# RECORD 11135c13b52c3fe849cfd9899af8e92968906ec5
# TITLE: Assessing the Impact of Wildfires on the Swedish Housing Market: a Case Study of the 2014 Västmanland Wildfire
# YEAR: 2026.0  VENUE: Economics of Disasters and Climate Change  DOI: 10.1007/s41885-026-00200-3
# AUTHORS: Tommaso Piseddu; David Stenvall
# PAGES: 27  CANDIDATE TABLE PAGES (1-based): [10, 13, 14, 16, 20, 21]
# ABSTRACT: The literature on wildfires and residential property prices is limited and primarily focuses on events in North America. There is a lack of studies examining this relationship in Europe. With the largest forest area in the entire EU, understanding this impact is particularly relevant in a Swedish context. In this paper, we investigate how the largest wildfire in Sweden’s recent history, the 2014 wildfire in Västmanland County, affected nearby housing prices and time-on-market. Using a difference-in-differences method, we find a significant negative effect on housing prices. Our most conservative estimate indicates an approximate 2.7% reduction in final selling prices for post-fire sales located within 20 km of the wildfire area. However, the negative effect is even larger when defining the treated area as a 5 km (-10.1%) or 10 km (-8.9%) distance to the fire, and these results are confirmed by a repeated sales model that includes only single- and multi-family houses. In a heterogeneity assessment with respect to the housing types, we find that our effects are mostly driven by the impact on single-family houses rather than apartments. For the former group, we find large negative and significant effects, while we fail to detect any significant one on apartments prices. The largest impact on prices occurs in the first months after the event and for those properties that are located along the trajectory of the wildfire’s smoke.


===== PAGE 1 (score 6, numeric-paren 0, table-words 0, var-words 6) =====

RESEARCH
Economics of Disasters and Climate Change (2026) 10:20
https://doi.org/10.1007/s41885-026-00200-3
Assessing the Impact of Wildfires on the Swedish Housing 
Market: a Case Study of the 2014 Västmanland Wildfire
Tommaso Piseddu1,2 · David Stenvall3
Received: 4 February 2026 / Accepted: 12 May 2026 / Published online: 25 May 2026
© The Author(s) 2026
Abstract
The literature on wildfires and residential property prices is limited and primarily focuses 
on events in North America. There is a lack of studies examining this relationship in 
Europe. With the largest forest area in the entire EU, understanding this impact is particu­
larly relevant in a Swedish context. In this paper, we investigate how the largest wildfire 
in Sweden’s recent history, the 2014 wildfire in Västmanland County, affected nearby 
housing prices and time-on-market. Using a difference-in-differences method, we find a 
significant negative effect on housing prices. Our most conservative estimate indicates 
an approximate 2.7% reduction in final selling prices for post-fire sales located within 
20  km of the wildfire area. However, the negative effect is even larger when defining 
the treated area as a 5 km (-10.1%) or 10 km (-8.9%) distance to the fire, and these re­
sults are confirmed by a repeated sales model that includes only single- and multi-family 
houses. In a heterogeneity assessment with respect to the housing types, we find that our 
effects are mostly driven by the impact on single-family houses rather than apartments. 
For the former group, we find large negative and significant effects, while we fail to detect 
any significant one on apartments prices. The largest impact on prices occurs in the first 
months after the event and for those properties that are located along the trajectory of the 
wildfire’s smoke.
Keywords  Wildfires · Housing prices · Natural disasters · Repeated sales · Sweden
MSC Codes  62J05 · 62J20
JEL Codes  C33 · Q51 · Q54 · R31
	
 Tommaso Piseddu
piseddu@kth.se; tommaso.piseddu@sei.org
1	
Department of Real Estate and Construction Management, Division of Real Estate Economics 
and Finance, Royal Institute of Technology (KTH), Stockholm, Sweden
2	
Stockholm Environment Institute, Stockholm, Sweden
3	
Department of Management and Engineering, Division of Economics, Linköping University, 
Linköping, Sweden


