Journal of Housing Economics 33 (2016) 128–142

Contents lists available at ScienceDirect

Journal of Housing Economics

journal homepage: www.elsevier.com/locate/jhec

A microsimulation of property tax policy in China

�

Jing Cao

∗, Wenhao Hu

School of Economics and Management, Hang Lung Center for Real Estate, Tsinghua University, 10 0 084 Beijing, China

a r t i c l e i n f o

Article history:

Received 20 July 2015

Revised 7 May 2016

Accepted 9 May 2016

Available online 16 May 2016

JEL classification:

H20

H27

P25

Keywords:

Property tax reform

Distributional effects

Incidence

Welfare analysis

China

a b s t r a c t

China is exploring possible property tax reform to stabilize the booming housing market as

well as providing sustainable revenue for the local government. In this paper we develop

a theoretical model of property tax reform to decompose potential impacts of property tax

reform in China. Then we used the China Family Panel Survey (CFPS) data to conduct a

microsimulation model to examine potential impacts and incidences of alternative prop-

erty tax designs in China. Our analyses suggest that a uniform property tax policy would

bring substantially heterogeneous impacts across different income groups as well as differ-

ent regions, mainly due to the differences in income distribution, housing prices and the

degree of the Housing Demolition program. In terms of property tax incidence, our simula-

tion suggests that utilizing tax revenue on the poor’s public housing subsidy may mitigate

the regressivity; in some case may even increase the overall social welfare. Finally, we use

the cross-sectional information in the Chinese Family Panel Survey (CFPS) data to simulate

for optimal tax scenarios for each region. Our microsimulation results provide some ini-

tial quantitative analysis in the literature and may shed some light on understanding the

impacts of future property tax reform in China.

© 2016 Elsevier Inc. All rights reserved.

1

b

p

m

c

c

s

C

c

t

(

T

D

g

h

1

. Introduction

With the housing boom for the last decade, China has

een involved in active discussion about imposing the

roperty tax or real estate tax to stabilize the housing

arket. Property tax is quite common in many developed

ountries, as a primary source of tax revenue for the lo-

al governments for the provision of local public goods,

uch as public school, police system and etc. However, in

hina, property tax still has not been implemented yet, ex-

ept some real-estate related taxes charged at the point of

ransaction.

� This project is supported by National Science Foundation of China

Project codes: 71422013 and 71173130), Hang lung Center for Real Estate,

singhua University Research Center for Green Economy and Sustainable

evelopment, and Tsinghua University Initiative Scientific Research Pro-

ram. ∗ Corresponding author.

E-mail address: caojing@sem.tsinghua.edu.cn (J. Cao).

ttp://dx.doi.org/10.1016/j.jhe.2016.05.004

051-1377/© 2016 Elsevier Inc. All rights reserved.

Recently, due to the concern about soaring housing

bubble, tight local government budget, rising income gap

between the rich who own many houses and the poor who

can hardly afford the rent, some pilot reforms of property

tax was implemented in Shanghai and Chongqing in 2011.

Before these experiments, property tax was just applied

to some business buildings or to the foreign companies in

China. Right now, whether these pilot programs should be

extended to the whole nation is still a highly debated open

question. Experiences from the developed world, such as

US and OECD countries, have provided some empirical evi-

dences that property tax can play a significant role in con-

trolling the housing price, restraining certain speculative

behavior ( Kuang et al., 2012 ).

Therefore, an open question is, if China implement the

property tax, how would various households being affected

under this new reform? Will they get worse off since their

out of pocket money will increase for paying the new tax?

If the landlord raise the housing rent, in an inelastic rental

market, renters may bear most of the burden, so it is likely

J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142 129

t

s

t

g

t

t

o

f

h

d

t

g

t

h

k

a

t

C

c

u

i

p

d

c

s

t

b

c

e

a

p

2

w

t

c

g

c

e

p

“

e

i

N

a

t

b

t

t

t

c

h

t

o

W

v

(

w

t

p

a

i

T

s

f

t

i

t

m

a

p

e

c

(

a

h

p

t

p

i

p

b

a

d

i

j

e

t

f

j

s

o

l

w

s

y

e

l

c

B

t

o

f

S

t

i

l

t

H

s

p

i

a

u

o be regressive. However, it might not be exactly true,

ince some poor households may get better off if they are

he beneficiaries of extra revenue spent on local public

oods. That is, the local government now with more local

ax revenue may provide more local public good, like bet-

er public primary and middle school, or increase supply

f cheaper subsidized housing projects for the poor. Un-

ortunately, the experiments in the two pilot cities (Shang-

ai and Chongqing) have not shed much light on potential

istributional effects considering such a diverse distribu-

ion of rich, medium and poor cities as well as big income

ap between the poor and rich in China. Therefore, given

he heterogeneity nature of the local property tax policies,

ow to design a proper property tax reform is also the

ey to the success of this policy. If we take both equity

nd efficiency criteria into consideration for the property

ax reform, how should we implement the property tax in

hina?

To address these issues, in this paper we exploit a mi-

ro data from the China Family Panel Survey (CFPS) to sim-

late for possible outcomes of various property tax policies

n China. We then use the microsimulation results to com-

are the effects of various property tax regimes regarding

ifferent designs on tax bases, tax rates as well as recy-

ling policies. Our results show that, given China’s broad

pectrum on the housing market, and increasing gap be-

ween the rich and poor, a uniform property tax may not

e appropriate for China, it is important to take into ac-

ount many local factors such as local housing prices, in-

quality patterns, to design more appropriate tax policies

nd recycling regimes to achieve better policy outcome in

ractice.

. Literature review

The impact of property on property value has been

idely studied in the developed countries. Three alterna-

ive views of the incidence of property tax have been dis-

ussed in the literature: the “traditional view”, which ar-

ues that the property tax would fully shift forward to

onsumers in the form of higher housing prices; the “ben-

fit view”, which suggests that the property tax is simply a

ayment for local public services, and finally the so-called

new view” of the property tax, which argues that prop-

rty tax is implicitly a distortionary tax on the use of cap-

tal within a local jurisdiction.

The traditional view dates back to Simon (1943) and

etzer (1966) , who took a partial equilibrium approach to

nalyzing the tax, focusing on the effects of increasing the

ax in a local housing market. The burden of the tax is

orne by consumers, and the traditional view argues that

his entire burden is borne by local housing consumers in

he form of higher housing prices, therefore implying that

he property tax inefficiently reduces the size of the lo-

al housing stock and its burden is borne in proportion to

ousing consumption.

However, the “benefit view” argues that the property

ax is beneficial to the consumers. This view was devel-

ped initially by Hamilton (1975, 1976) , Fischel (1974) and

hite (1975) , and is reviewed by Hamilton (1983) . This

iew is an extension of the famous Tiebout Model. Tiebout

1956) first mentioned the concept of “voting by feet”,

hich suggests that if consumers are fully mobile, then

hey will move to the community where their preference

atterns are best satisfied. Because of the tax completion

mong the communities, the local governments have more

ncentives to supply public goods and lower the tax rate.

iebout ignored local property taxation and instead as-

umed the existence of benefit taxes is implicitly in the

orm of head taxes. Following Tiebout, Hamilton assumed

hat individuals are sorted into local jurisdictions accord-

ng to their demands for local public services, and that

here are enough local tax expenditure packages to accom-

odate the heterogenous individual preference. Hamilton

rgued that such a “perfect capitalization” converts the

roperty tax into a benefit tax, at least in the long run

quilibrium (but not at the time when a tax change oc-

urs and is capitalized into property values). Yinger et al.

1988) have found empirical evidence that property taxes

nd local public service expenditures are capitalized into

ouse values, as predicted by the Hamilton model. The im-

lications of this “benefit view” are striking. First, it means

hat the property tax is effectively a user charge that is

aid in exchange for the benefits of local public service. It

s thus a nondistortionary tax. Second, as a benefit tax, the

roperty tax has no effects on the distribution of income.

The “new” view of the property tax, first developed

y Mieszkowski (1972) , subsequently extended by Zodrow

nd Mieszkowski (1986) , argues that the property tax is a

istortionary tax on the local use of capital, which results

n a misallocation of the national capital stock across local

urisdictions. Mieszkowski (1972) stressed that the partial

quilibrium analyses of the property tax that characterized

he traditional view was misleading, since it ignored the

act that the property tax was used by virtually all local

urisdictions and applied to a large fraction of the capital

tock. Zodrow and Mieszkowski (1986) suggest that the use

f a distorting property tax on mobile capital decreases the

evel of residential public services.

