Utpal Bhattacharya, Hong Kong University of Science and Technology Daisy Huang, Nanjing Audit University Kasper Meisner Nielsen, Hong Kong University of Science and Technology

1. Haunted Houses
Definition of Haunted Houses from Wikipedia:
“A haunted house is a house or other building often perceived
as being inhabited by disembodied spirits of the deceased
who may have been former residents or were familiar with the
property.”
Spillovers in Asset Prices: The Curious Case
of Haunted Houses
Utpal Bhattacharya, Hong Kong University of Science and Technology
Daisy Huang, Nanjing Audit University
Kasper Meisner Nielsen, Hong Kong University of Science and Technology

2. Research Questions
1) When a house becomes haunted, does its price drop? If
yes, by how much?
2) Does the haunted house affect the price of neighboring
houses? If yes, by how much?
3) If there is a negative spillover effect, why does that
happen? Fire sales? Or something else?

3. Why Are These Research
Questions Important to Answer?
Externalities—the notion that the consequences of an
economic activity can affect third parties—has been central to
the neo-classical critique of market organization.
While the theoretical modelling of externalities has made
significant advancements, the same cannot be said about the
empirical literature.
Empirically it is difficult to convincingly conclude that one
economic agent has an effect on an unrelated third party
without completely ruling out that shocks to the local economy
are affecting both.
This is especially true for housing.

4. Why Hong Kong?
1) Homogenous apartments in a block in an estate
all densely packed together – spillovers are easy
to identify
Details:
Hong Kong’s population of 7.3 million lives in a small area that is less than
25% of its 1,106 km2 land
Residential real estate mainly consists of units in high rise apartment
blocks sharing common facilities inside an estate
Market consists of many estates, but apartments inside each estate are
fairly homogeneous

5. 5
Example: Dawning Views
Fanling

6. 6
Example: Dawning Views

7. Why Hong Kong?
2) Locals take haunted houses very seriously. Sellers have to
disclose.

8. Why Hong Kong?
3) There are many new haunted houses every year: large
sample size

9. Why Hong Kong?
4) Unnatural death is an idiosyncratic negative shock that is
unlikely to be correlated with local demand or supply
economic shock affecting real estate prices.
Even if it is correlated, within an estate, it is unlikely that
local market conditions differentially affect haunted and
non-haunted units / floors / blocks because of the close
proximity.

10. 10
Data on HK Residential Real Estate
 All private residential real estate transactions in Hong Kong
between 2000 and 2015
 Data include
 Estate name and location
 Property characteristics (size, age)
 Transaction date and price
 We harvest data on haunted houses from 4 real estate websites
 HK Compass; Property HK; Squarefoot; Spacious
 We focus on the 214 most liquid estates with more than 1,000
transactions (~65 per year) during our sample period
 Corresponds to 50% of all private residential real estate
transactions during our sample period

11. 11
Liquid Estates

12. 12
Descriptive Statistics (1)

13. 13
Descriptive Statistics (2)

14. 14
Effect of Haunted Houses on Prices

15. The Formal Research Design
The dependent variable is the log price, yit, of apartment i in year-month t :
′ , (1)
where captures apartment fixed effects, is a vector of year-month fixed effects, is a vector of
apartment characteristics, and is an indicator for a haunted house due to an unnatural death occurring
in the last k years before t. House characteristics that change over time in our sample are interior size and
block age.
The coefficient of interest in equation (1) is θ. A negative θ reveals the discount for being haunted, and if
negative, the value of θ reveals the percentage discount.

16. 16
Empirical Specification
 is an indicator for a haunted
house due to an unnatural death
occurring in the last k years before t
 Unit
 Floor
 Block
…
 Estate
 To avoid spurious correlation we
exclude:
 Affected unit when we estimate the
effect on prices of the affected floor
 Affected floor when we estimate the
effect on prices of the affected block
 Affected block when we estimate the
effect on prices of the affected estate

17. 17
Spillover Effects of Haunted Houses
Unit Floor Block Estate
(1) (2) (3) (4)
Haunted house -0.1970***
-0.0466***
-0.0271***
-0.0120***
(0.049) (0.0015) (0.0019) (0.0014)
Size 0.0016***
0.0014***
0.0014***
0.0014***
(0.0001) (0.0001) (0.0001) (0.0001)
Age -0.0116***
-0.0171***
-0.0186***
-0.0213***
(0.0004) (0.0003) (0.0003) (0.0006)
Unit fixed-effects Yes Yes Yes Yes
Year-month fixed effects Yes Yes Yes Yes
Adj. R-squared 0.988 0.976 0.976 0.975
N 151,237 323,586 324,025 199,516

18. -20%
-5%
-3%
-1%
Spillover Effects of Haunted Houses

19. Other Results
1)The ripple effect is strongest for murders.

20. Other Results
2) Price recovery is slow.

21. Other Results
3) The most important contribution of our paper is that, since we
observe that negative spillovers exist even if the haunted house
is not sold, we can isolate the quality channel from the fire sale
channel. We find that the quality channel contributes significantly
to spillovers in assets prices.

22. Decomposition;
Quality Effect vs Fire Sales Effect
Quantity
Price
Demand for normal house
Demand for haunted house
Supply if haunted house is not sold
Supply if haunted house is sold
Quality Effect
Forced Sale
Effect

23. External Validity: Is Hong Kong Special?
Question 1: How prevalent worldwide is the belief in ghosts and the
supernatural, or is Hong Kong unique?
Answer: Belief in the supernatural is pervasive. We have data for 16
countries. In the United States, the percentage of people who believe in
ghosts range, depending on the survey, from 37% to 45%. In the U.K., the
percentage of people who believe in ghosts range, depending on the survey,
from 40% to 52%. The University of Pennsylvania has a center for research
on ghosts: the Penn Ghost Project.
Question 2: Is Hong Kong special with regards to its aversion to haunted
houses?
Answer: A study of U.S. case law suggests that it is illegal for a seller to hide
the fact that the property being sold is haunted. The website,
www.diedinhouse.com, claims to be the first of its kind.

24. 24
Conclusion
• We exploit the institutional setting in Hong Kong’s residential real
estate market to identify the effect of spillovers on asset prices
• We find a ripple effect of haunted houses on prices, as they drop by
• 20% for affected units
• 5% for units on the affected floor
• 3% for units in the affected block
• 1% for units in the affected estate
• The ripple effect is strongest for murders; price recovery is slow
• We show that negative spillovers exist even when the haunted
house (i.e. unit) is not sold – documenting (for the first time) that
asset quality contributes significantly to spillovers in asset prices

25. CONCLUSION