The Estimation of Property Tax Capitalisation in the Korean Taxation Context

The difference between public services and property tax rates is capitalised into home values. The aim of this research is to estimate the property tax capitalisation rate under a different taxation context of Korea, using a repeat sales method with short-term data on housing prices and estimated tax payments. In the operation of the property taxation, there is complexity that needs to be considered in the estimation of the property tax capitalisation rate. In this research, 32,101 apartment samples in Seoul are used for the analysis. Given these unique institutional circumstances, as a result of the analysis, the property tax capitalisation rate in Seoul was between 73.7% and 82.8% in the analysis periods.


Introduction
The difference between public services and property tax rates is capitalised in home values.This phenomenon is called a capitalisation effect (Yinger et al., 1988;Rosen & Gayer, 2007).Higher property tax payments lead to lower house values if all else being equal.Theoretically, housing prices drop by the discounted present value of property tax payments when the property taxes are imposed.This is called full or complete capitalisation.However, in the extensive study of Yinger et al. (1988), property taxes were partially capitalised.In spite of a simple concept of property tax capitalisation, it is challenging to estimate the degree of capitalisation (Yinger et al., 1988).
Different taxation policy contexts require different analysis methods.During the early 2000s, housing prices increased drastically in a selected few areas in Korea.According to a housing price index, housing prices in Seoul increased approximately 50% over the three years in the early 2000s.In response to the changes in the housing market, the Korean government utilised property taxation as a tool to stabilise housing prices and to improve housing affordability.The government announced more than 30 policies on property markets during the period 2003-7 (Jang, 2010: pp. 274-355).Nevertheless, little has been studied about the influence of the reinforced property taxes on housing prices in Korea.
Thus, the aim of this research is to estimate the property tax capitalisation rate under a different taxation context of Korea, using a repeat sales method with short-term data on housing prices and estimated tax payments.This study sheds light on the presence of the property tax capitalisation effect and proposes a new method to estimate that effect.

Institutional Context
The Korean government has three hierarchies: the central government, metropolitan governments (si/do, e.g.Seoul, Busan or Gyeonggi), and local governments (si/gun/gu).Each government levies property taxes, respectively.In order to calculate property tax rates precisely, this section reviews the structure of property taxation in Korea currently being operated in the three levels.
Firstly, there are many types of property taxes as a form of a surtax imposed by different levels of governments.These taxes include "Property Tax" 1 and "Gross Real Estate Tax (hereafter GRET)" and there are other taxes, based on the same property, such as "City Planning Tax" 2 , "Local Education Tax", "Facilities Tax", and "Special Tax for Rural Development".These taxes should be regarded as one of the property taxes because they are all levied to the same property using the same paying method.For example, when tax payers pay for the "Property Tax" to the local governments, they also have to pay for the "Facilities Tax" to the metropolitan government with the same tax bill.The "Local Education Tax" is additionally imposed to the owner with 20% of property tax payments.These taxes are included in one tax bill so that the governments may collect taxes easily.The collected money is distributed to each government.Likewise, the "Special Tax for Rural Development" is the surtax of 20% of the GRET.The GRET is imposed to expensive housing by the central government in order to stabilise housing prices, to recoup capital gains and to contribute to ba-lanced development between regions.Accordingly, it is easy to underestimate effective property tax rates when tax payers only consider the legal tax rate described in the Local Tax Act.To be precise, all these taxes, thus, need to be regarded as the property taxes.
Secondly, the Korean property taxation is progressive by values.Higher tax rates are applied to expensive houses, while lower tax rates, inexpensive ones.Progressiveness was strengthened due to the enforcement of the "Gross Real Estate Tax Act" in 2006.For houses below 40 million WON 3 , their "Property Tax" rate is only 0.15%.However, for houses more expensive than 600 million WON, their "Property Tax" rate becomes 1%.If the price is over 2 billion WON, the owner has to pay 2% of assessed housing prices.Table 1 summarises the details of tax rates by housing price bands.
Thirdly, local governments have a limited influence in the decision of property tax rates.For the purpose of equity between regions, the "Local Tax Act" bans local governments from changing tax rates.The elastic taxation meant that local governments had the authority to adopt a different tax rate within 50% of flexibility; the elastic tax rates were only operated in 2005 and 2006.During these periods, a few local governments adopted lower property tax rates than standard tax rates presented in the "Local Tax Act".However, equity was considered significant and the implementation of elastic taxation ended in 2006.Basically, the central and metropolitan government support the deficit after local governments finance local taxes at given property tax rates.So, there is a discrepancy between burdens of tax payers and benefits from the local governments.Even though owners of expensive properties pay for a large amount of property taxes on a higher tax rate, they benefit from public services similar to those who pay less due to an inflexible tax rate system discussed above.
Besides, additional complexity exists in the taxation.A tax payment-cap policy is in operation.The purpose of this policy is to prevent a rapid increase in tax payments caused from sudden changes in the real estate market.For example, a property tax payment should not exceed over 105% of the last payment when the price of the property is within 300 million WON, 110% for its price is laid between 300 and 600 million WON, and 150%, over 600 million WON.In addition, the aggregate amount of the "Property Tax" and the GRET should not exceed three times of the property tax payments in relation to the corresponding period last year.
This complexity under Korean taxation contexts creates difficulties in estimating the property capitalisation rate.Thus, the principle of the repeat sales model is used to estimate the property taxation capitalisation.

