The document analyzes the impact of a 2012 earthquake in Emilia-Romagna, Italy on firms located within and outside industrial districts (IDs) in the region. It reviews literature on natural disasters and economic performance, discusses theoretical perspectives on whether IDs strengthen or weaken firm resilience, and presents research questions on the differential impact on firms in/out of IDs. The methodology section describes using firm-level data from 2010-2013 and differences-in-differences estimations to address non-random spatial distribution of firms.
Natural disasters and firm resilience in Italian industrial districts - Giulio Cainelli, Andrea Fracasso, Giuseppe Vittucci Marzetti
1. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Natural disasters and firm resilience
in Italian industrial districts
Giulio Cainelli,1
Andrea Fracasso,2
Giuseppe Vittucci Marzetti3
OECD SPL, Trento
February 7, 2019
1Dpt. of Economics and Management, University of Padova, email: giulio.cainelli@unipd.it
2Department of Economics and Management and School of International Studies, University of
Trento, email: andrea.fracasso@unitn.it
3Department of Sociology and Social Research, University of Milano-Bicocca, email:
giuseppe.vittucci@unimib.it
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 1/29
2. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Layout
1 Aim and theoretical framework
Aim and background
Literature review: natural disasters and economic performance
Theoretical background: industrial districts and resilience
On natural disasters as natural experiments
2 Data and empirical methodology
The 2012 Emilia-Romagna earthquake
Data
Research questions
Empirical methodology
3 Estimation results
Average impact of the earthquake
Industrial district effect
4 Closing remarks
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 2/29
3. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Aim and background
Literature review: natural disasters and economic performance
Theoretical background: industrial districts and resilience
On natural disasters as natural experiments
Aim and background
The impact of natural disasters on economic performance and
growth has recently become an object of research.
Underlying idea and implicit assumptions:
most natural events are unpredictable and “random”;
natural events are exogenous shocks that can be used as natural
experiments to test economic hypotheses;
highly localized events impact mostly firms’ production and not final
demand, facilitating identification.
This work falls within this strand of the literature:
use a large sample of firms in Emilia-Romagna in the period
2010-2013, around the time of a localized earthquake sequence of
severe intensity (May 2012)
assess whether the location of firms within an industrial district
mitigates or exacerbates the impact of the disaster on their activity
and performance.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 3/29
4. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Aim and background
Literature review: natural disasters and economic performance
Theoretical background: industrial districts and resilience
On natural disasters as natural experiments
Natural disasters and economic performance
Most contributions focusing on the relationship between natural
disasters and economic performance are cross-country and use
macroeconomic data (Cavallo & Noy, 2009; Lazzaroni & van
Bergeijk; 2014).
These analysis do not allow to detect how local conditions and
individual factors interact with the shocks (Barone & Mocetti, 2014).
Only few studies investigate firms’ performance after a localized
major supply shock by using firm-level data.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 4/29
5. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Aim and background
Literature review: natural disasters and economic performance
Theoretical background: industrial districts and resilience
On natural disasters as natural experiments
Firm-level studies on the impact of natural disasters
Cole et al. (2013) use plant-level Japanese data to estimate the
impact of 1995 Kobe earthquake on firms’ survival.
Highly damaged firms face higher risk of exit.
In surviving firms:
value added and employment lower during the reconstruction and
higher afterwards;
productivity always higher.
Mel et al. (2012) investigate business recovery in Sri Lanka after the
2004 tsunami:
Affected firms lag behind unaffected comparable firms.
Direct aid helps recovery, more in services than in manufacturing.
Fabling et al. (2014) analyze the impact of the Canterbury
earthquakes in 2010-2011 in New Zealand:
Pre-shock profitability increases survival probability.
Coelli & Manasse (2014) investigate the impact of the floods in
Veneto in 2010:
After recovery, affected firms perform better than unaffected.
Aid transfers in the aftermath significantly contribute to the recovery.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 5/29
6. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Aim and background
Literature review: natural disasters and economic performance
Theoretical background: industrial districts and resilience
On natural disasters as natural experiments
Firm-level studies on the impact of natural disasters
Hayakawa et al. (2015) analyze how the 2011 flood in Thailand
affected the procurement patterns at Japanese affiliates of MNCs:
Natural disasters do not have persistent effects.