===== PAGE 10 (score 115, numeric-paren 83, table-words 8, var-words 7) =====

1 3
20 
Page 10 of 27
Economics of Disasters and Climate Change (2026) 10:20
In other words, the dataset is composed of single observations representing price dif­
ferences of two consecutive transactions for the same asset. The treatment group is in this 
approach composed of consecutive prices differences where the earlier transaction occurred 
before the date of the wildfire and the consecutive one occurred after it. That is, the wildfire 
occurred within the temporal bracket of two consecutive transactions.
A threshold distance is again imposed to identify the assets sufficiently close to the wild­
fire to be reasonably expected to have been affected by the event. Given that our variable of 
interest, the difference in consecutive selling prices, has been log-transformed, the depen­
dent variable of our repeat-sales approach is an approximation of the percentage variation 
of the final selling prices,
Table 2  Main results
Dependent variable:
Log(selling price)
(1)
(2)
(3)
(4)
(5)
Treatment
−0.585***
−0.648***
−0.028**
−0.049***
−0.069**
(0.019)
(0.016)
(0.011)
(0.011)
(0.032)
Post
0.278***
0.290***
0.284***
0.282***
0.513***
(0.005)
(0.004)
(0.003)
(0.003)
(0.013)
Living Area (m2)
0.005***
0.005***
0.005***
0.005***
(0.0001)
(0.0001)
(0.0001)
(0.00005)
Age
−0.004***
−0.0003***
−0.0004***
−0.002***
(0.0001)
(0.0001)
(0.0001)
(0.00005)
Age2
0.00000***
0.00000***
0.00000***
0.00000***
(0.00000)
(0.00000)
(0.00000)
(0.00000)
Number of rooms
0.076***
0.050***
0.052***
0.060***
(0.002)
(0.002)
(0.002)
(0.001)
W(Selling Price)
0.942***
0.949***
0.271***
(0.003)
(0.003)
(0.010)
Distance fire
−0.00000***
−0.00000
(0.00000)
(0.00000)
Distance forest
0.0001***
0.00002***
(0.00000)
(0.00001)
Distance train
−0.00001***
−0.00001***
(0.00000)
(0.00000)
Distance lake
−0.0001***
−0.0001***
(0.00000)
(0.00000)
Post * Treatment
−0.182***
−0.140***
−0.096***
−0.104***
−0.027**
(0.026)
(0.022)
(0.015)
(0.015)
(0.011)
Month F.E.
N
N
N
N
Y
Postal Code F.E.
N
N
N
N
Y
Treatment:
< 20 km
< 20 km
< 20 km
< 20 km
< 20 km
Observations
98,269
98,269
98,269
98,269
98,269
R2
0.062
0.335
0.682
0.694
0.829
Adjusted R2
0.062
0.335
0.682
0.694
0.826
Note: *p<0.1,**p<0.05,***p<0.01. The following variables are included as controls but omitted from the 
table: Type of housing


===== PAGE 11 (score 13, numeric-paren 0, table-words 5, var-words 3) =====

1 3
Page 11 of 27 
20
Economics of Disasters and Climate Change (2026) 10:20
With respect to a DiD approach, no assumption on the development of the trends is 
needed. However, large datasets are required, with the specific need to span several years. 
For this reason, we leave our original dataset unfiltered and consider all the transactions that 
occurred in Sweden between 2005 and 2019. Another challenge to the application of this 
method comes from the information available through the dataset as the geographic coor­
dinates in our dataset represent the centroid of the building structure. This introduces issues 
with apartments, because we are not able to distinguish between them. The geographical 
coordinates are the same for apartments located in the same building and we are not able to 
distinguish among them using other features such as the floor number and their sizes. We 
therefore proceed by dropping the observations pertaining to apartments when employing the 
repeat-sales approach. This represents only a minor issue, considering that most of the assets 
in the proximity of the wildfire are single- or double-family houses. The use of single- and 
multi-family houses only is, however, in line with the standard approaches adopted by the 
literature (Meldrum 2016; Wolf and Klaiber 2017). The use of the repeat sales method further 
reduces the size of the dataset as the assets that were sold only once within it are dropped.
The repeat sales approach is presented in Eq. (2):
	
