The effect of change in the property tax: evidence from the Swedish taxation reform*

The effect of change in the property tax: evidence from the Swedish taxation reform*

Timotheos Mavropoulos

¹

This version: October 16, 2014 Preliminary draft

Abstract

The paper investigates the effect of the property taxation reform introduced in Sweden after April 2007 on the market price of the affected properties (one-family houses). Inference is focused on discount rates, growth rates and the implied degree of tax capitalization. Stronger results for the capital compared to the rest of the country underline the importance of land scarcity for the degree of tax capitalization.

*I would like to thank Peter Englund for his advice during the completion of this term paper and encouragement whenever I found obstacles in proceeding. I would also like to thank Valueguard AB and in particular Lars-Erik Ericson for his support in data extraction process. Gratitude goes to participants of the NFN PhD conference 2014 and Kristian Miltersen (Copenhagen Business School).

1 Stockholm School of Economics (email: timotheos.mavropoulos@hhs.se)

1 Introduction

The degree of capitalization of property taxes and its effect on the market price of properties is an important subject to investigate. Property tax reductions shift the tax burden from current homeowners to the future ones. In the current paper I try to investigate if the tax effects are rationally accounted for by the market. In particular, it is interesting to know to which degree the Swedish property market for one-family houses capitalizes the property taxes, hence revealing the information about the expectations on the future taxation treatment of such properties. I will analyze a tax reform announced in April 2007¹. Until 2007 the tax was proportional to house value. From April 2007 a cap on the tax in SEK was introduced. This meant a sharp reduction of the tax for most properties in Stockholm and other major markets, but no reduction in other parts of the country. Though the magnitude of the tax effect is not obvious, it is clear that it should differ across different types of properties depending on their assessed value (e.g. major cities versus provincial areas).

The research literature on the capitalization of property taxes is rather scarce and mainly concentrated on the US property tax system (Bradley 2011, Bradbury et al. 2001). Most of the literature is cross-sectional in nature and asks whether differences across jurisdictions in taxes and local amenities get capitalized in house prices. Few studies deal with the dynamic issue of price effects due to a tax reform. A recent example is a study by Bradley 2011², where the effect of tax obligations of new homebuyers is explored in connection to the sale prices using data on homes sold in Michigan for the 1997-2007 period.

Introducing a cap on the taxable values of the real estate properties results in a substantial variation in tax obligations for existing homeowners vis-à-vis new entrants to the housing market. The study of taxation cap can reveal the degree of real estate market efficiency. In particular, if the existing homeowners with lower tax obligations due to the introduced cap are overcompensated as if such a cap is maintained permanently, the market is considered highly efficient. This is important information from the policy decision point of

1 A more detailed description of the reform history is given in the later section.

2Bradley Sebastien, Capitalizing on Capped Taxable Values: How Michigan Homebuyers are Paying for Assessment Limits, University of Michigan, 2011.

2

view. Elinder and Persson (2014), while exploring price responses and capitalization degrees from the Swedish real estate reform present results largely inconsistent with capitalization theory, emphasizing the role of behavioral theories of bounded rationality as a key explanation. I will attempt to contribute to the literature by exploring the effect of the taxation cap introduction in Sweden in 2007 while emphasizing the rational explanations, taking into account the implied discount rates and degrees of tax capitalization.

I proceed with reform history in the context of property taxation environment in Sweden, refer to capitalization theory, data, sample selection and methodology, present results and conclusion.

Property taxation in Sweden

To carry out property taxation the Tax Agency in Sweden completes real estate assessment according to a rolling schedule triennially, with different types of assessment units like one- or two-dwelling buildings, apartment buildings, units for agriculture and forestry, etc. The assessed value is set at 75% of the market value. From 2007, as will be described in a later section, a reform was introduced with government property tax on dwellings having been abolished and replaced by a local real estate fee. Since then the charge for single-family houses should not exceed 0.75% of the tax assessed value and reachingin total SEK6.387 billion in 2010. It should be noted that taxes on property are the relatively insignificant as a source of government tax revenue in Sweden with 2.4% share of total tax revenues as of 2009³.

Reform history

The Swedish Homeowners Association (VillaägarnasRiksförbund) - a national organization with 322.000 members (households) throughout Sweden promoting the interests of homeowners has been lobbying for years to abolish the property taxes on one-family houses.

