Mobile Edition
Print
Get the Mag
Weekly UpdatesSophisticated consumer choice modelling methodologies have been proposed, inter alia, by Di Clemente and Hantula (2003) and Guerin (2003). Methods based on stated (as opposed to revealed) choices allow us to identify consumer choice empirically using semi-structured interviews. Within this tradition of housing research known as perceptions/preferences modelling, many other specific methodological issues have gained attention too, such as group decision-making (Molin et al., 1999), neighbourhood aspect (Galster, 2001), how to incorporate supply side shortages (Rietveld and Wagtendonk, 2004), and transitional housing markets (Wang and Li, 2004). Multidimensional decision analysis techniques comprise a sophisticated tool for land use related or environmental problems that require behavioural or prescriptive treatments (see Gregory et al., 1997). While this genre fits well into the housing consumption and preference modelling tradition, it eschews the more philosophical debates within economic theory and methodology that tend to be attached to the use of stated choice methods. A related multicriteria decision making methodology has been applied for the determination of the utility degree and market value of real estate using experts' assessments of preference decisions on competing alternatives (see Kaklauskas et al., 2007).
Multidimensional value and benefit refers to a generic quality measure that goes beyond monetary value or transaction price (Gregory, 2000). According to this modelling tradition the elicited preference models are suited for making monetary values (i.e. market prices) commensurable with non-monetary (i.e. environmental, social, cultural, aesthetic, ecological etc.) values, for various locations or housing bundles (Kauko, 2002, 2003, 2006). In a related strand of inquiry, Daly and colleagues (2003) advocate the 'behavioural paradigm' in residential valuation, which puts more emphasis on the demand or consumer-driven factors related to preferences and intangible quality components.
3.2. The AHP and expert judgments
The analytic hierarchy process (AHP, Saaty 1977) technique is based on pair-wise comparison of elements (attributes or alternatives). In sharp contrast to the classical multi-attribute value-tree modelling approach, which is based on the assumption that utility functions can be explained, the AHP does not assume that the evaluator is able to express the overall elicitation of the problem as a single function. Instead, the AHP is based on the assumption that the relevant dominance of one attribute over another can be measured with a systematic, pair-wise comparison (e.g., Ball and Srinivasan, 1994). This elicitation of weights for the elements under comparison takes place in two stages: initially as ordinal ranks using the original scale of the input, and eventually cardinally as ratios between 0 and 1 derived via a matrix comparison of the ordinal ranks. (See Kauko, 2004, for full explanation).
In other words, the AHP requires numerical ordinal data and measured facts. The comparison begins at the lowest level, where the elements (attributes or alternatives) are usually elicited with an ordinal scale from 1 to 9, with the values corresponding to verbal expressions. A value of 1 means that 'both are of equal importance', and a value of 9 means that A has an extreme importance over B'. The comparisons are then converted into cardinal rankings. Balancing the pair-wise ranks in this way involves the use of measurement theory, as pair-wise judgments cannot be assumed consistent across the entire set of comparisons (e.g., Ball and Srinivasan, 1994). Finally, local weights are transformed into global weights. The most attractive choice is determined by aggregating the local priorities into global priorities. These procedures generate a value tree model, where the overall objective of the decision stands at the top of the hierarchy, with lower-level objectives or attributes at the lower levels (e.g., Zahedi 1986).
Methods such as the AHP have an added value in situations, when more nuanced insight into the context is needed. For this we need different variables than the ones readily available in registers, and as a consequence, a different way of collect the data--through interviews. Given that the method was developed by psychologists, and that it is often used on a set of less than ten attributes or cases, the problem is that is not suited for comparison of several observations simultaneously. Moreover, in order to cover all relevant interest groups, the set of respondents have to be selected meaningfully, not random. Hence the method of saturation is preferred instead of statistical sampling (see Poyhonen, 1998). Using this method each new respondent is expected to contribute with new information, and if that is not the case, including a new respondent does not bring added value that would offset the time costs involved in the interviewing. This implies that the number of respondents does not have to be high--depending on the application even one expert may be sufficient, because, rather than relying on statistical sampling, the use of AHP is always depending on qualitative information relating to the defined context as well as order and scale of variable levels (Kauko, 2002).