From the theoretical perspective, higher property tax

ill lower the property value. Oates (1969 ) uses the data

et of 53 northeastern New Jersey communities for the

ear of 1960, and finds that the property tax has a negative

ffect on the property value. Palmon & Smith (1998) fol-

ow the work done by Oates, and specify the importance of

apitalization; they also find the tax has a negative effect.

ai and Ouyang (2014) exploit the effect of property tax on

he housing price, taking advantage of a policy experiment

f property taxes in Shanghai and in Chongqing starting

rom January 2011. The counterfactual housing prices in

hanghai and Chongqing without the tax are estimated by

he housing prices of the strongly correlated provinces us-

ng a long monthly time series data. They find that the tax

owers the average housing price by 15% in Shanghai.

The decline in property value may further drag down

he consumption through the “housing wealth effect”.

arberler (1958 ) first put the property value into a con-

umption model, and notice that the change of the housing

rice will change the wealth of the residents. Then the res-

dents will adjust their consumption choice as well. Ludwig

nd Slok (2002) specify the realized wealth effect and the

nrealized wealth effect. Two channels through which the

130 J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142

p

u

f

e

2

C

U

t

t

t

p

C

t

o

o

t

e

m

c

g

h

p

h

w

m

d

i

w

M

a

h

a

c

h

a

i

e

M

a

t

t

c

m

f

F

u

o

o

t

b

s

a

i

s

b

C

i

u

S

roperty value will affect the consumption choice are fig-

red out: Liquidity constraints effect and Substitution ef-

ect. A lot of literatures have discussed how the prop-

rty value will affect the consumption choice. ( Muellbauer,

0 07; Muellbauer and Murphy, 20 08; Iacoviello, 2011 ).

ase and Quigley (2005 ) use the quarterly panel data of

.S. states during the 1980s and 1990s, and found that in

he U.S., the marginal propensity to consume with respect

o the property value is about 0.06 . However, in China

hese so called “housing wealth effects” are more com-

licated. Du Li (2010 ) uses a panel data with 172cities in

hina from 2002 to 2006, and finds that the increase of

he housing price actually decreases the total expenditure

f the residents. They argue that, unlike US, Chinese house

wners are more likely to be the potential buyers later, so

he rising housing price may further crow out consumption

xpenditures, so they can save more to buy more apart-

ent. In a recent paper , Yang (2014 ) considers the dual

haracters of the housing in China: both a consumption

ood and an investment good. She argues that living in a

ouse the residents not only pays the renting cost but also

ays the opportunity cost. The opportunity cost in China,

owever, is much higher, compared to the renting cost,

hen property tax is non-exist. According to Yang’s esti-

ate, the elasticity of the property value on the expen-

iture is −12.3% . They also highlight that, the heterogene-

ty in China makes it much harder to estimate the house

ealth effect.

Microsimulation is a useful tool for evaluating policies.

icrosimulation models allow simulating the effects of

policy on a sample of economic agents (individual,

ouseholds and firms) at the individual level. Bourguignon

nd Spadaro (2006) mentioned that the usefulness of mi-

rosimulation techniques in the analysis of public policies

as two aspects. First is the possibility of fully taking into

ccount the heterogeneity of economic agents observed

n micro-datasets. Second is the possibility of accurately

valuating the aggregate financial cost/benefit of a reform.

icrosimulation is widely used in welfare accounting

nd policy evaluating. Immervoll and Kleven (2007 ) use

he method to compares the effects of increasing tradi-

ional welfare to introducing in-work benefits in the 15

ountries of the European Union. By the microsimulation

odel, they can quantify the equity-efficiency trade-off

or a range of elasticity parameters and get the results.

or the housing market, Brownstone and Englund (1985 )

se the microsimulation method to estimate the effects

f the Swedish1983–85 tax reform on the demand for

wner-occupied housing.

In this paper, we apply the Slesnick–Jorgenson model

o calculate the changes in social welfare by combining

oth the efficiency and distributional equity criteria. The

ocial welfare function was first introduced by Jorgenson

nd Slesnick (1983) . This translog function is widely used

n estimating the welfare change caused by policies. It con-

iders both efficiency and equality, and the income distri-

ution is implicitly incorporated in the welfare calculation.

urrently, this method was been applied to estimate the

nfluence of the petroleum tax and natural gas price reg-

lation on the social welfare in the U.S. ( Jorgenson and

lesnick 1985a, 1985b ).

3. The pilot property tax reform in China

In 2011, a policy experiment of property tax took place

in Shanghai and Chongqing. Before that, property tax was

just charged on the foreign companies in China. The main

goal to impose property tax is to stabilize the housing

price and increase the financial income of the local govern-

ments. The two pilot experiments did have some impacts

on the property value. The central government in China

has not yet decided whether to impose the property tax at

nation-wide, but the new housing registration system and

network is currently under development to prepare for the

future property tax reform. There are four major issues to

consider before the implementation of property tax.

The first issue is about the tax base. In the U.S., the tax

base is the total value of the houses. However, in China, it

may be politically difficult to use housing value as tax base.

Such a tax would impose a very high burden for many

households, especially the poor who own the house but

may purchase earlier when the housing price is relatively

low. At the tax rate of 0.5%, a household with a house of

$20 0,0 0 0 will have to pay $10 0 0 per year. However, the av-

erage income per household in China in 2014 is only about

$15,0 0 0, which means the property tax may lower the ex-

penditure by 6.7%. Second, fairness is always an important

concern of the tax. The ratio of the property value to the

total wealth is likely to decrease as the total wealth in-

creases. So if we use the total value of the houses as the

tax base, then the property tax is more likely to be a re-

gressive tax. Third, in order to win more political support,

it is likely for the government to set a level under such

standard the property tax is waived, why above which the

owner has to pay for the extra tax, as implemented by the

pilot projects in Shanghai.

The second issue to consider is the tax rate. If the tax

rate is too low, it cannot achieve the goal to stabilize the

housing price. If the tax rate is too high, the property tax

would crowd out the consumption, and worsen the econ-

omy. In addition, a high property tax rate may lower the

property value by a large amount, which may hurt the in-

dustries not only real estate, but also construction, iron

and steel, cement, as well as all the other industries in a

chain. In some cases, the sharp decline in property value

may even trigger financial crisis, as the Asian financial cri-

sis happened in 2008. Still very few paper debate on what

would be the optimal property tax, but in practice it has

been widely accepted that it should be around 1%.

The third issue to consider is whether to set a uniform

rule of property tax policy on the whole China? In China,

regional disparity is becoming more and more severe. Ex-

cept Beijing, Shanghai, Shenzhen and Guangzhou, there is

a surplus of housing in many medium and small cities.

Therefore, a uniform tax rate of 0.5% or 1% may even lead

the real estate market crash and the ratio of empty apart-

ment rising. So whether to impose a heterogenous prop-

erty tax and let the local government to determine is also

an open question to ask.

The fourth issue is to determine how to use the in-

come from the property tax. In the U.S., property tax is

the major income resource of the local governments; most

of the income from the property tax is spent on the local

J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142 131

p

c

S

l

d

e

o

t

t

c

e

a

i

t

t

t

h

i

4

h

h

n

L

c

T

i

e

b

e

c

t

m

s

D

τi

t

x

∑

w

c

n

m

g

t

s

Table 1

Decomposition of the effects of property tax.

Term Name Sign

−θ · �τ/ ( r + τ j ) Tax out of pocket effect −−θ · �P j / P j (−θ · �Z j / Z j ) Price effect +

ɛ ep · �P j / P j Housing wealth effect −μ · �LG ij /LG ij Redistribution effect +

�V ij Overall effect ?

p

m

K

p

K

r

a

w

c

h

s

h

k

F

p

d

t

s

P

ɛ

v

V

�

i

e

ublic goods like education. In China, how to use the in-

ome remains under debate. In the policy experiment in

hanghai, the income was used for the supply of the pub-

ic renting houses for low income groups. So it is also a

ebate whether to use property tax revenue to subsidize

ducation or to subsidize public renting houses is also an

pen question.