display-eq-Eq u 2

(2)
Our dependent variable is therefore the difference between the log-transformed final selling 
price in period t and in period t-1 for each individual asset i. inline-eq-IEq12 is then a dummy 
variable which equals 1 if the wildfire occurred between these two consecutive transactions 
t and t-1. inline-eq-IEq13 is a dummy variable equaling one for the assets located within 10 
km from the perimeter of the wildfire. The rest of the control variables are the same as the 
ones employed in Eq. (2).
Results
Main Results
Table 2 depicts our main DiD results. As previously mentioned, the post dummy variable 
equals one for sales occurring after the wildfire, while the treatment dummy equals one for 
observations located within 20 km from the wildfire’s perimeter. Their interaction dummy, 
inline-eq- I Eq14 , is therefore considering the transactions that occurred after the event 
and within 20 km from this. In the first two columns we use sparse specifications without 
any spatial controls, only including the variables essential for the difference-in-difference 
analysis in column (1), while also adding some basic property characteristics in column (2). 
From these two columns, we begin by observing that prices in our treatment group differ 
from those in our control group, with the property prices in the former group being signifi­
cantly lower than in the latter. This is not surprising given that the wildfire took place in a 
rural area far from larger towns or cities, while the control group includes more observations 
in urban areas. Such evidence does not pose any challenge to the validity of our approach, as 
the DiD method assumes the trends to be parallel, which, again, we show to be reasonable 


===== PAGE 13 (score 80, numeric-paren 36, table-words 16, var-words 7) =====

1 3
Page 13 of 27 
20
Economics of Disasters and Climate Change (2026) 10:20
on average, final selling prices that were 10.1% (8.9%) lower than the control group, com­
pared to 2.7% lower prices estimated in Table 2. This is shown, respectively, in Column (2) 
and in Column (4) of Table 3. When expanding the treated distance to 25 km and including 
fixed effects, in column (6), we observe that the estimate is only marginally smaller than the 
− 2.7% reported in Table 2. Just like stated previously, the interpretation of the coefficient of 
interest, the interaction between post and treatment is not linear because the dependent vari­
able has been log-transformed. From Table 3, we can thus conclude that the effect on hous­
ing prices appears to have been most noticeable for sales that occurred in the proximity of 
the wildfire. Overall, the results suggest that the more distant from the wildfire, the smaller 
the impact on housing prices, corroborating the intuition that the wildfire is the event driv­
ing the results we observed. It is indeed reasonable to assume that as the treatment group 
is expanded to include asset more distant to the event, as when increasing the threshold 
distance to 25 km, the average effect in this group is diluted. At 27 km of distance from the 
perimeter of the wildfire the impact becomes statistically insignificant.
Robustness Analysis: Treatment Based on Postal Codes
We address in this section the issue related to the arbitrariness with which the distance 
thresholds were defined. The previous section has clarified that the direction of the results, 
i.e. a drop in selling prices, is not affected by the decision on the distance to use. What 
changes is the estimated effect, as more stringent distances identify narrower treatment 
groups that display, in absolute values, larger results. While informed by the previous lit­
erature, we acknowledge that the decision on the distance to use is still arbitrary. In order 
to further show that the results are unaffected by this arbitrariness, we employ postal codes 
as an alternative approach for identifying the treatment and the control group. In Table 4, 
we define the sales within the first ring of postal codes closest to the wildfire as the treated 
observations. The columns in this table are best compared with columns (4) and (5) of Table 
Table 3  Robustness analysis – alternative treatment distance
Dependent variable:
Log(selling price)
(1)
(2)
(3)
(4)
(5)
(6)
Treatment
−0.178***
−0.161***
−0.207***
−0.099**
−0.045***
0.096***
(0.038)
(0.055)
(0.028)
(0.045)
(0.010)
(0.020)
Post
0.279***
0.513***
0.280***
0.513***
0.283***
0.514***
(0.003)
(0.013)
(0.003)
(0.013)
(0.003)
(0.013)
Post * Treatment
−0.199***
−0.107**
−0.159***
−0.093***
−0.092***
−0.024**
(0.057)
(0.043)
(0.037)
(0.029)
(0.013)
(0.010)
Month F.E.
N
Y
N
Y
N
Y
Postal Code F.E.
N
Y
N
Y
N
Y
Treatment:
< 5 km
< 5 km
< 10 km
< 10 km
< 25 km
< 25 km
Observations
98,269
98,269
98,269
98,269
98,269
98,269
R2
0.693
0.829
0.694
0.829
0.694
0.829
Adjusted R2
0.693
0.826
0.694
0.826
0.694
0.826
Notes: *p < 0.1,**p < 0.05,***p < 0.01. The following variables are included as controls but omitted from the 
table: Type of housing, Living area, Age Age2, Number of rooms, W(Price), Distance to fire, Distance to 
lake, Distance to the train