In 2007 Fiscal Policy Bill of proposed guidelines on the economic and budgetary policy (Regeringens proposition 2006/07:100), dated 12th of October 2006, it was mentioned that the property tax will be at

3http://www.skatteverket.se/download/18.3684199413c956649b57c0a/1361442608379/10413.pdf

3

some point abolished but no details were provided (2006/07:1, volym 1, s. 147). The projection was that

“the reform will be implemented during fiscal year 2008”.Later on, the Proposition 2006/07:100 dated 10^th of April 2007mentioned that the property tax will be abolished (effectively from the taxation year 2008) and substituted with a “fee”(“fastighetsavgift”). The proposal was to make it 0.75% of Tax Assessed Value up to SEK4.500, with the details to be announced and discussed until the fall. It was also mentioned that the reform would be funded "within the sector" but did not describe exactly how. Finally, Proposition 2007/08:27 dated 25^th ofOctober 2007 announced the actual reform with the cap at SEK6.000 (increased fromSEK4.500 announcedpreviously). The funding of the reform was planned through an increase of the capital gains tax from 20% to 22% and the introduction of an 0.5% interest charge levied on the tax liability when the capital-gains tax can be rolled forward, which is the case when it is reinvested in housing. From the reform history it can be inferred that the reform dating is rather fuzzy, thus a definite pre-reform period can be determined as the earliest possible period ranging from a newly-conducted assessment in 2006 (i.e. January 2006) till the earliest possible rumors about the reform. The problem is that the mentioned rumors may have already started from October 2006, leading to a decision of making January 2006 the earliest possible pre- reform period to examine the effect of the 2007 reform on a set of properties assessed during 2006 and onwards.

Tax capitalization theory in the context of Swedish taxation reform

The standard model of property tax capitalization is presented in this section⁴. This model will be modified to accommodate the effects of the Swedish property tax reform, and in particular the discontinuous tax treatment after October 2007 of properties sold after the reform. A description of the model departs from the idea that market value of the property equals to the present value of the stream of housing services net of property tax obligations and then introduces the necessary features to account for

4To refer to a more thorough presentation of the basic model, see Ross and Yinger (1999) as well as Yinger(2005)

4

the reform. Assuming that constant continuous housing services R⁵ over the infinite lifetime of the house embody all associated public services and other amenitiesnet of tax payment, then we can investigate the effect of the reform in the following way. Before the reform the property tax was equal to 1% of the tax assessed value, or 0.75% of the market value⁶. After the reform the property tax equals to 0.75% of tax assessed value with a cap of SEK6.000 or, in terms of assessed value, SEK800.000. To illustrate the reform effect on the tax payment, observe the graph below:

Graph 1. Tax reform illustration: tax assessed value in SEK plotted against corresponding tax obligations in SEK

The cap introduced by the reform establishes a decrease in tax obligations proportional to the properties’ tax assessed value.

Expressing this in terms of market value before the reform with the real discount factor 𝑟 − 𝑔, where constant growth rate of housing services g and the constant discount rate𝜌 are assumed, one gets:

MV₀= R₀

ρ − g + 0,0075=TAV 0,75

while after the reform, for the properties below SEK800.000 in tax assessed value:

5 Note that this assumption has implications for the resulting magnitude of tax capitalization. When land is scarce, the total supply of land may be assumed to be fixed, with the long-run supply having positive slope. This, together with the long-run output growth in the economy which is likely to raise the total demand for land, is contributing to the increasing relative prices of land used for housing purposes, and hence the market price of houses. The housing services R in this case should adjust downwards due to an increased supply. As a result, the current theoretical calculations may overstate the resulting capitalization.

6Typically the assessment process aims to determine the assessment value at 75% of the market value one year before the assessment is conducted. Property owners fill in the detailed questionnaire where the characteristics of the house via hedonic indices determine the property tax assessed value.

0 2000 4000 6000 8000 10000 12000 14000

Tax obligation in SEK

Tax Assessed Value in SEK

Tax Reform Illustration

Tax before Tax after

5

MV′ = R0

ρ − g + 0.75 ∗ 0,0075= TAV(ρ − g + 0,0075) 0,75(ρ − g + 0.75 ∗ 0,0075)

and for the properties above SEK800.000 in tax assessed value and the tax cap of SEK6.000 (assuming the cap also grows at rate

𝑔

):

MV′ = R₀

ρ − g−6.000

ρ − g=TAV(ρ − g + 0,0075) 0,75(ρ − g) −6.000

ρ − g

The difference before and after the reform for the above SEK800.000 properties is thus:

MV^′− 𝑀𝑉0=TAV ρ − g + 0,0075 0,75 ρ − g −6.000

ρ − g−TAV

0,75=0,01TAV ρ − g −6.000

ρ − g

Expressing the difference before and after in terms of a ratio, we have for the above SEK800.000 tax assessed value properties:

𝑀𝑉′

𝑀𝑉₀=

TAV (ρ−g+0,0075) 0,75(ρ−g) −^6.000

ρ−g TAV

0,75

= 𝜌 − 𝑔 + 0.0075

𝜌 − 𝑔 −6.000 ∙ 0,75 TAV(𝜌 − 𝑔)

A table presenting the above ratio of capital gain ^{MV ′}_MV depending on different values of the real discount factor ρ − g and the tax assessed value TAV is given below:

Table 1. Ratio of capital gain due to reform for the properties with the tax assessed value above SEK800.000