The AHP has already been applied successfully in various contexts with different aims and objectives, data size and merits. In a classic demonstration Saaty (1990) showed how to select the best (single-family) house to buy for one (hypothetical) decision maker. In a similar setting, Ball and Srinivasan (1994) offered a rigorous evaluation on the use of AHP, and an application of the role of psychological factors in house selection (again one decision maker) in a suburb of Boston, US (see also Ong and Chew, 1996). Prior studies show that the optimal size of selected respondents for the AHP when used for housing quality and area assessment is about twenty (Nevalainen et al., 1990; but see also Bender et al., 1997, 1999, for the use of questionnaire survey). Most recently, the AHP has been applied for land mass appraisal in Donetsk, Ukraine by Kryvobokov (2005, 2006). In the realm of environmental externality related economic impact analysis AHP has been applied by Erkut and Moran (1991). Using related methodology, impact analysis has been conducted by Willis and Garrod (1993), McLean and Mundy (1998), and Strand and Vagnes (2001).
While the AHP fundamentally is not meant as a statistical survey method, the attribute weights elicited by the AHP can be understood as weights in housing market models, for example, as hedonic coefficients. It is however acknowledged that the experts may underestimate the intensity of the effects of each attribute compared to market participants (i.e. averaging). Kryvobokov (2006) compared the outcome of expert and market based methods by relating each resulting attribute weight or hedonic coefficient to the sum of all weights (or hedonic coefficients) in the model. This exercise showed that the weights resulting from the AHP and also those resulting from a direct questionnaire have systematically lower field ranges than the corresponding hedonic regression coefficients. In other words, the least important attributes were overvalued, and the most important attributes undervalued by the expert based models in relation to hedonic models.
On the other hand, existing market data may be an inappropriate mirror of market reality, in which case interviews become apt. When the equilibrium pertains to choices--not prices--the focus tends to switch to preferences and intangible quality features, and then the measured values can be all kinds of values--not just monetary ones. This way seen the method has two obvious and substantial benefits: (1) it allows for diversification of demand (and then indirectly also supply); (2) it ascertains an intangible elements in relation to perceptions (see Kauko, 2004). The method has certain problems however, such as the inevitable lack of robustness, the inherent property of the AHP restricts the elements to compare to very few, and the inability to perform direct comparison of validity with results obtained with methods based on revealed choice and market outcome data.
Lastly, each of the preference profiles is compared with known cases based on earlier research, but possibly from other circumstances. In this way the findings of the AHP together with prior information can be used for building an operational typology that covers the phenomenon, in this case location-specific housing consumption, as broadly as possibly. This methodology, referred to as naturalistic generalisation, implies that, rather than assuming any strategic reason for generating typical cases, the case repertoire is growing with each new result, with the eventual purpose in operational classification (see Johansson, 2005).
4. SEMI-STRUCTURED INTERVIEWS USING THE AHP
4.1. The study area
The Budapest inner city can be considered a special case, when relating the study to more Western type of housing market circumstances. Here the starting point is to note that in Budapest a Western style suburbia is yet to develop. While there are a lot of settlements around Budapest that grew very intensively after 1990 and a massive suburbanisation process took place, this suburbanisation process is showing signs of slowing down as the public transport system capacity is lagging behind--still ca. 60% of the transport within Budapest is public (Tosics, 2006). After the 1990s period of beginning suburbanisation, today two trends work against urban sprawl as Locsmandi (2006) notes: (1) the private housing boom within the city; (2) the lack of public investments in transportation facilities. On the other hand, in the inner city urban renewal and housing rehabilitation has been on the agenda constantly for the last twenty years. This is because the inner city area comprises a big problem for the whole image of the city--an increasingly important urban policy aspect. The housing market balance between supply and demand is not well developed after the socialist era as the existing stock is to a great extent derelict, especially with respect to the maintenance of communal areas and the facades, and lagging behind the preferences of those consumers with purchase power. As already noted, on a detailed level the situation is indeed very different than in the Western urban context. On a more general level, it is possible to nevertheless make some cautious links to certain contested urban and neighbourhood regeneration issues brought up in Western policy evaluation studies (see e.g. Cameron, 2006), in so far as the discussion concerns attracting the targeted population segment necessary for housing market renewal without neglecting the rights of the current residents. At the time of interviews (summer 2005) the market for private dwelling construction is very marginal at the citywide level; out of 800,000 units the output of the last ten years is ca 5%, which is the share of those households who are 'in the market'. (This figure has apparently increased somewhat since then, due to the activity of first time buyers and younger up-graders.) Within this group, the target for the analysis reported in this paper is the segment that comprises households who move to the outer part of the Pest inner city, that is to say neighbourhoods in the following Pest-side districts (around the circular Grand Boulevard, clockwise from north to south, see the right hand side of the map in Figure 1):
FEED