Though the exact proposal of implementing property

ax remains unknown yet, there is a high expectation that

he property tax will be implemented in near future. The

entral government of China has decided to form the real

state registration system, and the system begins to work

t March 1st 2015. This system is believed to build as an

nfrastructure for the next step of property tax reform for

he whole country.

To shed some lights on potential impacts of property

ax reform in China, next we describe a theoretical model,

hen we will conduct a micro-simulation model using the

ousehold housing survey data to simulate for potential

mpacts.

. The model

Consider an economy with J cities. In city j , there are N j

ouseholds. We assume that capital is mobile. Each house-

old i consumes three kinds of goods, housing service x 1 ij ,

umeraire non-housing goods x 2 ij and local public goods

G ij . We assume that each household i has a preference on

ity j , ɛ ij , which is different among households and cities.

he distribution of ɛ ij is continuous. Though the household

may gain different incomes in different cities, the differ-

nces are not influenced by property tax and somehow can

e reflected in the taste term ɛ ij . Then without loss of gen-

rality, we can assume the household i gains the same in-

ome in all the cities.

We first focus on the consumer behaviors. The migra-

ion effect will be discussed later.

ax u i j = θ · ln x 1 i j + ( 1 − θ ) · ln x 2 i j + μ · lnL G i j + ε i j

. t . ( Z j + τ j P j ) · x 1 i j + x 2 i j = e i (4.1)

The utility function of a household is in the Cobb-

ouglas form, where Z j is the rent of the housing service,

j is the tax rate of the property tax which is 0 at first, P j s the price of houses per unit, and e ij is the total expendi-

ure of the household.

1 i j ∗ =

θ

Z j + τ j P j e i , ·x 2 i j

∗ = ( 1 − θ ) e i (4.2)

∞

t=0

Z j

( 1 + r ) t

= P j → Z j = r P j (4.3)

here r is the interest rate.

Houses can be treated as durable goods. The houses

onsumed today from the existing housing stock and

ewly built ones.

In the production of house, capital and land are the two

ain inputs. In China, the supply of land is limited by the

overnment, which is fixed in the production. The real es-

ate company will determine how much capital and land

hould be used in the production of house to maximize its

rofit:

ax K 1 j

�1 j = P j K 1 j γ L j

1 −γ − r K 1 j − r j L j (4.4)

1 j is the capital invested in construction in city j, r j is the

rice of land in city j , L j is the land which is fixed.

1 j ∗ = L j ·

(γ P j

r

) 1 1 −γ

· Y 1 j ∗ = L j ·

(γ P j

r

) γ1 −γ

(4.5)

From the zero profit condition, we can get

j ∗ = ( 1 − γ ) ·

(γ

r

) γ1 −γ · P j

1 1 −γ (4.6)

What is the impact of property tax on property value

nd household utility?

We first analyze the medium and short term cases

hile in the long term households may migrate to other

ities. We assume that, in the medium term, the house-

olds can adjust their consumption choice, while in the

hort term the households are unable to adjust how much

ousing service they consume.

For the medium term, the equilibrium of housing mar-

et clears as follows:

θ

( r + τ j ) P j e i · N j = ( 1 − δ) F j + L j ·

(γ P j

r

) γ1 −γ

(4.7)

j is the houses volume in city j last period, δ is the de-

ression rate. The left hand side of the equation is total

emand of houses in this period and the right hand side is

he total supply of the houses in this period.

We can get the derivative of property value with re-

pect to the tax rate.

˙ J = −

{1

P j ( 1 − ε ep ) +

r + τ j

θ · e i j

· L j

N j

· γ

1 − γ·(γ

r

) γ1 −γ· P j

γ1 −γ

}−1

· 1

r + τ j

< 0 . (4.8)

ep is the elasticity of expenditure with respect to property

alue, as a result of housing wealth effect.

The indirect utility function of the household is

i j = ln e i − θ ln ( r + τ j ) − θ ln P j

+ θ ln θ + ( 1 − θ ) ln ( 1 − θ ) + μ · ln L G i j + ε i j (4.9)

V i j = −θ�τ

r + τ j

− θ�P j

P j + ε ep ·

�P j

P j + μ

�L G i j

L G i j

(4.10)

The impact of the property tax on the household’s util-

ty can be decomposed into four parts ( Table 1 ). The first

ffect is the tax out of pocket effect. As the tax rate of the

132 J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142

p

t

T

c

p

v

s

u

e

v

w

i

m

l

f

e

t

t

v

r

t

u

s

�

m

t

p

g

t

s

c

o

�

o

�

i

o

p

t

Table 2

The difference of the utility change between city j and k .

�V i j − �V ik P j P k

< ( L k N k

/ L j N j

) 1 −γ P j P k

> ( L k N k

/ L j N j

) 1 −γ

ɛ ep < θ + −ɛ ep > θ − +

W

roperty tax increases, the household needs to pay more

ax out of pocket, so the changes on utility are negative.

he second and the third effect are all related with the

hanges of the housing price. When the tax rate of the

roperty tax increases, the rents decrease as the property

alue decreases. The total expenditure on residential con-

umption will decrease, which increases the household’s

tility. This second effect we call it “Price Effect”. The third

ffect is the Housing wealth Effect. When the property

alue falls, the household may feel the loss in the wealth,

hich could lower their total expenditures. The last term

s called the “Redistribution Effect”. Since the local govern-

ent can use the revenue of the property tax to supply the

ocal public goods, then the welfare change would be dif-

erent on local residents since some would enjoy the ben-

fits while others may not. Given these four potential fac-

ors, the sign of the total effect remains unknown.

In the short term, the impact would be smaller. Since

he housing consumption choice is fixed, the property

alue may decline due to the rational expectation of the

ents. If the property value is unchanged in the medium

erm, then the utility is also the same.

i j = θ · ln x 1 i j + ( 1 − θ ) · ln [ e i − ( r + τ j ) · P j · x 1 i j ]

+ μ · ln L G i j + ε i j (4.11)

.t. x 1 ij is fixed

V i j = ε ep ·�P j

P j − θ

�τ

r + τ j

− θ�P j

P j + μ

�L G i j

L G i j

(4.12)

In the long term, we assume that the households can

igrate according to the Tiebout Hypothesis. We will see

he migration incentive of a household by comparing the

otential utility changes in different cities when the mi-

ration has not happened. To make it simple, we assume

he tax rate is same in all the cities, and we do not con-

ider the Redistribution Effect.

According to Eq. (4.10) , the difference of the utility

hanges under property tax if household i lives in city j

r city k is:

V i j − �V ik = ( θ − ε ep )

[− �P j

P j −

(− �P k

P k

)](4.13)

So we can see the difference depends on the differences

f the change rates of the housing prices.

Using Eq. (4.8) , we can get:

V i j − �V ik = ( θ − ε ep ) · �τ

r + τ

·{ [

1 − ε ep +

r + τ

θ · e i · γ

1 − γ·(γ

r

) γ1 −γ · L j

N j

· P j 1

1 −γ

]−1

−[

1 − ε ep +

r + τ

θ · e i · γ

1 − γ·(γ

r

) γ1 −γ · L k

N k

· P k 1

1 −γ

]−1 }

(4.14)

From Table 2 , we can see how the difference of the util-

ty change between city j and k changes with the change

f housing wealth effect, housing prices and land area

er capita. When ε ep is not large enough, which means

hat the housing wealth effect is dominated, then a city

with a lower level of L j N j

· P j 1

1 −γ become more preferable

under property tax. And when the housing wealth effect

is large enough, the result reverses. The empirical results

show that θ is usually higher than 20%, while ɛ ep is usually

smaller than 10%. So we can focus mainly on the case ɛ ep

< θ . From Eq. (4.8) , a city with a lower level of L j N j

· P j 1

1 −γ

will have larger decrease of housing price in percentage,

which shows it becomes more preferable under property

tax. And we notice that housing price has a stronger ef-

fect than the land per capita relatively. Though a city be-

comes more preferable does not mean a household will

surely move to the city, because of the preference of the

household, the continuity of the preference’s distribution

indicates that some migrations may occur from the cities

with higher levels of L j N j

· P j 1

1 −γ to those with lower levels.

In this paper, besides the decomposition calculation, we

use the Slesnick–Jorgenson model to calculate the changes

in social welfare by combining both the efficiency and dis-

tributional equity criteria.