ρ-g

TAV

800.000 1.000.000 1.500.000 2.000.000 5.000.000 10.000.000

∞

0,01 1,188 1,300 1,450 1,525 1,660 1,705 1,750

0,02 1,094 1,150 1,225 1,263 1,330 1,353 1,375

0,03 1,063 1,100 1,150 1,175 1,220 1,235 1,250

0,04 1,047 1,075 1,113 1,131 1,165 1,176 1,188

0,05 1,038 1,060 1,090 1,105 1,132 1,141 1,150

0,075 1,025 1,040 1,060 1,070 1,088 1,094 1,100

0,1 1,019 1,030 1,045 1,053 1,066 1,071 1,075

The table shows that with increase in the tax assessed value TAV the ratio of capital gain ^𝑀𝑉′

𝑀𝑉becomes very sensitive to the discount rate ρ-g.Note that after the reform, the difference in market values between the above SEK800.000 and below SEK800.000 tax assessed value properties is:

MV_>800′ ′ − MV_<800′ ′ =TAV ρ − g + 0,0075 0,75 ρ − g −6.000

ρ − g− TAV ρ − g + 0,0075 0,75 ρ − g + 0.75 ∗ 0,0075 =

= 0,0075 ρ − g + 0,0075

ρ − g ρ − g + 0.75 ∗ 0,0075 TAV−6.000 ρ − g

6

The theoretically derived relation above will be connected to the ones derived from regression analysis in the next section.

Investigating the tax gain after the reform we should note that there are two sources of this gain to be capitalized: a 0.25%TAV decrease from a 1%TAV tax and the cap itself. Since after the reform the tax on properties is 0.75%TAV with the cap of SEK6.000, the cap in SEKto be capitalized is:

Cap = 0.75% ∙ TAV − 6000

For properties with TAV above SEK800.000, thatimplies in terms of capitalization in SEK:

Tax Gain per year = 0.25%TAV+ 0.75%TAV− 6000 = 1%TAV− 6000

or

Tax Gain =0,01TAV ρ − g −6.000

ρ − g

as noted previously. Since the tax assessed value can be assumed unchanged by the reform (it is adjusted triennially), the driving force behind a change in market value for the properties

with TAV above

SEK800.000 is the capitalized tax gain ^0,01TAV_ρ−g ^−6.000_ρ−g

.

It should be noted that while the cap introduced by the reform could be assumed to stay constant and not grow at rate

g

, so we could rewrite the tax gain as ^0,01TAV_ρ−g −^6.000

ρ and infer about

𝜌

and

𝑔

separately.

However, the policy in the aftermath of the reform (as of 2014) has indeed made an increase in the initially proposed cap and in particular made the cap indexed to a disposable income index, with the 2014 cap being SEK7.074. That implies annual cap growth rate of 2.78245%, revealing information about g that can be further used in the analysis.

Furthermore, assuming the degree of tax capitalization β, one can express the tax gain as β ^0,01TAV

ρ−g −^6.000

ρ .

Note that since the

discount rate ρ

and the degree of tax capitalization β both enter the formula for the capitalized tax gain one needs to hold either of the two fixed to infer about the degree of tax capitalization or the implied real discount rate correspondingly. That is, one can infer about the degree of tax

7

capitalization by assuming particular real discount rates constant or, holding the assumption of full tax gain capitalization, infer about the implied real discount rates.

Data and method

The data comes from transactions on all individual homes (one-family houses) between individuals from 2007 until the latest available year from the real estate agencies (SvenskMäklarstatistikAB) when the transaction was made.

The question is whether the market during the period of 2007-2008 fully capitalizes the property tax is by no means simple to answer, given the fact that the discount rate cannot be determined not having previously assumed the degree of tax capitalization and, analogously, to determine the degree of tax capitalization one should assume a particular discount rate. One could in principle derive the implied discount rate while assuming full degree of tax capitalization or, having assumed particular discount rates, investigate the degree of tax capitalization implied from the property market values. In addition, a crucial step is to compare the tax capitalization in provincial areas versus the capital city, since the testable prediction is that in Stockholm the tax capitalization will be more profound due to land scarcity and higher market values.

Sample selection

The whole sample consists of 86.830 properties sold in the years 2006-2008 and assessed in 2006.

Note that the number of properties sold that are above and below SEK800.000 are increasing during 2006 due to the fact that their relative share in the market- compared to the properties assessed in 2003 - is increasing. However, in the sense that the differences between the above SEK800.000 group and below SEK800.000 group are not affected by the assessment year per se, the sample of properties assessed in 2006 is not systematically different from the sample assessed in 2003 (consult Appendix for the details).

8

Graph 2. Illustrating the dynamics of the number of properties sold quarterly during 2006-2008 and assessed in 2006.

From graph 2 above it can be inferred that the relative share of the properties below and above SEK800.000 remains relatively stable in the reform period 2007-2008, and for the above SEK800.000 group is 52%.