===== PAGE 14 (score 61, numeric-paren 23, table-words 14, var-words 7) =====

1 3
20 
Page 14 of 27
Economics of Disasters and Climate Change (2026) 10:20
2 which, except for the treatment variable, have the same model specification. This iden­
tification strategy, however, introduced significant limitations on the possibility to employ 
postal codes fixed effect, as they would present perfect multicollinearity with the treatment 
variable. The specification is therefore the same as in Eq. (1) with the exception that we can­
not exploit postal codes fixed effect. As we acknowledge the utility of maintaining spatial 
fixed effects in our estimations, we here include municipality fixed effects. Such an inclu­
sion spares our model from perfect multicollinearity while still allowing for a valid spatial 
control of non-observable factors. Figure 2 provides a visual rendition of this approach.
We find that the effect of the wildfire on housing prices is larger when we use treatment 
based on postal codes rather than distance. This is shown in Table 4. Without fixed effects 
for municipalities, we find a discount of approximately 15%. With such fixed effects, the 
impact is estimated at a statistically significant − 11.9%. We then use sales in both the first 
and second ring of postal codes as the treated observations in columns (3) and (4) of Table 
4. The effects in this table are somewhat lower than in some of the results from Table 3, but 
still highly significant. The results confirm our previous findings of significant impacts from 
the wildfire on final selling prices.
Robustness Analysis: Placebo Tests
In Table 5, we present the results from two different placebo tests. In columns (1) and (2), 
we shuffle the existing distances in our dataset so that each observation is randomly assigned 
a distance from another observation in the data. In column (3) and (4), we instead assign a 
randomly generated distance (between 0 and 100 km) to each observation drawing from a 
uniform distribution. If the observed effects on property prices in Tables 2, 3 and 4 are really 
driven by the wildfire, we expect our interaction term between the post and treatment vari­
able to be insignificant in the placebo tests. As depicted by Table 5, the is confirmed by the 
evidence. In the first placebo test (column 1 and 2), both interaction terms are insignificant. 
Dependent variable:
Log(selling price)
(1)
(2)
(3)
(4)
Treatment
−0.280***
−0.194***
−0.102***
−0.0348***
(0.037)
(0.034)
(0.014)
(0.015)
Post
0.280***
0.483***
0.282***
0.4848***
(0.003)
(0.0157)
(0.003)
(0.015)
Post * Treatment
−0.163***
−0.127***
−0.129***
−0.099***
(0.050)
(0.0.045)
(0.018)
(0.014)
Month F.E.
N
Y
N
Y
Postal Code F.E.
N
N
N
N
Municipality F.E.
N
Y
N
Y
Treatment:
1 st Ring 
Postal 
Codes
1 st Ring 
Postal 
Codes
1 st and 
2nd Ring 
Postal 
Codes
1 st and 
2nd Ring 
Postal 
Codes
Observations
98,269
98,269
98,269
98,269
R2
0.694
0.7522
0.694
0.7521
Adjusted R2
0.694
0.7518
0.694
0.7517
Table 4  Robustness analysis – 
alternative treatment using zip 
codes
Notes: 
*p < 0.1,**p < 0.05,***p < 0.01. 
The following variables are 
included as controls but omitted 
from the table: Type of housing, 
Living area, Age Age2, Number 
of rooms, W(Price), Distance to 
fire, Distance to lake, Distance 
to train
 