The indirect utility of household k can be estimated in

the form:

ln V k = ln p

′ · αp + 0 . 5 ∗ ln p

′ · B pp · lnp

− D (p) · ln

[M k

m 0 ( p, A k )

](4.15)

p is the price vector of the consumption sectors, B pp is the

demand elasticity matric, D ( p ) is a function of p, M k is the

total expenditure of household k, A k is the vector of de-

mographic variables and m 0 ( p, A k ) is the equivalent scale

of household k which is a function of prices and demo-

graphic variables. Because we are now considering the util-

ity of a household, we have to rule out the influence of the

household’s size. The equivalent scale can normalize all the

households to a standard household.

This translog indirect utility function is used widely in

measuring a household’s utility. Then we try to measure

the social utility.

The social welfare function we use here is:

= ln ̄V − γ (x ) ·[∑ K

k =1 m 0 ( p, A k ) · | ln V k − ln V̄ | −ρ

∑ K k =1 m 0 ( p, A k )

]− 1 ρ

(4.16)

In ̄V = In p

′ · 0 . 5 ∗ In p

′ · B pp · In p − D(p)

·∑ K

K=1 m 0 (p, A k ) · In

[M k

m 0 (p,A k )

]∑ K

K=1 m 0 (p, A k ) (4.17)

γ ( x ) is a function with respect to the allocation. This so-

cial welfare considers both efficiency and equality. The first

term is the average utility each household can get if the

J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142 133

Table 3

Summary statistics of CFPS data (2010).

Variable Obs. Mean Std. dev Min Max

ID 6609 344,489 130,647 110,001 621,842

Sample weight (number of households represented), 6609 84,032 92,609 439 2,107,389

Total expenditure, 6609 48,208 65,237 2100 996,0 0 0

Household size 6609 2.64 1.24 1 10

Number of students 6609 0.25 0.51 0 4

Live in own house(1:Yes; 0:No) 6609 0.79 0.41 0 1

Areas of the house they live in (m

2 ) 6609 105.7 83.9 5 10 0 0

Number of extra houses owned 6609 0.21 0.5 0 11

Areas of extra houses owned (m

2 ) 6609 20.62 63.64 0 1500

Value of extra houses owned (10,0 0 0 RMB) 6609 8.74 34.53 0 800

Value of the house they live in (10,0 0 0 RMB) 6609 42.67 56.88 0.6 720

Own-occupied housing rents (RMB/month) 6609 892.64 1301 50 30,0 0 0

e

a

c

i

i

d

e

5

5

t

T

p

t

m

j

t

9

m

w

c

n

t

o

p

T

h

w

p

m

a

m

t

s

e

o

o

o

t

a

o

r

t

a

e

o

e

p

h

t

t

t

T

s

t

o

a

o

i

g

p

g

W

p

m

d

r

C

D

t

t

w

2 Northeast China : Liaoning, Jilin, Heilongjiang; Centre China : Shanxi,

xpenditures are equally distributed among the households

ccording to their equivalent scales, which represents the

oncern to efficiency. The second term is the loss from the

nequality. ρ is from −1 to −∞ , representing the relative

mportance of equality in the social welfare. ρ is called the

egree of aversion of inequality. When ρ is −1 , the great-

st possible weight to the equity consideration is given.

. Data and microsimulation

.1. Data

The data we use to calculate the impact of property

ax is from the China Family Panel Survey (CFPS) 1 in 2010.

he data set contains 6827 urban households from 25

rovinces in China. This data set is uniquely constructed

o be nearly national representative sample, only some re-

ote areas are not included in the sample such as Xin-

iang, Tibet, Qinghai, Inner Mongolia, Ningxia and Hainan;

he remaining sampling 25 provinces counts for about

4.5% of total population in China. The CFPS was conducted

ulti-stage, and then in each stage implicit stratification

as employed. Appropriate sampling weights have been

onstructed to adjust for both sampling design and survey

on-responses. The detail sampling and weight construc-

ion was discussed in Xie and Lu (2015) .

The reason why we choose the data in 2010 is to rule

ut the possible influence of the policy experiment of the

roperty tax in Shanghai and Chongqing starting in 2011.

he data source has all the housing characteristics, house-

old income and expenditure, demographics information

e need to calculate the social welfare as well as the im-

act of property tax.

In the survey, the useful demographic include: how

any people are in the household, how many children are

t school, the ages of the members, the genders of the

embers, the occupations of the members and so on. With

he demographic variables, we can calculate the equivalent

cale of each household.

To calculate the household’s indirect utility, the total

xpenditure of the household is required to compute. The

wner occupied housing rent is an important composition

1 CFPS data is funded by 985 Program of Peking University and carried

ut by the Institute of Social Science Survey of Peking University.

H

C

G

N

f the total expenditure. However, it is usually hard to ob-

ain such information. In the survey, the households were

sked at what price they are willing to rent their houses

ut. We use these values to infer as their owner occupied

ental price. In our data, there are 169 households who ei-

her do not report their expenditures or report unreason-

ble values (less than $160 a year), 49 households who

ither do not report their own occupied residential rents

r report unreasonable values (less than $100 a year). We

liminate these missing or mistake samples and 6609 sam-

les are left as effective samples.

In the data, we can also get some information about the

ouseholds’ housing characteristics and ownership condi-

ion. For instance, the households were asked whether

hey own the houses they are living in, how large are

heir houses, how much is the house valued in the market.

hey were also asked whether they have other houses be-

ides the one they are living in. If the answer is yes, then

hey were asked again the same question about the sec-

nd house, what is the current market value, total areas

nd etc. The summary statistics of key variables used in

ur following simulation is given in Table 3.

In our data, there are 25 provinces, which were grouped

nto 7 regions. 2 The reason we group the data into 7 re-

ions is that there are just less than 300 samples in one

rovince on average, which is not large enough. The re-

ions are divided by economic and geographic conditions.

e regard Shanghai as an independent region rather than

utting it in Eastern Coastal areas, because Shanghai has

ore than 10 0 0 samples (15.6%) and it does have some

ifferences with the other coastal provinces.

All the parameters ( Table 4 ) in the translog form indi-

ect utility function are estimated in our previous paper by

ao et al. (2015) using the China Urban Household Survey

ata from 1992–2009.

There are three important matrix of the parameters in

he simulation, Ap, Bap and Bpp. Ap is a constant vector in

he simulation which satisfies i ′ •Ap = −1. Bap is a matrix

ith 11 demographic dummy variables, with which we can

enan, Hubei, Hunan, Anhui, Jiangxi; Southwest China : Sichuan,

hongqing, Guangxi, Guizhou, Yunnan; Northwest China : Shaanxi,

ansu; Coastal Areas : Shandong, Jiangsu, Zhejiang, Fujian, Guangdong;

orth China : Beijing, Tianjin, Hebei; Shanghai : Shanghai.

134 J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142

Table 4

Parameters in the simulation.

Matrix Variables Sectors

Food Consumer good Consumer service Residential cost

Ap ′ Constant −0 .3619 −0 .2189 −0 .1870 −0 .2323

Bap ′ Age 35–55 0 .0025 0 .0193 −0 .0127 −0 .0091

Age 55 + 0 .0011 0 .0210 −0 .0115 −0 .0106

Male −0 .0201 0 .0122 0 .0077 0 .0 0 02

Public 0 .0052 −0 .0169 −0 .0028 0 .0145

High school 0 .0159 −0 .0099 −0 .0199 0 .0140

College 0 .0522 −0 .0365 −0 .0479 0 .0321

Has aged −0 .0146 0 .0226 0 .0033 −0 .0112

Size 3 0 .0034 −0 .0149 −0 .0154 0 .0268

Size 4 + 0 .0101 −0 .0101 −0 .0171 0 .0171

East 0 .0283 0 .0251 0 .0015 −0 .0549

Middle −0 .0248 0 .0026 0 .0233 −0 .0011

Bpp Food −0 .0448 −0 .0418 −0 .1066 0 .1907

Consumer good −0 .0418 −0 .1157 0 .2302 −0 .0718

Consumer service −0 .1066 0 .2302 −0 .1082 −0 .0095

Residential cost 0 .1907 −0 .0718 −0 .0095 −0 .1139

Table 5

A proposal of property tax policy.

Tax base First house Free

Second house For the first two houses, each household

can have total area of 80/m

2 ∗(#member)

free of charge. The left parts are in

charge.

