|Title||Tourism, income, and jobs : improving the measurement of regional economic impacts of tourism|
|Source||Wageningen University. Promotor(en): Wim Heijman, co-promotor(en): Jack Peerlings. - Wageningen : Wageningen University - ISBN 9789054723509 - 188|
Agricultural Economics and Rural Policy
|Publication type||Dissertation, internally prepared|
|Keyword(s)||tourism - economic impact - income - employment - regional economics - models - tourism impact - visitor impact - toerisme - economische impact - inkomen - werkgelegenheid - regionale economie - modellen - impact van toerisme - impact van bezoekers|
|Categories||Tourism / Economics (General)|
Tourism can have a broad range of impacts, including impact on the economy, on the natural and built environment, on the local population, and on visitors themselves. This PhD thesis discussed the measurement of regional economic impacts of tourism, including impacts on output, value added, and employment caused by visitor expenditure. The focus was on the choice between models that can be used to calculate these regional economic impacts and the data requirements, usage, and further development of one specific model; the Input-Output (I-O) model.
The starting point of an I-O model is final demand, which is the value of goods and services bought by final users for the direct fulfilment of their needs and wants. In tourism this refers to the value of the goods and services bought by visitors. Final demand brings about a chain of production. First, goods and services that are part of final demand need to be produced. This requires production factors (i.e., capital and labour) as well as intermediate inputs. These intermediate inputs also need to be produced, again requiring production factors and a subsequent ‘level’ of intermediate inputs. Combining final demand and all ‘levels’ of intermediate inputs, an I-O model enables calculation of the total output required to satisfy final demand. An I-O model can be an appropriate choice for an economic impact analysis (EIA) in the following context:
Relevant data exist on (the change of) final demand, i.e. visitors expenditure per industry;
There is an I-O table on the appropriate spatial scale;
Impacts are analysed of (a change in) final demand;
The assumption ‘no scarcity of production factors’ is acceptable (which implies there are no relative prices changes, input substitution and redistribution of production factors among industries);
The assumption ‘no productivity changes’ is acceptable (final demand changes do not lead to productivity changes, e.g. employees working longer, harder or more efficiently);
There is interest in indirect impacts on output, value added, income and/or employment per industry, while there is little interest in induced impacts, spatial considerations, temporal consideration, social impacts, environmental impacts, and economic externalities. Indirect impacts are impact generated by the production of intermediary inputs.
Not all EIAs in tourism will be carried out within such a context. In some EIAs one or more of these conditions are not met. The overall goal of this research was to improve the measurement of the regional economic impacts of tourism by
Establishing criteria based on which an appropriate economic impact model can be selected for an EIA in tourism and;
Providing solutions for those situations where
an Input Output table on the appropriate spatial scale is not available;
and/or analysis is required of different ‘shocks’ than final demand changes;
and/or the assumption ‘no scarcity of production factors’ cannot be accepted (which implies there can be relative prices changes, input substitution and/or redistribution of production factors among industries);
and/or the assumption ‘no productivity changes’ cannot be accepted
without introducing prohibitive complexity and data demands to an I-O model.
This overall objective was subdivided into the following specific objectives:
Provide an overview and evaluation of the criteria for the selection of economic impact models.
Provide an explanation for the sign of the difference between regional I-O coefficients calculated between two alternative location quotient (LQ) methods, for all combinations of demanding and supplying industries.
To analyze medical tourism’s state-level economic impacts in Malaysia.
Address the limitations of I-O models and ‘upgrade’ the I-O model, without introducing the complexity and data collection costs associated with a full Computable General Equilibrium (CGE) model.
To include labour productivity changes, caused by a change in final demand in the tourism industries, into a non-linear I-O (NLIO) model.
Each of these specific objectives was discussed in a separate chapter. Chapter 2 discussed criteria to choose between economic impact models, when carrying out an EIA in tourism. Based on the literature review 52 potential criteria were identified. After consulting experts in tourism and/or EIAs 24 of these 52 criteria were identified as essential. These essential criteria were used to compare the five economic impact models that are most used in EIAs in tourism; Export Base, Keynesian, Ad hoc, I-O, and CGE models. The results show that CGE models are the preferred choice for many of the criteria. Their detail and flexibility potentially lead to more realistic outcomes. However, CGE models do not ‘score’ high on criteria related to transparency, efficiency, and comparability. Multiplier models (Keynesian, Export Base and Ad Hoc) score high on these criteria, but the realism of their results is limited. I-O models are an “in-between” option for many criteria, which explains their extensive usage in EIAs in tourism. Nonetheless, I-O models have some important disadvantages, most notably their strong assumptions (‘no scarcity of production factors’ and ‘no productivity changes’), which limit the realism of their results. Although the choice of a model should always depend on the specific context of each EIA, the general conclusion is that an ‘ideal model’ for many applications could be found somewhere in between I-O and CGE. The challenge, however, is to extend the I-O model, while keeping the complexity and data demands to a minimum. This conclusion provided the motivation for the application and further development of an NLIO model, in chapters 5 and 6.
Both I-O and NLIO models require the existence of an I-O table on the appropriate spatial scale. For a regional I-O analysis an I-O table needs to be available for the specific region. When such a table is not available, it can be created using LQ methods. The four most used LQ methods are Simple Location Quotient, Cross Industry Location Quotient, Round’s Location Quotient, and Flegg’s Location Quotient (FLQ). The size of the regional I-O coefficients (RIOCs), which are derived from a regional I-O table, directly influences the results of an EIA. An over- or underestimation of RIOCs can lead to over- or underestimation of economic impacts. It is therefore very important to understand the differences between LQ methods and the consequences for the RIOCs. Chapter 3 showed that the ranking in size of the RIOCs, generated by the four LQ methods, depends on the J-value of demanding industries (output of industry j on regional level divided by output of industry j on national level). The conditions were calculated under which FLQ, the LQ method which was developed to avoid overestimation, leads to the lowest RIOCs47. Although this chapter does not provide a complete answer to question which LQ method to use in an EIA it does show that a choice for the FLQ method could be motivated by the wish to arrive at a careful estimate of regional economic impacts and to avoid or limit overestimation.
