Keywords

1 Introduction

In Italy, property taxation is based on property cadastral values, which are computed using cadastral rents defined by the Italian land registry, dating back to the 1950s and 1960s. The land registry defined the cadastral rents still used today, across large geographical areas. These cadastral rents were defined empirically using a property classification that was based on physical features of properties, which no longer reflect the real estate market. Cadastral rents rely on the number of cadastral rooms. Cadastral room sizes belong to a predefined range rather than having a fixed number of square meters. The first cause of inequity of taxation is that they represent location using a geographical segmentation into large census zones, which no longer reflect real estate submarkets. For this reason, property location has a negligible impact on cadastral rents, although the impact of location on prices is well known (Bourassa et al. 2003, 2007; Goodman and Thibodeau 1998). The second cause of inequity of taxation is the use of cadastral rooms rather than actual sizes in square meters. Because of this, properties with different numbers of square meters can have the same number of cadastral rooms. Consequently, the householder of the smaller property pays the same taxes as the householder of the bigger property.

The aim of this paper is twofold: on one hand, we empirically measure the discrepancy between current cadastral values and market values, which is a measure of social inequity. Using a traditional hedonic approach (Rosen 1974), we identify the main factors that explain the spread between market and cadastral values, based on a case study. On the other hand, we propose a methodology to update current rents to match them more closely to market prices, while awaiting the next phases of the land registry reform, which will be a long process. The case study is Turin’s real estate market.

The paper is organized as follows. Section 2 introduces the empirical methodology used. Section 3 presents the case study and the databases used for the analysis. We discuss the empirical results in Sect. 4, and conclusions are presented in Sect. 5.

2 Empirical Analysis Design

The land registry defined cadastral values (CV) using the income approach (Regio Decreto Legge n. 652 of 1939 and Law no. 1142 of 1949). The benefits of property ownership were measured to define the cadastral rent (CR) of assets, which were divided into different categories (from category A to category F) according to their use. The biggest category was A, which includes ordinary assets and residences. Accordingly, cadastral rents were defined for assets in category A, while, for assets in other categories, the appraisal was performed case by case. In this study, we consider residences, which are assets belonging to category A. For assets in category A, cadastral rents (CR) are defined across census zones. Assets in the same category and census zone are divided into classes, according to their physical features. Then, for each class the cadastral rent by cadastral room, which we call unit cadastral rent (UCR), is defined. Finally, a property CR is the product of UCR and the number of the property cadastral rooms. Cadastral values (CV) are obtained using some coefficients defined by law, which are not connected with the interest rate. Formally, CV can be deduced by CR by the following:

$${\text{CV}} = {\text{CR}}* 1 2 6$$
(1)

Notice that CV is a linear transformation of CR. Indeed, we work directly on CR whenever this is possible. As discussed above, the low contribution of location to CR and the use of cadastral rooms as a proxy for size are the main factors that contribute to increasing the difference between cadastral values and market prices. Accordingly, we focus on these two aspects. With respect to location, there are some studies (Goodman and Thibodeau 1998; Bourassa et al. 2007) that conclude that geographical housing submarkets are more important in predicting house prices than the spatial statistics approach (Bourassa et al. 2003, 2010). Coherently and in accordance with the presidential decree 138/1998, we use Microzones to model location. Their importance for Turin real estate is also empirically supported in Fregonara and Semeraro (2013). Since, in Italy, information on market prices is not public and is difficult to retrieve, consequently we use listing pricing (LP) to proxy market prices (Taltavull and McGreal 2009; Curto et al. 2015), where listing prices are used to analyze the real estate market. Listing prices can be used to proxy market prices for the analysis of the two aspects considered in this paper: location and size. In fact, the explanatory power of location on listing price could be considered as a proxy for the explanatory power of location on market price, as empirically shown in Fregonara and Semeraro (2013).

The analysis performed in this work is developed in two steps. The first step aims to show that location and cadastral rooms are the main factors influencing the discrepancies between cadastral values and market values. The second step proposes an empirical procedure to update cadastral values to reduce the discrepancies with market prices.