Adjustments among suppliers by MNCs depend on ex-ante
knowledge of alternative sources.
Todo et al. (2013) and Tokui et al. (2017) investigate the role of
supply chains in firms’ recovery after the 2011 Great East Japan
earthquake:
Supply chains have two opposite effects:
make recovery harder because of higher vulnerability to network
disruption;
facilitate recovery through support from trading partners, easier
search for new partners, and agglomeration economies.
evidence is that the positive effects exceed the negative ones.
Cole et al. (2015) analyze the effect of clustering on survival in the
1995 Kobe earthquake in Japan:
Firms’ location in clusters reduces survival probabilities.
It does not impact much on firms’ performance after the shock.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 6/29
7. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Aim and background
Literature review: natural disasters and economic performance
Theoretical background: industrial districts and resilience
On natural disasters as natural experiments
Theoretical background: industrial districts and resilience
While Industrial Districts (IDs) generate positive externalities, at a
theoretical level it is not clear whether they strengthen or weaken firms’
vulnerability to large negative supply shocks:
+ higher resilience for:
Agglomeration externalities: higher productivity, profitability, survival
rates in good times.
Risk sharing via interlinking transactions (Dei Ottati, 1994; Cainelli,
Montresor & Vittucci Marzetti, 2012).
Fiscal stimulus and external aid flowing faster towards IDs, insofar
they have vantage positions in terms of signaling, lobbying and
political connections (Brioschi et al., 2002; Brusco, 1982; Brusco et
al., 1996; Cainelli & Zoboli, 2004; Noy, 2009).
– lower resilience for:
Shock transmission via supply chains (Henriet et al., 2012, Carvalho
et al., 2014).
Higher reliance on local public goods.
Localized lending relationships and risk sharing mechanisms,
increasing the probability of mass defaults (Cainelli, Montresor &
Vittucci Marzetti, 2012).
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 7/29
8. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Aim and background
Literature review: natural disasters and economic performance
Theoretical background: industrial districts and resilience
On natural disasters as natural experiments
Research question and analysis
As the theory does not tell whether IDs make firms more or less
resilient to disruptive supply shocks, the overall effect is an empirical
issue.
By using firm-level data in 2010-2013 for a large sample of firms in
Emilia-Romagna (hit by an earthquake sequence in May 2012), we
estimate the effect of natural disasters on affected firms and the
differential impact on those located within/outside IDs.
Cole et al. (2015) is the closest paper in spirit. Main differences
besides the natural disaster
We focus on IDs rather than clusters;
We apply different techniques and estimation methods.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 8/29
9. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Aim and background
Literature review: natural disasters and economic performance
Theoretical background: industrial districts and resilience
On natural disasters as natural experiments
Natural disasters as natural experiments: methodological
issues
We argue that natural disasters cannot be treated as “exogenous” in
econometric terms and that they do not give rise to “natural
experiments” cause:
The spatial distribution of firms is correlated with characteristics at
the firm level that affect firms’ performance and resilience.
The unconditional probability of a firm being “treated” (hit by the
shock) is not “as good as random”, for it is correlated with such
characteristics.
Example:
Due to specialization economies, firms in the same sector tend to be
spatially concentrated.
When an earthquake hits a region, the “treatment group” (the firms
hit) and the “control group” (firms not hit) are systematically
different, at least in terms of the sector they belong to.
This calls for caution and analytical tools to address the issue.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 9/29
10. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
The 2012 Emilia-Romagna earthquake
Data
Research questions
Empirical methodology
Emilia-Romagna earthquake in 2012
Part of the Emilia-Romagna region
(North-East of Italy) hit by a sequence of
major earthquakes between May 20 and June
6 2012.
Widespread structural damages:
historical buildings collapsed, and
warehouses and factories partially or totally
destroyed.