Third or more

houses

All the areas are taxable

Tax rate 1%/year a

a In our simulation, we also tried other rates, like 0.5% and 1.2%. We

found that the overall patterns are similar, the changes of the tax rates

wonld not affect the tax base; it will affect the magnitude levels that

property tax would pay.

g

s

v

c

d

i

h

t

5

C

f

t

c

(

a

t

τ

et the equivalent scale m 0 ( p, A k ). Bap shows how the con-

umption behavior changes according to the demographic

ariables. The demographic variables we choose here in-

ludes characters of the head of the household-age, gen-

er, occupation and education; whether has aged person

n the household; size of the household and location of the

ousehold. Bap satisfies i ′ •Bap = 0. Bpp is the price elas-

icity matrix, which satisfies Bpp

′ = Bpp and i ′ •Bpp

•i = 0.

.2. Microsimulation: the impact of property tax reform in

hina

According to the discussions on the property tax re-

orm, we conduct a microsimulation to examine the po-

ential impacts of property tax reform. Our proposal con-

entrated on most discussed proposals discussed recently

Table 5 ).

If a household with N members have K houses, whose

reas and values are M 1 , M 2 ,…, M K and V 1 , V 2 , …, V K . Then

he payment of property tax is:

T = τ∗[

K ∑

i =3

V i +

V 2

M 2

· ( M 2 − max { 0 , Cov er ∗ N − M 1 } ) ]

,

= 0 . 01 , Cover = 80 (5.1)

In this base case, because the first house owned by the

household is free of charge, naturally, quite a lot of the

households do not need to pay the tax.

As we can see from Fig. 1 , in most provinces, less than

10% of the households are involved in the property tax. In

Heilongjiang and Jilin, there are less than 3% households

needing to pay the property tax, while in coastal provinces,

more than 9% households are involved. The households are

richer in the coastal provinces and they will own more

houses in average. There are more households involved in

the property tax if more households in the province tend

to own more than one houses.

Another concern about the property tax is whether the

tax is progressive or regressive. Here, we use Suits Index to

compute and compare.

Suits Index is similar to the Gini Index, which is used to

show the distribution of tax burden. Like Gini Index, all the

households are sorted by income per capita from poorest

to richest. The accumulated percent income y is measured

on the horizontal axis, while the accumulated percent tax

burden for a tax policy x T x ( y ) is measured on the vertical

axis. Then Suits Index S x is calculated by the area under

the Lorenz curve:

S x = 1 − 2 ∗∫ 1

0

T x (y ) dy (5.2)

If the Suits Index is positive, it means the richer surfer

more tax burden, so the tax is progressive. Oppositely, if

the Suits Index is negative, it means the richer surfer less

tax burden, so the tax is regressive. From Fig. 2 , we can see

the result varies substantially across the country. It means

that property tax may have different distributional impacts

in different regions ( Fig. 3 ).

From the distribution, the richest 20% households suf-

fer from the most of the tax burden, and the richer house-

holds suffer with more tax burden. It seems the property

tax is progressive, because the richer have more houses

and tend to pay more tax. However, there are two char-

acteristics in the data need to be taken into considera-

tion. First, the ratio of the tax burden to the expendi-

ture of a household falls as the household becomes richer.

J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142 135

Fig. 1. Coverage map of affected households.

S

t

t

s

h

t

h

h

r

T

i

a

s

g

h

c

t

i

p

c

w

c

c

c

a

i

o

l

c

c

o

c

h

o

i

r

s

e

h

c

t

m

i

t

t

b

c

t

econd, in some regions like the Southwest and Shanghai,

he low income households may suffer more tax burden

han the middle income households. This may be not so

ignificant in the Southwest or in the North, but in Shang-

ai we see a significant increase of the tax burden on

he second lowest income households. It means some low

ouseholds may own more houses than the middle income

ouseholds, which is not consistent with the intuition that

icher households are more likely to own more houses.

his may due to the effects of Housing Demolition program

n China, that is, the old houses in the suburban areas

re demolished and the new buildings are constructed in-

tead. Usually, the households involved in the program can

et at least two houses for compensation. Because those

ouseholds live in the suburban areas initially, their in-

omes are usually below the average. The data really shows

he influence of this program. From Fig. 4 A, we can see

n all the regions, the average number of houses owned

er households approximately increases as the income in-

reases. However, when we just consider the households

ith two or more than two houses in Fig. 4 B, the low in-

ome households own more houses than the middle in-

ome households in average. Housing Demolition program

an explain this change, as the middle income households

re more likely to buy their second houses while the low

ncome ones are more likely to make it by the program.

Though the program happens almost in all the regions,

nly in some regions like Shanghai and the Southwest the

ow income households suffer more than the middle in-

ome households. We think this may be related with some

haracters like the household sizes and the housing areas

wned per capita.

From Fig. 5 , we see how the housing areas per capita

hange across different income groups and regions. Shang-

ai is quite different from the other 6 regions. All the

ther 5 regions except the North and Shanghai, the hous-

ng areas per capita increase as the households become

icher, mainly because the less income households own

maller houses and have more household members in av-

rage. Given waived area is based on 80 m

2 per capita, less

ousing areas per capita makes it easier for the low in-

ome households to escape from the tax. Unlike the North,

he low income households in Shanghai have about 8 m

2

ore areas per capita, which makes it harder to avoid the

nvolvement of the property tax.

Because of these heterogeneities, the impact of property

ax on the social welfare is quite complicated.

Now we try to simulate the impact of property tax on

he welfare. As mentioned in the model, the effects can

e decomposed into four parts. To make the four effects

onsistent with the translog welfare function, we regard all

he four effects as changes of the expenditures.

136 J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142

Fig. 2. The distribution of suits index of property tax reform.

0.00%

10.00%

20.00%

30.00%

40.00%

50.00%

60.00%

70.00%

80.00%

Northeast Centre Southwest Northwest Coastal North Shanghai

0-20% 20%-40% 40%-60% 60%-80% 80%-100%

Fig. 3. Distribution of property tax burden across income group and regions.

d

w

f

r

h

For the out of pocket effect, the decrease of the expen-

iture of the household equals the tax they need to pay,

hich is determined by how property tax is implemented.

For the price effect, we just consider the short-run ef-

ect. It means that we do not consider the change of the

esidential choice, when the property value decreases, the

ouseholds spend less on the residence, so that they can

spend more on the other goods. In this way, the price ef-

fect can be computed by the increase of the expenditure.

The price effect is related with how much the property

value will fall with property tax. We do not have an exact

estimation. In the paper by Bai et al. (2014 ), they estimated

that the property tax experiment lowers the property value

in Shanghai for 15%. Due to lack of alternative estimation,

J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142 137

1

1.05

1.1

1.15

1.2

1.25

1.3

1.35

1.4

Northeast Centre Southwest Northwest Coastal North Shanghai

0-20% 20%-40% 40-60% 60%-80% 80%-100%

2.00

2.10

2.20

2.30

2.40

2.50

2.60

Northeast Centre Southwest Northwest Coastal North Shanghai

0-20% 20%-40% 40-60% 60%-80% 80%-100%

A

B

Fig. 4. (A) Number of houses on average for the whole sample. (B) Number of houses in average for the restricted sample.

20

40

60

80

100

120

140

Northeast Centre Southwest Northwest Coastal North Shanghai

0-20% 20%-40% 40-60% 60%-80% 80%-100%

Fig. 5. Housing areas owned per capita.

138 J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142

Table 6

Scenarios of utilization property tax revenues.

Policy Qualification

No subsidy No household can get subsidy

Education subsidy Each child aged among 7 and 16 can be

subsidized

Housing subsidy A household owning no house or owning a

house with areas per capita less than 5 m

2

Double subsidy All the households that are involved in

education subsidy and housing subsidy

policy are qualified

w

h

h

p

t

s

s

t

l

M

b

h

p

p

a

e

a

t

e

h

i

l

t

t

t

t

c

s

h

d

c

s

e

l

t

h

i

t

s

c

y

s

m

e

e will try different values between 0% and 15% and see

ow the price effect will be under different circumstances.

For the housing wealth effect, it not only depends on

ow much the property value will change, but also de-

ends on the marginal propensity to consume with respect

o house wealth. In the U.S., many empirical studies have

hown that the marginal propensity to consume with re-

pect to house wealth is between 0.05 and 0.07. However,

he case is not the same in China. In China, house is more

ike an investment good rather than a consumption good.

any households in China do not own a house and plan to

uy one in the future. If the property value increases, the

ouseholds may need to save more, therefore the marginal

ropensity to consume in China may be negative. In this

aper, we will try different values between −0.06 and 0.06

nd see how the house wealth effect will be under differ-

nt circumstances.