Graph 3. Illustrating the relationship between the number of properties sold and the median market value 0

2000 4000 6000 8000 10000 12000 14000

Number of properties above and below SEK800.000

<800.000

>800.000

0 500000 1000000 1500000 2000000

0 2000 4000 6000 8000 10000 12000 14000

Number of properties and median Market value by quarter

Number of properties Median MV

9

Graph 4. Illustrating the relationship between the number of properties sold and the median tax assessed value

The graphs 3 and 4 show a slight decreasing trend in the market value and the tax assessed value during 2006. Comparing in to the data available from SCB, one notes that despite a slight decrease the median TAV value is in line with the one mentioned by SCB⁷. One can observe a slight increase in the median market value of properties in the treatment group after the reform. This slight increase actually occurs one quarter prior to the reform and persists for the period until third quarter of 2008:

Graph 5. Illustrating the relationship between the market value of the properties with the tax assessed value of below and above SEK800.000

7http://www.scb.se/en_/Finding-statistics/Statistical-

Database/TabellPresentation/?layout=tableViewLayout1&rxid=c576218a-468f-4b11-85ba-d5704813a83d 0

200000 400000 600000 800000 1000000 1200000

0 2000 4000 6000 8000 10000 12000 14000

Number of properties and median Tax assessed value by quarter

Number of properties Median TA value

0 500000 1000000 1500000 2000000 2500000 3000000

200601 200602 200603 200604 200701 200702 200703 200704 200801 200802 200803 200804

Market Value

Time (Quarters)

Median Market Value development over time for properties below and above SEK 800.000

MV (TAV<800000) MV (TAV>=800000)

10 Empirical analysis

Note that the degree of tax capitalization depends on supply elasticity. Poterba (1984) underlines that specification of the relationship between inputs of land and structures and outputs of housing services R is almost impossible to obtain. However, land scarcity in the capital and other major city areas vis-à-vis provincial areas is indicative for the corresponding difference in the degree of tax capitalization. In Stockholm, for example, the degree of capitalization is expected to be much higher compared to provincial areas or the whole country. The results for Stockholm presented later are supporting this prediction for the degree of tax capitalization.

In the methodology applied in the study the event (taxation reform) is a change in government policy and considered to be exogenous hence the data arise from a natural experiment. Note however the fuzzy timing of the reform which prevents a clear distinction between the pre-reform and post-reform periods.

Taking a sample of properties assessed in a particular year before the reform (2006) one can investigate if the market value of the properties with TAV above SEK800.000(“treatment group”) was affected disproportionally by the reform.

Presenting the results for Stockholm observe that all of the observations belong to the above SEK800.000 group. The following basic model in levels is first investigated:

MV_i= β₀+ β₁T + ε_i (1)

Where T denotes monthly time dummies for the whole period (2006-2008). Since this is a simple regression on a monthly dummy variable, the intercept is the average market price for single family houses in the base month of January 2006 (SEK 2.859.167 from the Table 2 in the Appendix) and the coefficient β1

on T is the difference in the average market price between houses sold in the base month of January 2006 and those sold in the month of the reform period, with the estimate indicating that on average, properties sold in for example, April 2007 are sold for SEK 1.425.000 more than those in the base moth of January 2006.

11

Adding the tax assessed values variable and the corresponding interaction term to the above basic model in levels, the following model is then estimated:

MV_i= β₀+ β₁T + β₂TAV_i+β₃T ∙ TAV_i+ ε_i (2)

It makes possible a direct comparison to the theoretically derived difference in the previous section.

Recall that:

MV^′− 𝑀𝑉₀==0,01TAV ρ − g −6.000

ρ − g

Thus β1 coefficient is related to ^6.000_ρ−g while β3 coefficient is related to^0,01_ρ−g , making possible an

inference about the implied discount rates and the degree of tax capitalization for Stockholm. In addition, with the help of coefficient β1 we can infer about the discount rate ρ, assuming the cap grows at rate g=2.78245%, while the coefficient β₃also provides inference about discount rate ρ assuming full degree of tax capitalization.Predicted negative sign for the coefficient β1is observed. Jump is noticed from January 2007 onwards, and acquires maximum magnitude around the time of Proposition 2006/07:100 dated 10th of April 2007 that mentioned that the property tax will be abolished. It should be noted that in terms of tax assessed values, the 1SEK of the tax assessed value of the properties sold in the reform period has translated in an at least 0.5SEK increase of market value of those properties compared to the base month.

For the implied discount rates and the degree of tax capitalization for Stockholm,observe the table below for ρ,ρ being consistent at around 4% inferring from both coefficients β1and β3for the period of the Proposition 2006/07:100 dated 10th of April 2007 which mentioned that the property tax will be abolished (effectively from the taxation year 2008) and substituted with a “fee” (“fastighetsavgift”):

Month (2007) ρ= g +6000/β1 ρ = g +0.01/β3

Feb 3,62% 4,38%

Mar 5,68% 5,13%

Apr 4,77% 4,57%

May 3,54% 4,02%

Jun 3,66% 4,11%

12

While for the period around Propostion 2007/08:27 dated 25 October 2007 which announced the actual reform with the cap at SEK6.000:

Month (2007) ρ= g +6000/β1 ρ = g +0.01/β3

Aug 3,77% 4,30%

Sep 6,19% 4,81%

Oct 3,54% 4,07%

Nov 3,64% 4,49%

Dec 3,57% 4,52%

Observe positive growth implied by the assumed full degree of tax capitalization and a relatively stablereal discount rate of 4% implied from the market.