===== PAGE 15 (score 10, numeric-paren 0, table-words 4, var-words 2) =====

1 3
Page 15 of 27 
20
Economics of Disasters and Climate Change (2026) 10:20
In the second test (columns 3 and 4), the magnitudes of the coefficients for interaction terms 
are weak and have the opposite sign of what we found in our earlier results. The coefficient 
in column (3) is insignificant, but we can observe a weakly significant interaction term in 
column (4). Overall, we argue that these tests provide support for the validity of our previ­
ous results that identified the wildfire as the main driver of the estimated effects.
Robustness Analysis: Repeat-Sales
In order to provide more robustness to the results presented in the previous sections, we 
address the same research question as before by employing a different method. The trade-
offs that the use of a repeat-sales approach entails have been discussed before in the Method 
section. We just stress that, while on the one hand this method relaxes the stringent assump­
tions that are necessary in order to obtain reliable results from a DiD application, on the 
other hand, the number of observations is significantly reduced and the only type of assets 
we can focus on are single- and multi-family houses. This results in a final dataset for this 
analysis with only 5,865 suitable observations. The treatment group used in Table 6 is made 
of couples of consecutive sales that occurred within 10 km from the limits of the wildfire 
and where the first transaction occurred before the wildfire and the following one after it. In 
order to effectively implement this additional strategy, two more variables are created. The 
former, labelled as treatmentd, is a dummy variable equalling 1 for the assets located within 
Fig. 2  Map of the postal codes around the area affected by the wildfire. The red area represents the burnt 
area. The first ring of postal codes that border this are depicted in yellow; the second ring of postal codes 
is in purple. Every green dot represents a transaction
 


===== PAGE 16 (score 67, numeric-paren 28, table-words 12, var-words 10) =====

1 3
20 
Page 16 of 27
Economics of Disasters and Climate Change (2026) 10:20
10 km from the perimeter of the wildfire. The latter, treatmentt, indicates the transactions 
pairs where the first transaction occurred before the wildfire and the second one occurred 
after it. The interaction between these two is the key variable of our strategy, equalling one 
for those transactions pairs that occurred within 10 km of the wildfire and where we com­
pare the transaction after the wildfire to the one, for the same asset, that occurred before it. 
In a sense, we are estimating how housing prices evolved, by keeping the assets character­
istics fixed, i.e. by comparing it to its previous price, before and after the event. The results 
reveal that, on average, the single-family houses in the treatment group experienced a price 
appreciation that was 11% to 17% lower than that experienced by the assets in the control 
group. This one is formed by single-family houses within 10 and 100 km from the wildfire. 
Dependent variable:
Δ(Selling price)
(1)
(2)
Treatmentd
−0.029
−0.051
(0.038)
(0.039)
Treatmentt
0.207***
0.114***
(0.006)
(0.008)
Treatmentd*Treatmentt
−0.183**
−0.115*
(0.073)
(0.070)
Years controls
N
Y
Postal Code F.E.
N
Y
Observations
5,865
5,865
R2
0.171
0.257
Adjusted R2
0.169
0.248
Table 6  Robustness test – repeat-
sales approach
Notes: 
*p < 0.1,**p < 0.05,***p < 0.01. 
The following variables are 
included as controls but omitted 
from the table: Living area, 
Number of rooms, Distance 
from fire, Distance from forest, 
Distance from lake, Age Age2, 
W(Price), Plot area, Distance 
from train
 