Data
The Korean central government assesses housing prices of all housing in Korea for tax purpose on a yearly basis viaa government agency-Korea Appraisal Board (KAB).The appraised prices make it possible to calculate the amount of individual property tax payments.
The unit of the analysis in this research is an individual apartment (high-rise condominium).Extensive data for three years about housing (appraised) prices between 2006 and 2008 was used for this research.A sample consists of 32,101 apartments, which account for 2.6% of total apartment stocks in Seoul (see Table 2).Apartment is a major housing type in Seoul, accounting for 54.2% of the housing stock (Census in 2005).Almost new housing construction is composed of apartments, for example, 76.5% in 2006, and 100% of housing renewal projects were a form of high-rise apartments (Ha, 2010).Thus, the sample of this research is large enough to represent Seoul's housing market.
The To calculate the amount of tax payments and effective tax rates, following four factors were considered in the Korean context: appraised apartment prices, tax-rate bands, the elastic tax rates in 2006, and the property tax payment cap4 .These factors made the amount of property tax payments change every year.The effective tax rate is house's property tax payment divided by its market value (Yinger et al., 1988).The summary of calculated tax payment and their effective tax rates are presented in Table 3.
Based on aggregated tax payments and appraised prices, effective property tax rates were 2.059 mill in 2006, 1.990 mill in 2007, and 2.139 mill in 2008, respectively.

Methodology
This study takes advantage of the principle of a repeat sales model.The repeat sales model was developed by Bailey et al. (1963).Then, Case & Shiller (1987) used this model to estimate variations in housing prices and to create an housing price in-  (McMillen, 2008).The principle of the repeat sales model is that, when one house is transacted, the price matches to the previous transaction price of the same house so that the difference between the two prices can be measured precisely.The changes in housing prices are estimated by a regression analysis.The rationale of this matching is to remove any expected biases.
Every single house has its unique locational and structural characteristics.Accordingly, only when the house is compared to the same house without any structural alterations, can real changes of housing prices be correctly measured.The repeat sales model adds time-variables to the hedonic price model used for a cross-sectional analysis.As the same house is compared in calculation of housing price changes, the repeat sales model rules out the effects of other variables, such as lot size, bedrooms, toilets, and garages, as long as the house was not renovated between the two transaction periods.
In the model of the initial repeat sales model by Bailey et al. (1963), a natural log was taken to calculate the changing rates of housing prices.The natural log makes left-sides directly a variation rate in housing prices.P f is the first transaction and P s is the second one in Equations ( 1) and ( 2) (Choi et al., 2010).In In , , 1 In In where X i is characteristic of the house.
i  is coefficient.
s  and f  are error terms.
As Equation (1) subtracts Equation (2), the equation is simplified to conduct a regression analysis.In Equation (3), the explanatory variables in a general hedonic price model are deleted.
1 In v The left side is the rate of variations in housing prices; all individual housing characteristics are removed in the right side.This is the most advantageous feature in the repeat sales model simplifying model specification.D t in Equation ( 3) is a dummy variable.It becomes one when transaction occurs, otherwise zero.β t is the coefficient to be estimated by conducting a regression analysis.β t can be transformed to a housing price index 5 .
McMillen (2003) and Kim & Lee (2004, 2005) modify the repeat sales model.They divide variables into time-varying variables and time-fixed variables to discover how influence of the time-varying variables has changed, by estimating the coefficient of the time-varying variables, where time-fixed variables are eliminated.Likewise, this paper utilises the principle of the repeat sales model to estimate the property tax capitalisation effect under the Korean taxation context 6 .When there are no property taxes, the hedonic price model can be specified like Equation (4).
where, is the housing price before imposing the property tax.P X is attributes of the house.0  is an intercept.k  is a coefficient.When the effective property tax rate is  , the amount of property taxes is P   P .