Now note the implied degree of tax capitalization βimplied from two different sources, in particular:

β =^β1(ρ−g)

6000 and β =^β3_0,01^(ρ−g) :

For implied degree of tax capitalization from an assumed discount rate and coefficient β₁, that is β =^β1(ρ−g)

6000 with g= 0,027825we have:

Month (2007)

ρ

3,50% 4,00% 4,50%

Aug 72,13% 122,39% 172,65%

Sep 21,02% 35,66% 50,31%

Oct 94,53% 160,41% 226,28%

Nov 83,79% 142,18% 200,57%

Dec 91,02% 154,44% 217,86%

While the implied degree of tax capitalization from an assumed discount rate and coefficient β₃, that is β =^β3(ρ−g)

0,01 with g= 0,027825 we have:

Month (2007)

ρ

3,50% 4,00% 4,50%

Aug 47,36% 80,36% 113,36%

Sep 35,38% 60,03% 84,68%

Oct 55,68% 94,48% 133,28%

Nov 41,90% 71,10% 100,30%

Dec 41,33% 70,13% 98,93%

13

Observe that the degree of tax capitalization is proportional to the increase in the difference of real discount rate and the growth rate as well as undercapitalization for comparatively smaller discount rates.

For the entire country a following regression is run:

MVi= β0+ β1TAVi+ β2Tr + β3T+β4Tr ∙ TAVi+ β5T ∙ TAVi+ β6Tr ∙ T + β7Tr ∙ T ∙ TAVi+ εi(3)

where T denotes monthly time dummies for the whole period (2006-2008) and Tr dummy variable for the Treatment group (properties above SEK800.000), one infers to what degree the tax gain is capitalized into the market value.

Observe that the coefficient β₆ is the double interaction term of the monthly dummy with the affected properties (treatment group) dummy while the coefficient β₇ is the triple interaction term of the monthly dummy with the affected properties (treatment group) dummy and their tax assessed value TAV.

The coefficientβ7showsthe average monthly increase in market value per 1SEK in tax assessed value of the houses due to the reform, provided that properties with tax assessed values below and above SEK 800.000 did not appreciate for other reasons.

Coefficients β6 and β7also yield the degree to which tax gain is capitalized, i.e. the degree to which tax gain is translated into the increase of the market value of the properties in the treatment group given particular discount and growth rates.The estimates of these coefficients are to be compared to the theoretical difference: 0,0075 ρ−g+0,0075

ρ−g ρ−g+0.75∗0,0075 TAV−^6.000

ρ−gfrom page 5. In particular, 0,0075 ρ−g+0,0075

ρ−g ρ−g+0.75∗0,0075 is to be related to β₇ while β₆to ^6.000_ρ−g. In addition, with the help of coefficient β₆ we can infer about the discount rate

ρ, assuming the cap grows at rate g=0,0278245, while the

coefficient β7also provides inference about

discount rate

ρ, assuming full degree of tax capitalization.

After the initial reform, that is from 12 October 2006 as mentioned in Government proposition 2006/07:1 where the property tax was decided to be abolished, we would expect a kink in the relation between MV and TAV, proportionality up to SEK 600.000 (due to initially discussed cap of SEK 4.500) and a steeper slope above. Two sets of coefficients for the interaction TAV and time will be estimated, one for TAV

14

below and one for TAV above SEK600.000 (“treatment group”). Finally, since the proposition 2007/08:27 dated 25 October 2007 announces the actual reform where the cap is at SEK6.000 (not 4.500 as previously announced), that is the treatment group consists now of properties above SEK800.000. The results for the coefficient of interest (β7)are presented in the tables4-5 in the appendix (base month: Jan 2006). Observe a significant jump in market value in the case of SEK6.000 cap - that is a much more significant price impact of tax capitalization from January 2007 onwards.

For the implied discount rates and the degree of tax capitalization in the whole country,observe the table below for ρ for the period around the Proposition 2006/07:100 dated 10th of April 2007 mentioned that the property tax will be abolished (effectively from the taxation year 2008) and substituted with a “fee”

(“fastighetsavgift”):

Month

(2007) ρ=g+6000/β6

Feb 3,91%

Mar 3,71%

Apr 4,06%

May 3,99%

Jun 5,92%

While for the period around Propostion 2007/08:27 dated 25 October 2007 which announced the actual reform with the cap at SEK6.000:

Month

(2007) ρ=g+6000/β6

Aug 3,96%

Sep 3,74%

Oct 3,68%

Nov 4,45%

Dec 3,97%

Observe positive growth rates implied by the assumed full degree of tax capitalization and the approximately 4% real discount rate implied from the market.