For redistribution effect, the property tax revenue is

ssumed to be used in redistribution. The effect means

hat if the government uses the income from the prop-

rty tax to provide more local public good, some house-

olds may benefit from such programs. The redistribution

s very complicated due to the demand characteristics of

ocal public goods. We assume our microsample the dis-

ribution of such demand characteristics will not vary in

he short and medium term with respect to the property

ax policy, so we can use our micro-sample to simulate for

his distributional effect of benefits. For example, if the lo-

al government uses the income from the property tax to

ubsidize education, we can regard it as a subsidy to the

ouseholds with children at school.

Before we decompose the four effects, we now intro-

uce four possible revenue utilization regimes that are dis-

ussed the most in China ( Table 6 ).

The first revenue utilization policy is to assume that no

ubsidy is given on any households. For instance, the rev-

nue can be used to pay back the previously cumulative

ocal government debt. We use this case as the benchmark

o compare redistribution policies. Under this policy, the

ouseholds cannot get any subsidy, so the welfare for any

ncome groups is negative.

The second policy scenario is to assume the property

ax is used to enhance local public school system, so the

ubsidy is implicitly given to the households who have

hildren aged among 7 and 16. In China, the mandatory

ears of schooling are nine years, and most children enter

chool at 7. The policy is similar to that in the U.S., where

ost of the property tax revenue is spent on local public

ducation system.

The third policy scenario is to assume that the property

tax revenue can be used to subsidize housing for the poor

people. At current standard, only people who do not own

any houses or own a house with areas per capita less than

5m

2 are eligible to apply for the public renting house. In

the pilot experiment of property tax in Shanghai, the ma-

jor part of the property tax revenue has been used to sub-

sidize public renting house program.

The fourth policy scenario is to assume that the prop-

erty tax is used to both enhance public school system and

subsidize housing for the poor people. This policy com-

bines the second and third policies, so eligible households

may benefit from both policies. The government can have

different choices on spending how much of the tax rev-

enue on these two systems. In the simulation, three sub

scenarios have been examined (the public school system

will get 40%, 50% and 60% of the total tax revenue respec-

tively; and vice versa for subsidizing housing for the poor

at 60%, 50%, and 40%).

In our simulation, we apply the provincial jurisdiction

as the border line for each region to redistribute the prop-

erty tax revenue through these benefit programs. Based

on these assumptions, we can use our micro sample to

identify how many households are affected by the specific

policies. For instance, under the second policy scenario,

the difference of the ratio of households with children

aged among 7 and 16 is not so large among the regions,

in general between 15% and 25%. However, for the third

policy, the affected population varies substantially across

provinces ( Fig. 6 ).

There are two main factors may affect the eligibility

number of households to apply for public renting house.

First, if the average income in one province is too low,

then the ratio to apply for Public Rental Program is rela-

tively high. There are six provinces can be explained this

way: Heilongjiang, Jilin, Shanxi, Shaanxi, Gansu and Yun-

nan. Second, if the housing price is too high; so many

households cannot afford a house so more people would

choose renting instead, such as Beijing and Shanghai. For

the provinces whose average income is higher enough and

the housing price is not so high as in Beijing and Shang-

hai, for example the coastal provinces like Zhejiang and

Jiangsu, the ratio is lower significantly. With our micro

simulation exercise, we can see how the social welfare will

change under various redistribution policies ( Table 7 ).

We can see from Table 7 , in the benchmark case of no

subsidy at all, the social welfare would not change a lot,

because quite a lot of households are not influenced by

the property tax. The social welfare decreases under the

policy with no subsidy, and the welfare in Shanghai and

the coastal area decreases the most. This is because more

households own more houses in Beijing and Shanghai, so

the tax burden has a negative effect on the social welfare

where benefit program does not exist yet. For the other

two scenarios, we can see that for both the education

subsidy policy and the housing subsidy policy, the social

welfare would increase, and the housing subsidy policy

always deliver better performance. In general, whether a

household has a child aged between 7 and 16 is weakly

correlated with the income of the household; while a

household with lower income is more likely to own no

J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142 139

Fig. 6. The distribution of population eligible to apply for public rental program in China.

Table 7

Effects of four property tax utilization policies.

Region No subsidy (%) Education

subsidy (%)

Housing subsidy

(%)

Double subsidy

Edu:40%

Double subsidy

Edu:50%

Double subsidy

Edu:60%

Northeast −0.0126 0.0183 0.0216 0.0206 0.0202 0.0199

Centre −0.0583 0.0361 0.0809 0.0663 0.0619 0.0574

Southwest −0.0592 0.0272 0.0289 0.0303 0.0306 0.0306

Northwest −0.0274 0.0359 0.0402 0.0392 0.0388 0.0384

Coastal −0.1057 0.0692 0.1087 0.1034 0.0997 0.0954

North −0.0576 0.0397 0.0553 0.0514 0.0498 0.0481

Shanghai −0.1581 0.0187 0.1324 0.1011 0.0902 0.0782

h

e

t

l

r

l

t

c

N

s

h

h

s

N

b

h

a

T

l

t

w

t

t

s

r

i

f

d

t

p

ouse or own a house with small areas. Intuitively, the

ducation subsidy policy redistribute the revenue among

he households without children but not directly target

ow-income group; while the housing subsidy policy can

edistribute the revenue directly to the households with

ow income. In result, the latter mitigate more inequality

han alternative regimes. And we can see that the edu-

ation subsidy policy has weaker effects in Shanghai and

ortheast. It may be due to the correlation between the

tudent numbers of a household and the income of a

ousehold ( Table 8 ). Though in all areas it seems that a

ousehold with higher income is more likely to have fewer

tudents, the correlation is much weaker in Shanghai and

ortheast. The second reason may be relevant to the num-

er of students per household. The student number per

ousehold in Shanghai is much lower than other places,

nd Northeast is next lower among the rest six regions.

he lower number of students per household means that

ess proportion of households are involved in the educa-

ion subsidy policy, which makes the effect of the policy

eaker. The double subsidy policy performs better than

he education subsidy policy alone, but still worse than

he housing subsidy policy, as expected. Among the three

ub scenarios of double subsidy policies, the more tax

evenue are spent on the public school system, the less

ncrease on the social welfare.

In all the scenarios, the housing subsidy policy per-

orms better than the education subsidy policy and the

ouble subsidy policy for all the regions. So we just need

o focus on the housing subsidy policy.

Now we can decompose the four effects under the

roperty tax policy with housing subsidy.

140 J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142

Table 8

Statistics about the number of students.

Region Correlation between number of

students in a household and

income per capita

Number of students

per household

Northeast −0.0835 0.1912

Centre −0.0997 0.2778

Southwest −0.1419 0.3139

Northwest −0.1056 0.2725

Coastal −0.1039 0.2925

North −0.1129 0.2554

Shanghai −0.0751 0.1049

t

r

r

d

c

i

w

p

f

i

a

r

t

i

T

A

Besides the benchmark tax rate of 1%, we also tried al-

ernative rates: 0.5% and 1.2%. This range reflects the cur-

ent discussed upper and lower bound of the property tax

ate range. From Table 9 , we can see the four effects un-

er the three tax rate scenarios have similar patterns. The

hanges of the tax rates would not affect the tax base,

t will only affect the magnitude levels that property tax

ould pay, as well as the overall welfare. Compared to

rice effect and housing wealth effect, out of pocket ef-

ect and redistribution effect is much smaller. The reason

s that every household is influenced by the price effect

nd house wealth effect, while only a few are involved in

edistribution and payment of property tax. We can see

hat the effects differ a lot among the regions. The hous-

ng wealth effect dominates the other three effects. If the

able 9

. Decomposition of four effects of property tax with tax rate of 1%.