This natural disaster:
has not yet been covered in the literature;
is interesting for in the region:
the industrial density is high;
there are several Industrial Districts (IDs). Source: Wikimedia.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 10/29
11. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
The 2012 Emilia-Romagna earthquake
Data
Research questions
Empirical methodology
Data
Bureau van Dijk financial information on about 26,000 firms
(manufacturing and KIBS) located in Emilia-Romagna during the
period 2010-2013
Industrial district
Earthquake
Total
No Yes
No 14,937 1,886 16,823
Yes 6,940 2,522 9,462
Total 21,877 4,408 26,285
Dependent variables:
Turnover.
Tangibles.
Bank debt/sales ratio.
Value Added (VA).
Production value.
Return On Equity (ROE): Net Income/Equity %.
Return On Sales (ROS): EBIT/Net Sales %.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 11/29
13. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
The 2012 Emilia-Romagna earthquake
Data
Research questions
Empirical methodology
Distribution by Province
Province no earthquake - no district earthquake - no district no earthquake - district earthquake - district Total
No. Col % Row % No. Col % Row % No. Col % Row % No. Col % Row % No. Col %
Bologna 5931 39.7 85.6 940 49.8 13.6 2 0.0 0.0 57 2.3 0.8 6930 26.4
Ferrara 201 1.3 17.2 840 44.5 72.0 75 1.1 6.4 51 2.0 4.4 1167 4.4
Forl`ı-Cesena 964 6.5 51.0 0 0.0 0.0 925 13.3 49.0 0 0.0 0.0 1889 7.2
Modena 1391 9.3 24.6 0 0.0 0.0 2498 36.0 44.1 1773 70.3 31.3 5662 21.5
Parma 2686 18.0 93.8 0 0.0 0.0 177 2.6 6.2 0 0.0 0.0 2863 10.9
Piacenza 1076 7.2 85.3 0 0.0 0.0 185 2.7 14.7 0 0.0 0.0 1261 4.8
Ravenna 867 5.8 52.6 0 0.0 0.0 781 11.3 47.4 0 0.0 0.0 1648 6.3
Reggio nell’Emilia 298 2.0 9.1 106 5.6 3.2 2244 32.3 68.2 641 25.4 19.5 3289 12.5
Rimini 1523 10.2 96.6 0 0.0 0.0 53 0.8 3.4 0 0.0 0.0 1576 6.0
Total 14937 100.0 56.8 1886 100.0 7.2 6940 100.0 26.4 2522 100.0 9.6 26285 100.0
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 13/29
14. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
The 2012 Emilia-Romagna earthquake
Data
Research questions
Empirical methodology
Research questions
1 What was the average effect of the earthquake sequence on firms’
performance in the short- and medium-term?
2 How did the location in an ID mediate the impact of the earthquake
on firms’ performance?
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 14/29
15. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
The 2012 Emilia-Romagna earthquake
Data
Research questions
Empirical methodology
Estimating the average impact of the earthquake
To estimate the average effect of the earthquake on firms’ performance
we employ two alternative approaches:
Difference-In-Differences (DID).
Propensity Score Matching (PSM) in:
levels;
first-differences.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 15/29
16. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
The 2012 Emilia-Romagna earthquake
Data
Research questions
Empirical methodology
Estimating the average impact of the earthquake: DID
The DID approach compares the change in the performance of firms
located in an area hit by the earthquake with that of firms placed in
areas not affected by the disaster, after controlling for a number of
firm- and area-specific characteristics.
The impact of the earthquake is captured by estimating either:
yit = ai + β0tt + β1ei + β2ei tt + uit (1)
or:
∆yi = δ0 + δ1ei + δ2Xi + νi (2)
yit is the performance variable of interest (turnover, tangibles,
debt/sales, VA, production, ROE, ROS) for the firm i in period t;
tt is a time dummy equal to 1 for the period after the earthquake
and 0 otherwise (t ∈ {0, 1} is the pre/post earthquake period);
ei is the earthquake dummy, equal to 1 if the firm is located in an
area hit by the earthquake;
Xi is a vector of firm-level controls
(sector, ID dummy, year of incorporation).
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 16/29
17. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
The 2012 Emilia-Romagna earthquake
Data
Research questions
Empirical methodology
Estimating the average impact of the earthquake: DID
Equation (1):
Fixed-Effects (FE) is a consistent estimator.