2.1 Factors Influencing Spread in Cadastral Rents and Asset Prices

The first step of analysis is performed using a traditional hedonic approach. We specify hedonic models to explain the two dependent variables: listing price (LP) and cadastral rent (CR). We focus on the explanatory power of Microzones, number of rooms and size measured in square meters. The variable Microzone is a nominal variable with disordered modalities: the attractiveness of each Microzone depends on the subjective perceptions of sellers and buyers. Therefore, Microzones are specified by dummy variables. We empirically computed the coefficient of determination corresponding to linear regressions to explain listing prices, cadastral rents, and the differences between listing prices and cadastral values (DP):

$${\text{DP}} = {\text{LP}} - {\text{CV}}.$$
(2)

We considered two hedonic models, corresponding to two sets of explanatory variables. Formally, the first model is:

$$Y_{j} = \alpha_{j0} + \sum\limits_{i = 1}^{n} {\alpha_{ij} X_{ij}^{CR} + \varepsilon_{j} ,\quad j = 1, \ldots ,k = 1,2}$$
(2.1)

where Y j, j = 1, …, 4 are LP, CR, and DP, respectively, and the explanatory variables \({X^{CR}}_{ij}\) are the variables defining cadastral rents: category, class, and number of cadastral rooms. The hedonic weight α ijk , j = 1, … n i , i = 1, …, n assigned to each variable is equivalent to the contribution level of these characteristics to the price values (Rosen 1974); α j0 is the model intercept; and ε j the error term. The second model is:

$$Y_{j} = \alpha_{j0} + \sum\limits_{i = 1}^{n} {\alpha_{ij} X_{ij}^{LP} + \varepsilon_{j} ,\quad j = 1, \ldots ,k = 1,2}$$
(2.2)

where Y j j = 1, …, 4 are LP, CR, and DP, respectively, and the explanatory variables \({X^{LP}}_{ij}\) are the main factor influencing market prices: Microzone, quality of the building, and size (Curto et al. 2015). The hedonic weight α ijk , j = 1, … n i , i = 1, …, n assigned to each variable is equivalent to the contribution level of these characteristics to the price values (Rosen 1974); α j0 is the model intercept; and ε j the error term. The hedonic weights in Eqs. (2.1) and (2.2) are estimated using traditional least-squares estimates. The coefficient of determinations (R2) of the two regression models measures the proportion of variation of the dependent variables (LP, CR, DP) explained by the model.

2.2 Operational Proposal

The second step of analysis proposes a methodology to adjust CR across Microzones to introduce cadastral rents per square meter. We introduce a location-adjustment coefficient for each Microzone to apply to CR—and CV—as follows. If P i is the sample mean of assets prices in the Microzone, i = 1, …, N, where N is the number Microzones into which the area under appraisal is divided, the coefficients c i are defined by:

$$c_{i} = \frac{{P_{i} }}{{\frac{{\sum\nolimits_{i = 0}^{40} {P_{i} } }}{N}}} = \frac{{N P_{i} }}{{\sum\nolimits_{i = 0}^{40} {P_{i} } }}$$
(2.3)

Hence, each coefficient is the ratio between the mean listing price observed in the Microzone i, i = 1, …, 40 (P i ), and the arithmetical mean of P i , i = 1, …, N, i.e., \(\overline{P} = \frac{{\sum\nolimits_{i = 0}^{40} {P_{i} } }}{N}\). Then, the adjusted rent ACR j (I) of a property I located in Microzone j, j = 1, …, N is given by:

$$ACR_{j} (I) = c_{j} CR_{j} (I)$$
(2.5)

where CR j (I) is the cadastral rent of property I. The relationship between the adjusted rent ACR i (I i ) and ACR j (I j ) of two properties I i and I j , belonging respectively to submarkets i and j, depends only on the mean prices in the two submarkets (Curto et al. 2014).