Dependent variable:
Log(selling price)
(1)
(2)
(3)
(4)
(Rand.) 
Treatment
0.017
0.016*
−0.007
−0.007
(0.011)
(0.008)
(0.006)
(0.004)
Post
0.280***
0.513***
0.277***
0.511***
(0.003)
(0.013)
(0.003)
(0.013)
Post * 
(Rand.) 
Treatment
−0.011
−0.010
0.010
0.008*
(0.015)
(0.011)
(0.006)
(0.005)
Month F.E.
N
Y
N
Y
Postal Code 
F.E.
N
Y
N
Y
Treatment:
Re-shuffle
Re-shuffle
Unif. 
Distribution
Unif. Dis­
tribution
Observa­
tions
98,269
98,269
98,269
98,269
R2
0.691
0.829
0.691
0.829
Adjusted R2 0.691
0.826
0.691
0.826
Table 5  Robustness analy­
sis – randomized treatment 
assignment
Notes: 
*p < 0.1,**p < 0.05,***p < 0.01. 
The following variables are 
included as controls but omitted 
from the table: Type of housing, 
Living area, Age Age2, Number 
of rooms, W(Price), Distance to 
fire, Distance to lake, Distance 
to train
 


===== PAGE 17 (score 9, numeric-paren 0, table-words 4, var-words 1) =====

1 3
Page 17 of 27 
20
Economics of Disasters and Climate Change (2026) 10:20
Despite the exclusion of apartments, the reduction in the number of observations and the 
change in the method employed, the results from the repeat-sales approach are comparable 
to what is found in Table 2. The fact that the results in Table 6 are slightly higher, may hint 
to the fact that the overall results in Table 2 may be driven by the impact that the wildfire had 
on single-family houses. We investigate such a possibility more in detail in the next section.
Robustness Analysis: Temporal Trends and Heterogenous Effects
The heterogeneity of the impact on single- and multi-family houses and on apartments is 
addressed in the two graphs of Fig. 3. This same figure also explores the temporal trend of 
the impact. Figure 3a) is produced by estimating Eq. (1) and introducing interaction terms 
between monthly dummy variables and the dummy variable equaling 1 for the assets sold 
within 20 km from the limits of the wildfire, i.e. the treatment variable in Eq. (1). In this 
sense, our approach is similar to the one explored by Zhu et al. (2016). By doing so, we 
are not only simply assessing the temporal trend of the estimated effects, but we are also 
Fig. 3  Temporal trends of the estimated impact with respect to the pre-treatment period. Notes: houses in 
blue, apartments in red
 