Now note the implied degree of tax capitalizationβ implied from different discount rates and coefficients β₆ and β₇, with g=0,0278245:

15 β =β₆(ρ − g)

6000 and

β = β₇

0,0075 ρ−g+0,0075 ρ−g ρ−g+0.75∗0,0075

For implied degree of tax capitalization from an assumed discount rate and coefficient β6 , that is β =^β6₆₀₀₀^(ρ−g) we have:

Month (2007)

ρ

3,50% 4% 4,50%

Aug 61% 104% 146%

Sep 75% 127% 179%

Oct 80% 135% 191%

Nov 43% 73% 103%

Dec 60% 102% 144%

While the implied degree of tax capitalization from an assumed discount rate and coefficient β₇, that is β = _{0,0075ρ−g+0,0075}^β7

ρ−g ρ−g+0.75∗0,0075

we have:

Month (2007)

ρ

3,50% 4% 4,50%

Aug 42% 74% 107%

Sep 44% 77% 110%

Oct 44% 77% 111%

Nov 20% 35% 51%

Dec 38% 67% 97%

Observe that the degree of tax capitalization is proportional to the increase in the difference of real discount rate and the growth rate as well as the fact that the tax is capitalized less compared to Stockholm.

Several conclusions can be drawn from tables 3-5 and graphs 6-8 in the appendix. The cap of SEK6.000 resulted in a much more significant price impact regarding a tax capitalization, the point estimates become more precise in the period around the Proposition 2006/07:100 dated 10th of April 2007and improve further as one goes from a country level to the capital city level.

16 Robustness check

As a robustness check, the regional analysis is performed as follows: the provinces (kommuns) are divided into two groups: one group is consistent of the kommuns where more than 90% of the properties have a tax assessed value above SEK800.000, and the other – where more than 90% of the properties have a tax assessed value below SEK800.000. The regression for the kommuns is:

MV_i= β₀+ β₁TAV_i+ β₂Gr₁+ β₃Gr₃+ β₄T+ ¹⁰_i=5β_idouble interaction terms+ β₁₀TAV∙ T ∙ Gr₁+ β₁₁TAV∙ T ∙ Gr₃+ ε_i(4)

where T denotes quarterly time dummies for the whole period (2006-2008) and Gr dummy variable for the group that the property sold belongs to (Gr1 denotes kommuns with the majority of properties below SEK800.000, while Gr3 those above). From the regression results one infers that the tax gain is capitalized to a larger degree in geographical areas where the majority of properties are theoretically affected by the taxation reform. Consult the table below for further details (1Q 2006 is a base quarter) and observe that the timing of the jump is around the Proposition 2006/07:100 dated 10th of April 2007,mentioning that the property tax will be abolished (see table 6 in the Appendix):

It is noted that for the post-reform period the kommuns with the majority of properties above SEK800.000 show significant jump regarding their market value. Descriptive graphs for the kommuns where the majority of the properties (more than 90%) have their TAV above SEK800.000, and those where the majority of the properties (more than 90%) have their TAV below SEK800.000 are provided below. There are 24.325 properties in the kommuns where the majority of the properties have TAV above SEK 800.000, and 54.871 properties in the ones where the majority of the properties have TAV below SEK 800.000:

17

Graph 9. Geographical areas (kommuns) where properties sold over a 2006-2008 period have in majority tax assessed value above SEK800.000

Graph 10. Geographical areas (kommuns) where properties sold over a 2006-2008 period have in majority tax assessed value below SEK800.000

Results and Conclusion

The paper investigated the effect of the property taxation reform introduced in Sweden after April 2007 on the market price of the affected properties (one-family houses). Observe that the degree of tax capitalization is proportional to the increase in the difference of real discount rate and the growth rate as well as a much more profound degree of capitalization in the capital of Stockholm compared to the provincial areas. The degree of tax capitalization is also very sensitive to the real discount rate, which under

1480000 1500000 1520000 1540000 1560000 1580000 1600000 1620000 1640000

0 500 1000 1500 2000 2500 3000 3500 4000

Kommuns with the majority of the properties (more than 90%) having TAV above SEK 800.000

Number of properties Median GroupHigh TAV

0 100000 200000 300000 400000 500000 600000 700000 800000

0 1000 2000 3000 4000 5000 6000 7000 8000

Kommuns with the majority of the properties (more than 90%) having TAV below SEK 800.000

Number of properties Median GroupLow TAV

18

full degree of tax capitalization reaches 4%. An increase in real discount rate from 3,5% to 4,5% nearly doubles the degree of tax capitalization.