In chapter 4 the FLQ method was used to create RIOCs for nine Malaysian states. These RIOCs were used to calculate state-level economic impacts of medical tourism based on regional I-O models. It was shown that impacts related to non-medical expenditure of medical tourists (USD 273.7 million) are larger than impacts related to medical expenditure (USD 104.9 million) and that indirect impacts (USD 95.4 million) make up a substantial part of total impacts (USD 372.3 million). Data limitations implied that strong assumptions were required to estimate final demand by medical tourists, specifically regarding their non-medical expenditure and allocation of this expenditure to industries of the I-O model.
In chapter 5 the I-O model was “upgraded” to a NLIO model, by replacing the Leontief production function, underlying the I-O model with a Constant Elasticity of Substitution (CES) production function. Thereby the main drawback of the I-O model, the need to accept the assumption of ‘no scarcity of production factors’ was thus eliminated. The analysis performed showed that, for large changes of final demand, an NLIO model is more useful than an I-O model because relative prices changes are likely, leading to substitution and redistribution of production factors between industries. The NLIO takes this into account. Impacts can be higher or lower than in the I-O model, depending on assumptions about capacity constraints, production factor mobility and substitution elasticities. Relative price changes, substitution, and redistribution are less likely for a small change of final demand. In that case most realistic results are achieved by accepting assuming ‘no scarcity of production factors’, as in case of the I-O model. To analyze impacts of other types of ‘shock’ than final demand changes, such as a change of subsidies, an I-O model is not an option. A more flexible model is required, such as a NLIO model. A NLIO model requires additional assumptions and/or data. First, researchers need to choose the appropriate assumption regarding the functioning of factor markets and production factor mobility between industries. Second, the NLIO model forces the researcher to specify the substitution elasticities, instead of implicitly assuming an elasticity of zero (as in the I-O model). Compared to a CGE model, the NLIO model offers the advantage that it is not dependent on the existence of a Social Accounting Matrix (SAM) on the appropriate spatial scale, while the production structure is identical. Furthermore, using a CGE model introduces additional complexity as it requires the specification of the relationships between income and final demand, including issues such as income transfers and income taxation.
In chapter 6 labour productivity changes, that result from final demand changes were included into the NLIO model, thereby integrating productivity changes. A differentiation was made between real and quasi productivity changes and productivity changes for core and peripheral labour. Real productivity changes (changes that enable the production of more output per unit of labour) were integrated by introducing Factor Augmenting Technical Change (FATC) based on an endogenous specification. Quasi productivity changes (substitution of labour by other inputs which automatically leads to higher labour productivity) were already integrated into the NLIO based on the CES production function. The differentiation between core and peripheral labour was integrated by a smaller potential change of FATC for peripheral labour, implying less room for productivity changes. The NLIO model with and without FATC was applied to calculate impacts of a 10% increase of expenditure in tourism in the province of Zeeland in the Netherlands. Accounting for FATC leads to less usage of labour in the tourism industries as productivity increases allow output to be produced using fewer inputs. This implies lower marginal costs, which leads to lower output prices. These relative input and output price changes stimulate substitution and quasi productivity changes. To what degree the NLIO with FATC leads to more realistic results than the NLIO without FATC depends vitally on the specification of FATC, the differentiation between core and peripheral labour, and the labour supply function. All these elements require additional assumptions and/or data.
For some EIAs the NLIO is an improvement compared to the I-O model because it does not require the assumption ‘no scarcity of production factors’ to be accepted. In the NLIO with FATC neither the assumption of ‘no scarcity of production factors’ nor the assumption of ‘no productivity changes’ is required. In chapter 7 are discussed considerations related to the acceptance or rejection of these two assumptions. Rejection of ‘no scarcity of production factors’ can be appropriate in EIAs in large regions, of large changes of final demand, in regions with limited or no unused labour and capital, in long term analyses, in regions with low factor mobility from and to other regions, and for impact analyses (instead of significance analyses). Acceptance or rejection of the assumption ‘no productivity changes’ depends on the degree to which labour productivity changes can be expected as a result of a final demand change, a consideration which requires expert judgment.
This research makes several contributions to the measurement of the regional economic impacts of tourism:
24 essential criteria that can be used to select a model for application in an economic impact analysis. Although the decision which criteria to consider, and how to weigh these criteria, should always be made on a case specific basis the essential criteria provide a good starting point
This thesis provides additional insights into the differences between the regional I-O coefficients and total output multipliers generated by the four LQ methods. Furthermore, it was shown that a choice for FLQ could be motivated by the wish to avoid or limit overestimation of regional economic impacts.
The NLIO model with endogenous factor augmenting technical change enables a calculation of economic impacts of tourism in contexts where the I-O model is not the most appropriate choice. The NLIO model namely allows for measurement of different ‘shocks’ than final demand changes and can be applied in context where the assumptions ‘no scarcity of production factors’ and/or ‘no productivity change’ are untenable. When applying an NLIO model, the added realism compared to the I-O model needs to be weighed against the need to make additional assumptions, collect additional data, and deal with the more complex nature of this model. In this perspective the NLIO model does compare favourably to the CGE Model, often presented as a more realistic alternative to the I-O model, because it does not depend on data on the relationships between income and final demand (i.e. the need for a SAM).