===== PAGE 20 (score 58, numeric-paren 23, table-words 13, var-words 6) =====

1 3
20 
Page 20 of 27
Economics of Disasters and Climate Change (2026) 10:20
Impact on Time-on-Market (TOM)
The list of variables available in the housing transactions dataset this study builds on allows 
us to investigate whether the event had any significant impact on other key indicators of the 
housing market, such as time-on-market (TOM). For some of the observations in our data­
set, we have information about the date of posting of the sale and the date of the sale. We are 
therefore able to compute a simple difference between these two variables and measure the 
time the asset spent unsold on the market. It must be cautioned, however, that the informa­
tion on the date of posting is not available for every transaction in the dataset, and that we 
therefore result to working with a dataset that is a subset of the original one. After removing 
observations with missing information, our dataset to investigate the impact of the 2014 
Västmanland wildfire on the assets’ TOM drops to 85,558 transactions, with an average 
TOM of 52 days. The estimation proceeds by following the specification in Eq. (1) by sim­
ply replacing the dependent variable with the number of days spent on the market unsold.
Table 8 presents the results from investigating this additional channel through which 
the wildfire may have impacted the local housing market in Sweden. The results are in 
line with both the theoretical predictions of search models and with the empirical evidence 
from the previous literature (Wheaton 1990; Krainer and LeRoy 2002; Hayunga and Pace 
2019), highlighting the positive correlation between housing prices and TOM. The wildfire 
is found to have significantly decreased the TOM of the affected areas, with assets located 
within 20 km from its perimeter experiencing an average drop of TOM everything else con­
stant, of 10 days (Column (1) in Table 8) or about 16% (Column (2) in Table 6). The findings 
could also be better understood in the context of the literature investigating the impact of 
exogenous shocks on the housing market in Sweden. Piseddu (2025) investigates the impact 
Dependent variable:
Log(selling price)
(1)
(2)
Treatment
−0.042***
−0.068**
(0.011)
(0.031)
Post
0.282***
0.515***
(0.003)
(0.013)
Smoke
−0.164***
0.041
(0.029)
(0.047)
Post * Treatment
−0.100***
−0.022*
(0.015)
(0.012)
Post * Smoke
0.008
−0.0004
(0.040)
(0.030)
Treatment * Smoke
−0.053
0.427***
(0.056)
(0.123)
Treatment * Post * Smoke
−0.085
−0.113*
(0.077)
(0.060)
Month F.E.
N
Y
Postal Code F.E.
N
Y
Observations
98,267
98,267
R2
0.694
0.830
Adjusted R2
0.694
0.827
Table 7  Assessing the impact of 
wildfire smoke on housing prices
Notes: 
*p < 0.1,**p < 0.05,***p < 0.01. 
The following variables are 
included as controls but omitted 
from the table: Type of housing, 
Living Area, Age Age2,Number 
of rooms, W(Price), Distance 
from fire, Distance from forest, 
Distance from lake, Distance 
train
 


===== PAGE 21 (score 71, numeric-paren 40, table-words 7, var-words 12) =====

1 3
Page 21 of 27 
20
Economics of Disasters and Climate Change (2026) 10:20
of the news of the Fukushima accident on the housing markets surrounding nuclear facili­
ties in Sweden and identifies a negative correlation between housing prices and TOM: while 
a drop in housing prices is observed, TOM is found to increase, two signals of increased 
adversity to risk from homebuyers. The diverging evidence between our findings and those 
in Piseddu (2025) could be reconciled by considering the nature of the events studied. In 
the case of the 2014 Västmanland wildfire, homebuyers could count on the fact that the 
forest will grow again and that the supply of natural amenities will be restored, providing a 
justification for the shortened TOM that is observed. In the case of the presence of nuclear 
facilities, the risk will not be removed, resulting in additional negative impacts that affect 
not only housing prices but also TOM.
Lastly, Figure A2 in the Appendix reveal that the wildfire had no significant impact on 
the number of transactions in the affected areas. This is confirmed both in absolute terms 
and in percentages, with the average quarterly number of transactions displaying its usual 
seasonality and the total number of transactions in Q3 and Q4 of 2014 surpassing those of 
Q3 and Q4 of 2015.
Dependent variable:
TOM (Days)
Log(TOM) (Days)
(1)
(2)
Treatment
6.008
0.161
(12.839)
(0.108)
Post
−18.128***
−0.351***
(5.280)
(0.044)
Living Area (m2)
0.206***
0.002***
(0.019)
(0.0002)
Age
−0.225***
−0.002***
(0.031)
(0.0003)
Age2
0.001***
0.00001***
(0.0001)
(0.00000)
Number of rooms
0.384
0.020***
(0.487)
(0.004)
W(Selling Price)
2.917
0.083**
(4.155)
(0.035)
Distance fire
−0.0004
−0.00000
(0.0004)
(0.00000)
Distance forest
−0.002
−0.00003
(0.002)
(0.00002)
Distant train
0.001*
0.00001***
(0.0004)
(0.00000)
Distance lake
−0.008***
−0.0001***
(0.001)
(0.00001)
Post*Treatment
−10.334**
−0.164***
(4.586)
(0.039)
Month F.E.
Y
Y
Postal Code F.E.
Y
Y
Observations
85,558
85,558
R2
0.095
0.183
Adjusted R2
0.079
0.168
Table 8  Assessing the impact of 
the wildfire on time-on-market 
(TOM)
Notes: 
*p < 0.1,**p < 0.05,***p < 0.01. 
The following variables are 
included as controls but omitted 
from the table: Type of housing
 