As a concluding remark, note that the study was performed for the entire country as well as for the capital city. Since the market price of a house includes the price of land, the tax capitalization channel may be lying through the increase of the land prices. However, in the period 2005-2008 the land prices for housing - as approximated by the price of agricultural land - were increasing at an almost constant rate

8

, something which implies that that in the reform period of 2007-2008 the driving force behind the market price increase in the housing market was a capitalized tax gain capitalized to a higher degree in the capital city of Stockholm in comparison to the rest of the country.

8See Sorensen, 2013, ”The Swedish housing market: trends and risks”, agricultural land prices analysis, page 14.

Appendix

Table 2.Simple regression (1), Stockholm

Months 2006 2007 2008

Jan 618599 495564

(585236) (579485)

Feb 183864 1.340e+06** 1.137e+06**

(609963) (551382) (563641) Mar 280833 1.357e+06** 1.236e+06**

(609963) (548694) (582802) Apr 294992 1.425e+06*** 1.484e+06***

(589280) (551115) (557201) May 846449 1.455e+06*** 1.159e+06**

(568516) (541575) (556442) Jun 380064 1.247e+06** 532917

(597320) (543816) (572196)

Jul 124833 310833 697976

(774757) (682227) (860562) Aug 347227 1.219e+06** 1.037e+06*

(571414) (558406) (601107) Sep 809485 1.367e+06** 999452*

(556442) (542568) (569921) Oct 642599 1.365e+06** 561262

(553398) (538754) (605298)

Nov 669495 931310* 705098

(564760) (540674) (607566) Dec 831083 1.068e+06* 459333

(660715) (561570) (660715)

Constant 2.859e+06***

(522341)

N 3157

R-squared 0,043

Table 3.Estimation results for regression (2), Stockholm

Coefficient: β1 (T) Coefficient: β3 (TAV_x_dT) Coefficient:

β2 (TAV) Constant

Months 2006 2007 2008 2006 2007 2008

Jan ^{-664,728* -247,322 0.558*** 0.376 1.731*** -190,470}

^{(378,351) (488,459) (0.213) (0.285) (0.0658) (162,150)}

Feb ^-9,587 -718,142** -1.099e+06*** 0.0285 0.627*** 0.889***

^{(303,410) (317,121) (421,802) (0.165) (0.162) (0.227)}

Mar ^{734,618** -206,700} -1.062e+06** -0.384** 0.426*** 0.931***

^{(294,370) (310,497) (458,567) (0.151) (0.152) (0.256)}

Apr ^{-160,501 -301,685 -388,217 0.0788 0.560*** 0.606***}

^{(302,717) (326,199) (310,723) (0.163) (0.172) (0.158)}

May ^-189,174 -790,819** -391,204 0.162 0.805*** 0.596***

^{(244,567) (330,546) (289,647) (0.112) (0.168) (0.150)}

Jun ^-13,078 -683,902** -657,360** 0.0177 0.751*** 0.737***

^{(328,701) (339,332) (319,163) (0.164) (0.178) (0.175)}

Jul ^{-562,141 214,066} 1.351e+06*** 0.376 0.219 -0.529**

^{(633,834) (425,923) (368,207) (0.364) (0.244) (0.209)}

Aug ^49,102 -603,122** 505,692 0.0122 0.660*** 0.0290

^{(208,283) (281,990) (372,597) (0.102) (0.137) (0.207)}

Sep ^{-134,829 -175,738 -458,150 0.157 0.493*** 0.531***}

^{(255,987) (272,839) (341,028) (0.123) (0.138) (0.170)}

Oct ^-169,233 -790,467** -280,652 0.259* 0.776*** 0.325***

^{(273,376) (325,155) (212,289) (0.143) (0.167) (0.0989)}

Nov ^-155,360 -700,672** -361,526 0.241 0.584*** 0.352***

^{(346,951) (272,589) (222,142) (0.185) (0.140) (0.107)}

Dec ^{562,262 -761,052* -110,483 -0.134 0.576** 0.155}

^{(436,447) (439,715) (348,052) (0.249) (0.232) (0.171)}

N ³¹⁵⁷

R-squared ^0.831

Graph 6.Coefficient β3 point estimates and confidence intervals over a 2006-2008 period, Stockholm

Graph 7.Coefficient β7 point estimates and confidence intervals over a 2006-2008 period, cap SEK4.500

Graph 8.Coefficient β₇ point estimates and confidence intervals over a 2006-2008 period, cap SEK6.000 -1.5

-1 -0.5 0 0.5 1 1.5

-1.5 -1 -0.5 0 0.5 1 1.5

-1.5 -1 -0.5 0 0.5 1 1.5

Table 4. Coefficient β7, the cap of SEK4.500 Table 5. Coefficient β7, the cap of SEK6.000 Coefficient: β7 (TAV_x_dTr_x_dT) Coefficient: β7 (TAV_x_dTr_x_dT)

Months 2006 2007 2008 Months 2006 2007 2008

Jan ^{0.121 0.292} Jan ^{0.358 0.535**}

(0.286) (0.292) (0.225) (0.271)

