class="body-paragraph">Adams, P. D. & Parmeter, B. R. (1993) The Medium-term Significance of International Tourism for the State Economies (Canberra, Bureau of Tourism Research). </p><p class="body-paragraph">Almon, C. (1991) The INFORUM approach to interindustry modelling, <strong><em>Economic Systems Research</em></strong>, 3, pp. 1-7. </p><p class="body-paragraph">Armington, P. (1969) A theory of demand for products distinguished by place of production, IMF Staff Papers, 16, pp. 159-178. </p><p class="body-paragraph">Batey, P. W. J. & Madden, M. (1981) Demographic-economic forecasting within an activity-commodity framework: some theoretical considerations and empirical results, Environment and Planning A, 13, pp. 1067-1083. </p><p class="body-paragraph">Batey, P. W. J., Madden, M. & Weeks, M. J. (1987) Household income and expenditure in extended input-output models: a comparative theoretical and empirical analysis, Journal of Regional Science, 27, pp. 341-356. </p><p class="body-paragraph">Batey, P. W. J. & Weeks, M. J. (1987) An extended input-output model incorporating employed, unemployed and in-migrant households, Papers of the Regional Science Association, 62, pp. 93-115. </p><p class="body-paragraph">Beaumont, P.M. (1990) Supply and demand interaction in integrated econometric and input-output models, International Regional Science Review, 13, pp. 167-181. </p><p class="body-paragraph">Blackwell, J. (1978) Disaggregation of the household sector in regional input-output analysis: some models specifying previous residence of workers, Regional Studies, 12, pp. 367-377. </p><p class="body-paragraph">Bureau of Industry Economics (1984) Tourist expenditure in Australia, Research Report 16 (Canberra, Bureau of Industry Economics). </p><p class="body-paragraph">Dewhurst, J. H. Ll. & West, G. R. (1990) Closing interregional input-output models with econometrically determined relationships, in: L. Anselin & M. Madden (eds), New Directions in Regional Analysis: Integrated and Multiregional Approaches (London, Pinter). </p><p class="body-paragraph">Dixon, P. B., Parmenter, B. R., Sutton, J. & Vincent, D. P. (1982) ORANI: A Multisectoral Model of the Australian Economy (Amsterdam, North-Holland). </p><p class="body-paragraph">Dixon, P. B., Parmenter, B. R., Powell, A. & Wilcoxen, P. (1992) Notes and Problems in Applied General Equilibrium Economics (Amsterdam, Elsevier). </p><p class="body-paragraph">Conway, Jr, R. S. (1990) The Washington projection and simulation model: a regional interindustry econometric model, International Regional Science Review, 13, pp. 167-181. </p><p class="body-paragraph">van Dijk, J. & Oosterhaven, J. (1986) Regional impacts of migrants' expenditures: an input-output vacancy chain approach, in: P. W. J. Batey & M. Madden (eds), Integrated Analysis of Regional Systems (London, Pion). </p><p class="body-paragraph">Dervis, K., De Melo, J. & Robinson, S. (1982) General Equilibrium Models for Development Policy (Cambridge, Cambridge University Press). </p><p class="body-paragraph">Evans, W. D. (1954) Input-output computations, in: T. Barna (ed), The Structural Interdependence of the Economy (New York, Wiley). </p><p class="body-paragraph">Government Statistician's Office (1990) Experimental Input-Output Tables Queensland 1985-86 (Brisbane, Government Statistician's Office). </p><p class="body-paragraph">Hamilton, T. R. & Pongtanakorn, C. (1983) The economic impact of irrigation development in Idaho: an application of marginal input-output methods, Annals of Regional Science, 17, pp. 60-71. </p><p class="body-paragraph">Harrigan, F., McGregor, P., Dourmashkin, N., Swales, K. & Yin, Y. P. (1991) The sensitivity of output