Region Out of

pocket

effect

Price effect Housing wealth e

min (pp =

0)

mean c max (pp =

15%)

min (pp =

15%, ee =

0.06)

mea

Northeast −0.013% 0% 0.138% 0.275% −0.672% −0.1

Centre −0.062% 0% 0.177% 0.352% −1.572% −0.6

Southwest −0.065% 0% 0.240% 0.477% −1.456% −0.7

Northwest −0.029% 0% 0.184% 0.365% −0.911% −0.2

Coastal −0.110% 0% 0.203% 0.404% −1.830% −0.9

North −0.059% 0% 0.222% 0.441% −1.400% −0.5

Shanghai −0.173% 0% 0.258% 0.511% −2.395% −1.3

(B) Decomposition of four effects of property tax with tax rate of 0.5%.

Northeast −0.005% 0% 0.138% 0.275% −0.672% −0.0

Centre −0.027% 0% 0.177% 0.352% −1.572% −0.0

Southwest −0.026% 0% 0.240% 0.477% −1.456% −0.0

Northwest −0.013% 0% 0.184% 0.365% −0.911% −0.0

Coastal −0.048% 0% 0.203% 0.404% −1.830% −0.0

North −0.028% 0% 0.222% 0.441% −1.400% −0.0

Shanghai −0.071% 0% 0.258% 0.511% −2.386% −0.1

(C) Decomposition of four effects of property tax with tax rate of 1.2%.

Northeast −0.016% 0% 0.138% 0.275% −0.672% −0.0

Centre −0.074% 0% 0.177% 0.352% −1.572% −0.0

Southwest −0.076% 0% 0.240% 0.477% −1.456% −0.0

Northwest −0.034% 0% 0.184% 0.365% −0.911% −0.0

Coastal −0.135% 0% 0.203% 0.404% −1.830% −0.0

North −0.070% 0% 0.222% 0.441% −1.400% −0.0

Shanghai −0.224% 0% 0.258% 0.511% −2.395% −0.1

a pp means the pencentage of price falling with respect to property tax. b ee means the marginal propensity to consume of housing wealth. c The mean is the average of the values with all possible parameters of pp andd We use the housing subsidy policy for the redistribution effect. e This case is based on Yang’s result.

propensity to consume is positive like in the U.S., then

property tax may lower the welfare; otherwise, the welfare

may increase due to the increase of consumption as there

is no need to save so much with the falling of property

value. To estimate the propensity to consume in China is

very important to compute the impact of property tax on

the welfare, which will be involved in our further work.

Now naturally, a question occurs, if China is going to

implement the property tax, should all the country-side

impose the uniform property tax rate or let the local gov-

ernment decide the rate within its jurisdiction? If so, what

is the optimal policy cross regions?

As discussed above, price effect and house wealth effect

are not influenced by how the property tax is implemented

and how the revenue is redistributed. And we use housing

subsidy policy to be the redistribution policy.

To address these issues, we compare two alternative

policies for local governments to opt in to choose their op-

timal policy:

(1) All the housing areas are taxed, similar to the US

case.

(2) The first house is free of charge, and there are some

areas free of charge for the first two houses, the left

areas are taxed, similar to the Shanghai case.

Under the first policies, the increase of tax rate has two

different effects. On one hand, as the tax rate increases, all

the households need to pay more tax, which will lower the

ffect a , b Redistribution

effect d Whole effect

n c max (pp =

15%, ee =

−0.06) e

min (pp =

15%, ee =

0.06)

mean c max (pp =

15%, ee =

−0.06)

72% 0.533% 0.023% −0.387% 0.144% 0.818%

56% 1.017% 0.084% −1.199% 0.198% 1.390%

12% 1.120% 0.021% −1.023% 0.214% 1.552%

67% 0.769% 0.037% −0.538% 0.196% 1.142%

04% 1.191% 0.111% −1.425% 0.219% 1.596%

23% 1.007% 0.056% −0.962% 0.225% 1.445%

95% 1.394% 0.122% −1.935% 0.272% 1.854%

17% 0.533% 0.017% −0.369% 0.117% 0.787%

66% 1.017% 0.073% −1.124% 0.093% 1.314%

72% 1.120% 0.045% −0.907% 0.156% 1.535%

27% 0.769% 0.034% −0.524% 0.144% 1.091%

87% 1.191% 0.118% −1.406% 0.073% 1.515%

50% 1.007% 0.059% −0.912% 0.149% 1.384%

29% 1.395% 0.160% −1.831% 0.063% 1.800%

17% 0.533% 0.041% −0.389% 0.107% 0.781%

66% 1.017% 0.164% −1.253% 0.042% 1.286%

72% 1.120% 0.102% −0.860% 0.117% 1.511%

27% 0.769% 0.081% −0.557% 0.122% 1.078%

87% 1.191% 0.248% −1.618% −0.026% 1.464%

50% 1.007% 0.133% −1.028% 0.101% 1.354%

28% 1.394% 0.342% −1.948% −0.063% 1.740%

ee in the range.

J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142 141

0.00%

0.02%

0.04%

0.06%

0.08%

0.10%

0.12%

0.14%

0.16%

0.18%

0.20%

0 10 20 30 40 50 60 70 80

Northeast

Center

Southwest

Northwest

Coastal

North

Shanghai

Fig. 7. The welfare change with waived areas for different regions.

w

d

t

e

a

t

c

c

t

t

m

i

m

t

m

t

p

A

h

m

m

c

p

a

s

b

r

v

w

i

m

f

d

v

c

t

t

s

Table 10

Suggested policy under two cases.

Region U.S. case Shanghai case a

Tax rate with

no waived

areas (%)

Welfare

change

(%)

Extra waived

areas per capita

(m

2 )

Welfare

change

(%)

Northeast 0.4 0.058 0 b 0.063

Center 0.2 0.099 10 b 0.158

Southwest 0.1 0.011 70 b 0.037

Northwest 0.5 0.025 20 b 0.064

Coastal 0.1 0.063 20 b 0.142

North 0.2 0.061 50 b 0.068

Shanghai 0.3 0.073 30 b 0.190

a In Shanghai Case, the tax rate is set to be 1%. b The suggested policy.

6

h

q

t

w

p

t

w

m

p

p

w

c

p

i

m

f

f

U

e

t

elfare; on the other hand, more tax revenues can be re-

istributed from the rich to the poor, which will increase

he welfare.

Under the second policies, both the tax rate and the ar-

as free of charge can change. We first argue that at given

reas free of charge, the lower the tax rate is, the lower

he welfare will be. Given areas free of charge, the tax base

an be determined. And because the first house is free of

harge, most of the low income households do not need

o pay the tax. Then the first effect we mentioned above

hat the welfare will decrease since more tax to pay seems

uch weaker. It’s obvious that the first effect will be dom-

nated by the second effect. The decrease of the tax rate

akes less revenue to be redistributed so that it will lower

he welfare. But the tax rate is not the higher the better as

entioned, so we choose 1% as a reasonable tax rate for

he second policies.

When we consider the change of the areas waived for

roperty tax charges, there are also two different effects.

s the areas waived decreases, more low income house-

olds will be involved in the tax. The poor needs to pay

ore tax, which lowers the welfare. On the other hand,

ore revenue can be redistributed as the areas free of

harge decreases, which will increase the welfare.

Now we can use this framework to simulate different

olicies and compare. Here we focus on some policies that

re relatively political feasible (for instance the tax rate

hould be smaller than 1% to avoid political resistance and

ig shock to the economy), but still achieve satisfactory

esults than other policies. Fig. 7 shows how the welfare

aries with the change of waived areas in each region. So

e can get some rough idea of the optimal waived areas

n Shanghai case. Table 10 gives a summary of such ad-

issible policies after comparing their corresponding wel-

are changes. We can see that these admissible policies are

ifferent among the regions and the welfare changes also

ary a lot among the regions. In all regions, the Shanghai

ase policy performs better than the U.S. case policy. So in

he further implementation of the property tax, how to set

he waived areas is important to consider in the policy de-

ign.

. Conclusion

Despite many years of discussion whether or not, and

ow to impose a property tax reform in China, very little

uantitative study has been conducted on the analysis of

he impacts of potential property tax reform in China. So

e still have little understanding of the effects if China im-

ose the property tax, and what would be the incidence on

he households, that is, who will benefit and who will get

orse off due to these reforms. Since China’s real estate

arket are so different across the whole nation, the appro-

riate policy is obviously would be quite different across

rovinces.