===== PAGE 22 (score 9, numeric-paren 0, table-words 1, var-words 7) =====

1 3
20 
Page 22 of 27
Economics of Disasters and Climate Change (2026) 10:20
Discussion and Policy Implications
The comparison between our results and those produced in other studies introduces the 
need for a discussion on the potential factors behind these. Looking, for instance, at the 
results produced by McCoy and Walsh (2018), our estimate in Table 3, with a treatment 
group made of assets within 5 km from the perimeters of the wildfire in column (2), 
display results that are similar to the 11% contraction that the authors observed for the 
assets within 2 km from the wildfires in Colorado. It must, however, be noted, that our 
results are mostly driven by the short-term impact that the wildfire had in the few weeks 
immediately after, while for the case of Colorado the impact is found to still be signifi­
cant even after several months after the event. It must also be noted that the authors focus 
on several wildfires, occurring in areas close to each other, that have a total affected area 
that is about 10 times the size of the event we studied in Sweden. While the area studied 
by McCoy and Walsh (2018) has been affected by several events, we are not aware of any 
other major wildfire in the years preceding or following our period of reference in our 
study area. One could argue that the longer temporal effects identified in Colorado may 
be driven by constant action of several consecutive wildfires that act as a reminder of the 
vulnerability of the area. On the contrary, the wildfire in Västmanland may be perceived 
as a single, unlikely, event in the area. Our results are also in line with some of the esti­
mates provided by Mueller et al. (2009) and this study is, to the best of our knowledge, 
one of the first to estimate the impact of the number of wildfires on housing prices. While 
focusing on a different market, in the US, the results presented by the authors corroborate 
the explanation that we provided here: the results reveal that the number of wildfires is 
a significant factor in explaining the impact of climate change risk, and that, while one 
wildfire has a marginal impact of roughly 10%, two wildfires reduce housing prices, on 
average, by 22%. In that sense, the single event in our study area and timeframe may 
make it easier for the housing market to recover significantly quicker than was previously 
estimated.
Another aspect that may explain the relatively quick recovery of housing prices in the 
aftermath of the wildfire is the pace of vegetation restoration. The direct damages from 
the wildfire can be reasonably expected to have impacted prices on the market, but while 
only few assets were actually damaged or destroyed (MSB, 2015), it can be argued that 
the main driver behind the impact of wildfires on housing values is the loss of values in 
the amenities: access and visibility of forests have been previously identified as substantial 
positive drivers of housing prices (Poudyal et al. 2010; Belcher and Chisholm 2018). The 
post-event analysis of the Västmanland fire proposed by Granath et al. (2021) presents evi­
dence of a rapid pace of vegetation regrowth in the area. Such an occurrence may explain 
the short-lived impact of the event studied here. The relatively low number of assets dam­
aged or destroyed by the event, 71 (MSB, 2015), may also be a factor that explains the 
short-lived impact, compared to other occurrences of more destructive wildfires (Kiel and 
Matheson 2018; Kim and Yoon 2025). Expected utility theory may offer a further explana­
tion of the results observed. Most of the assets sold in the rural portions of our dataset are 
intended to be used as summerhouses. If the utility of the buyer is only affected by the use 
of these assets that can be done of them in summer, an event at the end of the season is not 
expected to significantly affect the market for a long period. If the buyer values the utility