multipliers to alternative technology and factor market assumptions: a computable general equilibrium analysis, in: J. Dewhurst, R. Jensen & G. Hewings (eds), Regional Input-Output Modelling: New Developments and Interpretations (Aldershot, Avebury). </p><p class="body-paragraph">Horridge, J. M (1991) A general equilibrium model of Australia's second largest city, IAESR Working Paper 2/1991, Institute of Applied Economic and Social Research, University of Melbourne, Melbourne. </p><p class="body-paragraph">Israilevich, P. R., Hewings, G. J. D., Sonis, M. & Schindler, G. R. (1994) Forecasting structural change with a regional econometric input-output model, paper presented to the Regional Science Association International British and Irish Section Meeting, Dublin. </p><p class="body-paragraph">Israilevich, P. R., Mahidhara R. & Hewings, G. J. D. (1993) Forecasting with regional input-ouput tables, REAL Discussion Paper 93-T-1, University of Illinois, Champaign-Urbana, IL. </p><p class="body-paragraph">Joun, R. Y. P. & Conway, Jr, R. S. (1983) Regional economic-demographic forecasting models: a case study of the Washington and Hawaii models, Socio-Economic Planning Sciences, 17, pp. 345-353. </p><p class="body-paragraph">Lahiri, S. (1976) Input-output analysis with scale dependent coefficients, Econometrica, 44, pp. 947-961. </p><p class="body-paragraph">Leontief, W. (1953) Dynamic analysis, in: W. Leontief et al. (eds), Studies in the Structure of the American Economy (New York, Oxford University Press). </p><p class="body-paragraph">McGregor, P. G. & Swales, J. K. (1994) An investigation into a neo-classical interpretation of regional input-output analysis, Strathclyde Papers in Economics 94/1, University of Strathclyde, Glasgow. </p><p class="body-paragraph">Miernyk, W. H., Bonner, E. R., Chapman, Jr, J. H. & Schellhamer, K. (1967) Impact of the Space Program on a Local Economy: An Input-Output Analysis (Morgantown, WV, West Virginia University Press). </p><p class="body-paragraph">Miyazawa, K. (1976) Input-output Analysis and the Structure of Income Distribution (Berlin, Springer). </p><p class="body-paragraph">Oosterhaven, J. & Folmer, H. (1985) An interregional labour market model incorporating vacancy chains and social security, Papers of the Regional Science Association, 58, pp. 141-155. </p><p class="body-paragraph">Rattso, J. (1982) Different macroclosures of the original Johansen model and their impact on policy evaluation, Journal of Policy Making, 4, pp. 85-97. </p><p class="body-paragraph">Tiebout, C. M. (1969) An empirical region input-output projection model: the state of Washington, 1980, Review of Economics and Statistics, 51, pp. 334-340. </p><p class="body-paragraph">Treyz, G.I. (1993) Regional Economic Modelling: A Systematic Approach to Economic Forecasting and Policy Analysis (Boston, MA, Kluwer). </p><p class="body-paragraph">West, G. R. (1993) Economic significance of tourism in Queensland, Annals of Tourism Research, 20, pp. 490-504. </p><p class="body-paragraph">West, G. R. (1994) The Queensland impact and projection model: the household sector, <strong><em>Economic Systems Research</em></strong>, 6, pp. 363-383. </p><p class="body-paragraph">West, G. R. & Bayne, B. A. (1990) Visitor Expenditure Analysis Model: An Economic Assessment Modelling System of Tourism Activity in Queensland (Brisbane, Queensland Tourist and Travel Corporation). </p><p>~~~~~~~~</p><p class="body-paragraph">By GUY R. WEST </p><p class="body-paragraph"></p><p class="body-paragraph">Guy R. West, Regional and Urban Economics Research Unit, Department of Economics, University of Queensland, St Lucia, Queensland 4072, Australia. </p>
扫码加好友，拉您进群
全部版块
我的主页
收藏