The second proposal to update CR consists of defining a rent per square meter (square meter cadastral rent, SMCR for short), according to actual rents per cadastral room (UCR). In that, we assume the value as unit rent of asset I to be:

$$SMCR(I) = \frac{UCR(I)}{{s_{middle} (I)}}$$
(2.7)

where UCR(I) is the rent per cadastral room of asset I and s middle (I) is the arithmetical mean between the minimum size and the maximum size per cadastral room of asset I (the range of the number of square meters per cadastral rooms are provided in Table 4). Then we can find recalculated cadastral rents (RCR) of asset I by multiplying SMCR(I) and property size, and we obtain:

$$RCR(I) = SMCR(I)*s(I)$$
(2.8)

where s(I) is the size of asset I measured in square meters. Recalculated cadastral rents (RCR) are compared with current rents. Finally, we apply the location coefficients c j , j = 1, …, n also to RCR, to analyze the distribution of rents across Microzones after the introduction of SMCR and the application of location adjustment coefficients.

3 The Case Study

We perform the empirical analysis on a case study of the city of Turin. Turin is divided into four census zones and forty Microzones. Microzones were defined by Politecnico di Torino in 1999, using a territorial information system and performing a factorial analysis and a cluster analysis. The factors considered include price level, building characteristics, accessibility, presence of services, and green areas, as provided by the law 138/1998. Microzones are numbered from 1 to 40, fanning out from the center of the city to the suburbs. Figure 1 compares Microzones with the four census zones in which Turin is divided.

Fig. 1
figure 1

The four census zones and the 40 census Microzones of the city of Turin

Full size image

As one can see in Fig. 1, the census zones are big areas that do not reflect local amenities. In the same census zone, there are heterogeneous assets. In particular, in census zone 1, one can find prestigious assets as well as low-quality assets. Coherently, people living there belong to various social status. Instead, Microzones are a geographical segmentation, and they are able to explain almost 40% of market prices, as empirically measured in Fregonara and Semeraro (2013), and of listing prices, as shown in Curto et al. (2015).

3.1 Data

The analyses use two separate property databases of the Turin Real Estate Market Observatory (TREMO). TREMO was founded under an agreement between Politecnico di Torino, Turin’s Municipality, and Turin’s Chamber of Commerce, with the institutional aim to collect and analyze data from the real estate market. We consider the TREMO sample of assets on sale in 2013, which consists of 566 properties, and we call it the TREMO sample. For each asset, we consider the characteristics: Microzone, size (measured in square meters), and building quality. We defined five building quality levels, which consider several building characteristics, such as building materials, age of the building, and also cadastral category and class. The highest level corresponds to attractive properties while the lowest corresponds to council houses. We used dummy variables to model the quality-of-building levels. Descriptive statistics for each Microzone are provided in Curto et al. (2014). The second database is the property of Turin Municipality, and we call it CDB (cadastral database). The CDB contains information on each asset in the Land Registry. Information collected for residential assets are cadastral rent, category, class, number of cadastral rooms, and location, which is provided through map sheets, the number, and the subdivision. Unfortunately, the identification code used by TREMO does not allow for a one–to-one association between the TREMO and CDB databases, since address codes identifies buildings and not apartments. Nevertheless, the CDB sample consists of 49,305 data and includes the assets with the address codes sampled in TREMO sample. Descriptive statistics for the CDB sample are provided in Curto et al. (2014).

To analyze discrepancies between CR and LP, we link the two databases using address code and floor. We found one-to-one correspondence for 129 data, which we collect in a new sample named BDM. The sample of 129 properties provides information from the two databases TREMO and BDC. Despite the small sample size, the statistics in Table 1 highlight the need to update current rents and include the marginal contribution of location on property value. Table 3 includes Microzones with at least three observations; Microzones with at least seven observations are in bold.

Table 1 Sample JOIN: listing prices and cadastral values—under the assumption of residence—mean values for Microzones
Full size table

Table 2 provides the mean and the range of size across properties with different numbers of cadastral rooms and exhibits a large variability of size in correspondence to the same number of cadastral rooms. From this, cadastral room number seems not suited to be a proxy for size. In this sense, descriptive statistics seem to justify our approach.

Table 2 Sample JOIN: size for given number of rooms statistics
Full size table

4 Empirical Results

This section presents the results of the empirical analysis performed. Firstly, we analyze the main factors explaining CR, LP, and the difference between LP and CV, i.e., DP. Secondly, we update the current cadastral rents and discuss the results.