 should be divided by the discount rate (r) in order to acquire the total present value of property taxes.When the property tax is in operation, the hedonic price model can be expressed like Equation (5) 7 .
In Equation ( 5), 1 t T r  is a net present value of property tax payments and β is a coefficient that indicates the capitalisation rate.If β equals one, property taxes are fully capitalised.Housing prices decrease by the present value of increased property taxes.On the other hand, ifequals zero, housing prices do not change regardless of newly added property taxes.In Equation ( 5), the period of the left side-the housing price (P)-is t while, on one of the right side-the amount of property tax (T) -is t − 1. P t is closely tied to 1 t P  as effective tax rates and the amount of total property taxes are decided by housing prices, causing an auto-correlation problem that creates difficulty in estimating by OLS.Due to the Korean taxation context discussed in the Section 2, one more step is required.In order to remove the auto-correlation problem, the repeat sales method is modified as suggested by McMillen (2003) and used by Kim & Lee (2005).In the period t + 1, housing prices including property taxes become Equation (6).
If Equation ( 6) is subtracted from Equation ( 5), the attributes of the house are eliminated.It is possible when there have been no substantial physical changes to housing attributes between t and t + 1.Finally, Equation ( 7) can be reached to a simplified form to conduct a regression analysis.
  , and this intercept signifies changes of hous-ing prices due to the socio-economic conditions such as income levels, the demand for or the supply of housing, public services and employment.These factors are not spatially specific, thus can be equally applied to all areas in Seoul. 1  is the coeffi- cient that represents the change in housing prices caused by the changes of property taxes. 1 In addition to the changes in market trends such as the demand and the supply, local influences caused by location-specific changes in amenity, infrastructure, and accessibility, can differ between districts over the analysis period.For example, the construction of an outer circular highway around Seoul was completed in 2006.Due to the new highway, accessibility was improved in the northern part of Seoul, which influenced housing prices.Housing supply varies between districts.For example, while there was no new apartment construction in Nowon, 27,559 new apartments were built in Songpa in 2007-2008.Conditions of supply and demand in each region create a discrepancy in housing price movements.It is necessary to consider these effects are different in each district.Thus, district dummy variables are added to Equation ( 4) to measure the location-specific influence over time.For example, for Seochogu, only the Seochogu's district dummy variable equals one, otherwise zero.District dummy variables explain time-varying effects between districts.

Result of Analysis
As a result of a regression analysis8 , almost all variables are statistically significant at the 95% confidence level (see Table 4).The intercept-17,537-plays an important role in interpreting the trend in the housing market.With taxation effects and location-specific differences controlled, the intercept explains that housing prices increased by 17,537 thousand WON on average between 2007 and 2008.When referring to the average housing price in 2007, the average change in housing prices over the one year was a 6.3% increase (=17.537/279.04).This is attributable to the provision of public goods, macroeconomic factors, and the demand for and the supply of housing.
Only four district dummy variables are not significant at the 5% significance level, but two of them are statistically significant at the 90% confidence level (see Table 5) 9 .Two districts, Eunpyeong and Gwanak, show that the dummy variable is not significant at the 10% of the significance level.This is due to variations in housing prices from house to house in Eunpyeong and Gwanak are so large that the statistics cannot conclude that the change in housing prices is different from the base district, Gangnam.For example, there have been redevelopment projects, called New Town Development, in Eunpyeong, since 2002.Thus the infrastructure provided concentrated on specific areas.This caused larger variations in housing prices in Eunpyeong.Housing prices in the new residential development in Eunpyeong increased much more than outside of the develop- 7 Theoretically, capitalisation takes place at the time of the announcement of the tax change.However, one year time lag is necessary given the policy context.Property taxation is complicated enough to recognise because they consist of different kinds of taxes in Korea.In addition, the amount of property taxes is billed to home owners in June and September whereas appraising is done in January and noticing, in April.Thus, this study assumes that property taxes in year t have an influence on housing prices in year t + 1, taking a one-year time-lag in Equation ( 5).The assumption about the one-year time-lag is reasonable as far as processes of appraisal, notice, and billing are concerned.& Wheaton, 1996: p. 207).However, deciding on a discount rate has been problematic in estimating the property tax capitalisation rate (Yinger et al., 1988).Some studies adopt a nominal discount rate instead of a real discount rate and some use higher rates than the real rates, thus making the capitalisation rate higher.A real discount rate needs to be used and income tax on interest and tax deduction should be considered when measuring the real discount rate (Yinger et al., 1988).According to the Bank of Korea, the long-term interest rate was 3.98% in 2006 and 4.47% in 2007.The income tax rate on interest income was 15.4%.Considering the income tax rate on the interest, the real discount rate becomes 3.37% and 3.78%, respectively.Accordingly, the capitalisation rate of the property taxes becomes 73.7% and 82.8%.Thus, it is reasonable to conclude that the degree of property tax capitalisation was somewhere between 73.7% and 82.8% (see Table 6).As studied by Yinger et al. (1988), the result shows partial capitalisation that means the effect could transfer to the future owner (Palmon & Smith, 1998) or to tenants thus resulting in an increase in rents.