To shed some lights on these questions, in this paper

e used the China Family Panel Survey (CFPS) data to

onduct a microsimulation model to examine possible im-

acts and incidences of alternative property tax regimes

n China. We use the cross-sectional information in the

icro-data to decompose and simulate the potential ef-

ects of property tax policies, and found the impacts dif-

er quite significantly across regions and across households.

sing the Slesnick–Jorgenson to combine both criteria of

fficiency and equity, we simulated for various property

ax scenarios for each provinces in our sample, our results

142 J. Cao, W. Hu / Journal of Housing Economics 33 (2016) 128–142

s

t

t

g

p

b

t

a

g

w

p

s

i

t

h

b

i

t

b

o

t

m

m

f

s

i

r

d

r

l

e

i

m

t

t

g

w

s

t

w

s

m

i

w

o

fl

e

t

t

b

e

t

c

R

B

uggest that property tax, as a local tax, would achieve bet-

er performance if the policy design is catered to match

he unique distribution of the housing ownership, income

roup and consumption distributions in each area.

With the microsimulation focused on the incidence of

roperty tax reform, we found the tax is more likely to

e progressive in many cases. Redistributing the property

ax revenue would somewhat mitigate the regressivity,

nd the direct subsidy on the poor’s public housing pro-

ram usually achieve better performance in terms of social

elfare.

Our study has provided some crude estimates on the

otential impacts of property tax reform in China for the

hort-run and medium run. It seems that the property tax

n Shanghai case performs better than the U.S. case. In

he future property tax reform, it is very likely the first

ouse owned by a household or certain floor area would

e waived from taxation. Considering most households live

n an apartment typically between 80 and 120 square me-

ers, for a typical household with three household mem-

ers, a waived area of 30–40 per capita would lead to the

utcome that most of the households in China do not need

o suffer the tax burden. If the households have two or

ore houses or apartments, then the waived area would

atter quite a lot. In our experiments and simulations, we

ound out to maximize social welfare, it is important to

et different waived area per capita at local level consider-

ng huge heterogeneity across whole China. In addition, the

evenues from the property tax should be used to subsi-

ize housing for the poor people. Similarly, the optimal tax

ates are also supposed to be different in different regions,

ess than 1% seems to be more acceptable. Many potential

ffects of the property tax remain unknown, nevertheless

t is even harder to implement the policy in practice. To

ake it easier to be accepted, at the beginning, property

ax can be imposed on the property owners having more

han two or three houses or apartments at lower rate, then

radually expand the coverage and raise the tax rate, along

ith better local public good provision.

In the future when the panel data or repeated cross-

ection data are available with both household income dis-

ribution, housing information, housing price changes as

ell as detail housing characteristics, we can extend our

tatic model and cross-sectional simulation to a dynamic

icrosimulation, then we may also assess the longer term

mpacts of the property tax policy on housing market, as

ell as household behaviors and social welfare. Finally, in

ur simple model here we did not allow for the capital

ow and labor migration in our model. According to lit-

rature, property tax differences across regions would po-

entially lead to factors move around, so this would lead

o a future study when more detail capital flow and la-

or migration information is available from the previous

mpirical evidence, then we can improve current studies

o better understand how housing prices and property tax

hanges lead to these changes.

eference

ai, C , Li, Q. , Ouyang, M. , 2014. Property taxes and home prices: a tale oftwo cities. J. Econom. 180 (1), 1–15 .

Bourguignon, F. , Spadaro, A. , 2006. Microsimulation as a tool for evaluat-

ing redistribution policies. J. Econ. Inequal. (4) 77–106 .

Brownstone, D. , Englund, P. , Persson, M. A Microsimulation Approach”,1985. “Effects of the Swedish 1983–85 tax reform on the demand for

owner-occupied housing. Scand. J. Econ. 87 (4), 625–646 . Cao, J., Ho, M., and Liang, H. “Estimating Household Consumption Func-

tion in Urban China,” CCWE working paper, 2015. Case, K.E. , Quigley, J.M. , Shiller, R.J. , 2005. Comparing wealth effects: the

stock market versus the housing market. Adv. Macroecon. 5 (1), 1–

32 . Du, L. , Pan, C.Y. , Zhang, S.Y. , Cai, J.N. , 2010. “How do house prices af-

fect consumption: promote or suppress? empirical evidence from172 prefecture-level cities in China”. Zhejiang Soc. Sci. (8) 24–

30 . Fischel, W.A. , 1974. Fiscal and environmental considerations in the lo-

cation of firms in suburban communities a nontechnical digest. In:

Proceedings of the Annual Conference on Taxation Held under theAuspices of the National Tax Association-Tax Institute of America, 67,

pp. 632–656 . Haberler, G. , 1958. Prosperity and Depression. Harvard University Press .

Hamilton, B.W. , 1975. Zoning and property taxation in a system of localgovernments. Urban Stud. 12, 205–211 .

Hamilton, BruceW. , 1976. Capitalization of intrajurisdictional differences

in local tax prices. Am. Econ. Rev. 66, 743–753 . Hamilton, BruceW. , 1983. A review: is the property tax a benefit tax?",

local provision of public services. The Tiebout Model after Twenty-fiveYears. Academic Press, pp. 85–107 .

Immervoll, H. , Kleven, HenrikJ. , Kreiner, ClausT. , Saez, E. , 2007. Welfare re-form in European countries: a microsimulation analysis. Econ. J. 117,

1–44 . Jorgenson, D.W. , Slesnick, D.T. , 1983. Individual and social cost of living

indexes. In: Diewert, W.E., Montmarquette, C. (Eds.), Price Level Mea-

surement. Ottawa, Canada, pp. 241–323 Statistics . Jorgenson, D.W. , Slesnick, D.T. , 1985. Efficiency versus equality in

petroleum taxation. Energy J. 6, 171–188 . Jorgenson, D.W. , Slesnick, D.T. , 1985. Efficiency versus equality in natural

gas price regulation. J. Econom. 30, 301–316 . Kuang, W.D. , Zhu, Y. , Liu, J.T. , 2012. The impact of property tax upon

housing prices: evidence from OECD countries. Financ. Trade Econ. (5)

121–129 . Iacoviello, M., “Housing Wealth and Consumption”, FRB International Fi-

nance Discussion Paper, 2011 Ludwig, A., Slok, T., 2002, “The Impact of Changes in Stock Prices and

House Prices on Consumption in OECD Countries”, IMF Working Pa-per, issue. 1, 2002, pp. 1–36.

Mieszkowski, P. , 1972. The property tax. An excise tax or a profits tax? J.

Public Econ. 1, 73–96 . Muellbauer, J. , 2007. Housing, credit and consumer expenditure. In:

Proceedings Economic Policy Symposium – Jackson Hole, pp. 267–334 .

Muellbauer, J. , Murphy, A. , 2008. Housing markets and the economy: theassessment. Oxf. Rev. Econ. Policy 24 (1), 1–33 .

Netzer, D. , 1966. Economics of the Property Tax. Brookings Institution,

Washington DC . Oates, W.E. , 1969. The effects of property taxes and local public spending

on property values: an empirical study of tax capitalization and thetiebout hypothesis. J. Political Econ. 77 (6), 957–971 .

Palmon, D. , Smith, B.A. , 1998. New evidence on property tax capitaliza-tion. J. Political Econ. 106 (5), 1099–1111 .

Simon, HerbertA. , 1943. The incidence of a tax on urban real property. Q.

J. Econ. 59, 398–420 . Tiebout, C.M. , 1956. “A pure theory of local expenditures. J. Political Econ.

64 (5), 416–424 . White, M.J. , 1975. Firm location in a zoned metropolitan area. In: Edwin

S., Mills, Wallace E., Oates (Eds.), Fiscal Zoning and Land Use Controls.Lexington Books, Lexington, MA, pp. 175–202 .

Xie, Y. , Lu, P. , 2015. The sampling design of the china family panel studies(CFPS). Chin. J. Sociol. 1 (4), 471–484 .

Yang Z., Zhang H., Zhao L.Q., “Dual Role of Housing Consumption and In-

vestment: Reexamine Correlation of Housing and Consumption in Ur-ban China”, Working paper, 2014.

Yinger, John , Bloom, HowardS. , Boersch-Supan, Axel , Ladd, HelenF. , 1988.Property taxes and house values. The Theory and Estimation of Intra-

jurisdictional Property Tax Capitalization. Academic Press, San Diego,CA .

Zodrow, G.M. , Mieszkowski, P. , 1986. Pigou, Tiebout, property taxation,

and the underprovision of local public goods. J. Urban Econ. 19,356–370 .