4.1 Factors Influencing Cadastral Rents and Market Prices

This section analyzes the main factors explaining LP and CR. Results of hedonic analysis are provided in Table 3a, b. Table 6a explains CR, LP, and DP using factor defining cadastral rents, i.e., category, class, and cadastral rooms. It shows that the factors defining rents are not able to fully explain LP, while they contribute 37% to explain the difference between CV and LP.

Table 3 Hedonic regression analysis
Full size table

Table 3b explains CR, LP, and DP using the main factor influencing LP, i.e., size, Microzone, and building quality, which explain 87% of LP. These factors are able to explain only 37% of CR, and they are responsible for the difference between CR and LP (\({{\text{R}}^{ 2}}_{\text{adj}}\) = 0.85). These results highlight the fact that location and size are able to explain the current spread between CR and LP. We also split the above regression into two separate regressions to measure the impact of location and building quality on LP and CR separately. The first regression has Microzones as an explanatory variable, and we found \({{\text{R}}^{ 2}}_{\text{adj}}\) = 0.49 for LP and \({{\text{R}}^{ 2}}_{\text{adj}}\) = 0.22 for cadastral rents. The second regression has building quality as explanatory variable and we found \({{\text{R}}^{ 2}}_{\text{adj}}\) = 0.45 for LP and \({{\text{R}}^{ 2}}_{\text{adj}}\) = 0.23 for cadastral rents. These results underline that factors which influence current market prices do not influence cadastral rents.

The next section proposes a simple methodology that could be applied by local administrations to update CRs; it relies on a redistribution of CRs across Microzones and on the introduction of unit CR per square meter to overcome the drawbacks of cadastral rooms as a proxy for size. We notice that also building-quality classifications should be updated to redefine cadastral categories and classes; nevertheless this action requires a long process, which is part of the Land Registry reform and is beyond the scope of the present paper.

4.2 Rents Adjustment

Following the procedure outlined in Sect. 2.2, we compute the location-adjustment coefficient c i , i = 1, …, 40 in Eq. (2.3), where P i , i = 1, …, 40 are the mean LP for the 40 Microzones. The location-adjustment coefficients are provided in Table 4. Then, for each asset in the CDB sample, we obtain the adjusted cadastral rents (AR) by applying Eq. (2.5).

Table 4 Adjusted rents by Microzone
Full size table

In Table 4, we provide the mean values of CR and of AR for each Microzone.

Notice that, in the Microzones with the location-adjustment coefficients greater than one, the adjusted rents would increase as opposed to those with the location-adjustment coefficients smaller than one. Following the procedure in Sect. 2.2, we computed SMCR for each asset in the BDM sample using Eq. (2.7), and then we computed the recalculated cadastral rents RCR for each asset in the sample using Eq. (2.8). Table 5 provides statistics of RCR across Microzones. Recalculations yield result higher than current rents, suggesting that everybody declares the minimum admissible number of cadastral rooms per property.

Table 5 CR recalculated using UCR per m2
Full size table

Finally, we apply the location-adjustment coefficient to RCR of each Microzone to re-distribute RCR across them. Table 6 exhibits, for each Microzone, sample statistics for final cadastral rents (FCR), obtained applying location coefficients to RCR.

Table 6 Final rents: descriptive statistics
Full size table

Notice that the application of the location-adjustment coefficient to RCR leads to FCR, which are higher than current rents in the most desirable Microzones and lower that current rents in the less attractive Microzones.

5 Conclusions

This paper analyzes discrepancies between current cadastral values, which are defined by Land Registry and are used as a base for property taxes, and market prices, which are represented by listing prices. We focus on two main aspects: location and size. These two aspects are proven to be the main factors influencing the discrepancy between cadastral values and listing prices. In fact, using hedonic approach, we show that location has a negligible impact on current cadastral rents and explains the differences between cadastral rents and listing prices. We then propose a simple methodology to update rents, which makes it possible to incorporate the value of location and to overcome the current use of cadastral rooms as a proxy for size. The methodology proposed could be applied by local administrations while waiting for Land Registry reform, in accordance with the recent regulation provided by Legislative Decree no. 23 of 11 March 2014. We show that updated rents could be better as proxy listing prices and that they can reproduce the variation of prices across spatial real estate submarkets.