Feb ^{0.296 0.0606 0.229} Feb ^{0.127 0.475*** 0.390**}

(0.269) (0.234) (0.246) (0.221) (0.183) (0.185)

Mar ^{-0.134 0.335 -0.198} Mar ^{-0.0400 0.564*** 0.0171}

(0.313) (0.241) (0.320) (0.196) (0.188) (0.329)

Apr ^{-0.251 0.181 0.168} Apr ^{0.252 0.407** 0.485***}

(0.268) (0.240) (0.229) (0.194) (0.185) (0.174)

Maj ^{-0.0623 0.188 0.104} Maj ^{0.205 0.467*** 0.277}

(0.264) (0.227) (0.262) (0.195) (0.173) (0.247)

Jun ^{-0.216 -0.0479 -0.173} Jun ^{0.329* 0.195 0.0911}

(0.250) (0.264) (0.286) (0.197) (0.251) (0.281)

Jul ^{0.0564 -0.268 -0.604**} Jul ^{0.230 -0.154 -0.513*}

(0.254) (0.310) (0.281) (0.203) (0.329) (0.270)

Aug ^{-0.0819 0.315 -0.0994} Aug ^{0.135 0.504*** 0.197}

(0.253) (0.223) (0.229) (0.186) (0.177) (0.180)

Sep ^{-0.0398 0.179 0.0471} Sep ^{0.176 0.522*** 0.406**}

(0.246) (0.232) (0.228) (0.178) (0.186) (0.182)

Okt ^{-0.170 0.172 -0.127} Okt ^{0.201 0.526*** 0.191}

(0.230) (0.223) (0.232) (0.175) (0.177) (0.185)

Nov ^{0.0890 0.0654 -0.00662} Nov ^{0.191 0.241 0.311*}

(0.256) (0.227) (0.225) (0.188) (0.174) (0.177)

Dec ^{-0.238 0.0749 0.154} Dec ^{0.138 0.457** 0.408**}

(0.258) (0.239) (0.245) (0.191) (0.178) (0.199)

N 86830 N 86830

R-squared ^{0.833 R-squared 0.834}

Table 6.Coefficients β10 and β11 for the period 2006-2008 Year-Quarter Coefficient:

β10(TAV_T_Gr1)

Coefficient: β11 (TAV_T_Gr3)

2006-Q2 0.0636 0.134

(0.151) (0.140)

2006-Q3 -0.0682 -0.0114

(0.143) (0.133)

2006-Q4 -0.181 -0.115

(0.134) (0.124)

2007-Q1 -0.123 0.139

(0.130) (0.121)

2007-Q2 -0.0608 0.360***

(0.127) (0.118)

2007-Q3 -0.0896 0.313**

(0.131) (0.122)

2007-Q4 0.000346 0.254**

(0.127) (0.118)

2008-Q1 -0.642*** -0.184

(0.129) (0.120)

2008-Q2 0.186 0.691***

(0.126) (0.116)

2008-Q3 -0.409*** 0.00123

(0.130) (0.122)

2008-Q4 -0.0561 0.174

(0.129) (0.121)

N 86830

R-squared 0.835

Tables show the median Tax Assessed Values and number of properties sold by sales year and assessment years. Last rows and columns are, correspondingly:

median TAV of the properties sold during the whole period, by assessment year (2003-2012), median TAV of the properties sold for all assessment years, by year (2005-2012):

Sales year Median TAV by : Assessment year

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Median

2005 663000 662000 913000 1055000 669000

2006 607000 591000 619000 1040000 1975000 689500 717000

2007 498000 475000 511000 853000 1031000 828000

2008 605000 483000 445000 855000 805000 1046000 1175000 832000

2009 657000 707000 424000 924000 870500 900000 1029000 636500 936000

2010 711000 358000 646500 874000 790000 820000 1176000 838000 401500 1132500

2011 707000 2130000 394000 727000 597000 847000 1177000 1186000 865500 1493500 1173000

2012 514000 454000 779000 530000 889000 1232000 1190500 1250000 1128000 1222000

Median 637000 652000 613000 893000 818000 891000 1167000 1173000 1232000 1133000

Sales year Number of transactions by : Assessment year

Total

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

2005 15510 17407 383 444 33744

2006 10126 2644 14487 11501 8 2 38768

2007 716 127 1579 39274 89 41785

2008 79 22 113 18644 17240 72 111 36281

2009 15 2 13 12250 3622 12977 8956 28 37863

2010 7 3 6 951 173 2014 35542 650 28 39374

2011 3 1 5 121 46 175 20018 15007 482 90 35948

2012 4 1 35 15 60 13682 4192 14314 3498 35801

Total 26460 20206 16587 83220 21193 15300 78309 19877 14824 3588 299564

The observations in bold, totaling 86.830, are the ones used in the regression analysis. Note from the first table that Median TAV of the properties sold during the whole period and assessed in 2006 is 46% higher compared to the Median TAV of the ones assessed in 2005.