Conclusion
This paper focuses on the property tax capitalisation effect under a unique Korean taxation context.The model employed in this research is simple but based on reasonable assumption in which the time-fixed variables can be eliminated.While a conventional hedonic price model includes locational and structural variables, the modified repeat sales model utilises variations in property tax payments, which simplifies model specification.The model in this research used three years data on housing prices and tax payments.
Given these unique institutional circumstances, as a result of the analysis, the property tax capitalisation rate in Seoul was between 73.7% and 82.8% in the period 2007-8.This is partial In terms of policy objectives, the increase in property tax rates is conducive to changes in housing prices and increases in public funds.However, there is a fundamental issue about the rationale for property taxes.The purpose of the property tax is to ensure public goods which enhance property values, not to depreciate home values as an intervention tool in the housing market.Bruekner (1979) suggested that an efficiency is in the point that a property tax locally financed supports house values sufficiently in that region.Although Korean property taxes finance public expenditure, it is difficult to regard that the Korean property taxation is efficient due to the lack of a close connection between property taxes and public expenditure.Equity concerns have been more emphasised than taxation efficiency in Korea.
This study contributes to evidence of property tax capitalisation in a different policy context using the principle of the repeat sales model.As a result of the analysis, it is obvious that additional property taxes are reflected in home values.Property tax capitalisation can be understood as a natural result in response to policy changes.Open Access

f
 and s are the intercept composed of macro-economic factors, housing preferences and other factors that can influence housing prices., , 1 this research uses appraisal prices, the principle of the analysis relies on the repeat sales model that use transaction data.H.M. KIM ET AL.

Table 1 .
Legal tax rates and tax base in Korea.The Local Tax Act and Gross Real Estate Tax Act.
average housing price was 222,375 thousand WON in 2006, 279,042 thousand WON in 2007 (a 25.5% increase in relation to 2006), and 290,946 thousand WON in 2008 (a 4.3% increase in relation to 2007).More details are in Appendix.

Table 3 .
Estimated effective tax rates and tax payments by districts.using transaction prices.Since Bailey et al. and Case & Shiller, the repeat sales model has been widely used in generating a housing price index and understanding property markets dex

Table 4 .
Descriptive statistics of dependent and independent variables.

Table 5 .
Result of a regression analysis.P t+1 − P t ; ** adj-R square: 0.4831; *** Notes: The unit is thousand WON, and Gangnam is the basement of district-dummy-variables.ment area 10 .Yongsan has the highest dummy-variable coefficient.Yongsan is in the middle of the three central business districts in Seoul, but the quality of houses is in general not as good as other residential areas.Recently, there have been redevelopment projects all over the Yongsan area.The dummy variable reflects the effect of the projects on housing price changes in Yongsan.Yangcheon and Songpa show the biggest decrease in housing prices.These two districts including Gangnam belonged to the most expensive areas in Seoul.by the discount rate.The discount rate is an opportunity cost that the home-owners can accrue when they invest in other options instead of housing.Simply, the longterm interest rate could be used as the discount rate (DiPasquale * Dependent variable: 

Table 6 .
Interest rates and capitalisation rate.
For appraisal housing price and sample, the sample of this study; Source: Ministry of Land, Transport and Marine Affairs.