The parameter of interest is the coefficient attached to the
interacting term ei tt (β2).
Equation (2):
The model is in first-differences and includes firm-specific controls
possibly affecting the rate of change.
OLS is a consistent estimator.
The parameter of interest is the coefficient attached to the dummy
ei (δ1).
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 17/29
18. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
The 2012 Emilia-Romagna earthquake
Data
Research questions
Empirical methodology
Estimating the average impact of the earthquake: PSM
An alternative approach to quantify the average effect of the
earthquake is the Propensity Score Matching (PSM).
PSM controls for confounding factors in the estimation of the impact
of the treatment by ensuring that the comparison is performed using
treated and control units that are as similar as possible.
Steps:
The pre-treatment characteristics of the firms are summarized in a
single variable (the propensity score) by means of a probit/logit
estimation;
Similar treated and control firms are matched;
The average effect of the treatment on the treated is computed as
the average difference between the values of the variable of interest
for the treated and control firms in each pair of matched firms.
This approach requires that the sample contains enough couples of
treated and control units with the same propensity score.
If industries are entirely concentrated in one area, no control group is
actually available.
The PSM can be applied to levels and first-differences.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 18/29
19. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
The 2012 Emilia-Romagna earthquake
Data
Research questions
Empirical methodology
Estimating the effect of industrial districts
To analyze the possible influence of being in an ID on the impact of
the earthquake, in Eq. (1) and (2) we add a district dummy with
the associated interacting terms aimed at capturing differences in
the average impact of a unique treatment, i.e. the earthquake, for
district vs. non-district firms.
yit = ci + γ0tt + γ1di tt + γ2ei tt + γ3ei di tt + it (3)
or
∆yi = π0 + π1di + π2ei + π3ei di + π4Xi + νi (4)
where di is a district dummy (equal to 1 if the firm is in a district).
The parameters of interest are γ3 in Eq. (3) and π3 in Eq. (4).
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 19/29
20. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Average impact of the earthquake
Industrial district effect
Time series of mean turnover and VA for firms hit/not hit
by the earthquake
(a) Turnover (b) Value Added
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 20/29
21. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Average impact of the earthquake
Industrial district effect
Time series of mean tangibles and debt-sales ratio for firms
hit and not hit by the earthquake
(c) ln Tangibles (d) Bank debt/sales ratio
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 21/29
22. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Average impact of the earthquake
Industrial district effect
Average impact of the earthquake: estimation results
PSM in levels Panel FE: 2010/11– First difference: 2010/11–
Dependent
variable
2012 2013 2012 2013 2012 2013
OLS PSM OLS PSM
ln(turnover) -.0033 .0318 -.0281∗∗
-.0003 -.0225∗∗
-.0223∗
.0150 .0138
(.0362) (.0367) (.0118) (.0150) (.0114) (.0123) (.0145) (.0158)
ln(tangibles) .0036 .0382 .0015 .0299∗
.0010 -.0004 .0301∗
.0295∗
(.0457) (.0463) (.0117) (.0158) (.0117) (.0126) (.0164) (.0175)
debt/sales -.3758 -.0487 .8761∗∗∗
1.107∗∗∗
.7413∗∗∗
.7469∗∗∗
1.012∗∗∗
.9642∗∗∗
(.4747) (.4792) (.2646) (.3302) (.2807) (.2967) (.3403) (.3579)
ln(value-added) -.0377 -.0210 -.0471∗∗∗
-.0094 -.0451∗∗∗
-.0442∗∗∗
-.0051 -.0065
(.0366) (.0374) (.0129) (.0151) (.0134) (.0142) (.0151) (.0161)
ln(production) -.0117 .0301 -.0368∗∗∗
.0002 -.0282∗∗∗
-.0289∗∗∗
.0136 .0123
(.0354) (.0358) (.0111) (.0140) (.0108) (.0115) (.0137) (.0147)
ROE .2943 1.056∗∗
-.8054 .1686 -.4791 -.3078 .3343 .4914
(.5623) (.5366) (.5366) (.5606) (.5739) (.5970) (.5753) (.6102)
ROS -.8103∗∗∗
-.4231∗
-.6227∗∗∗
.0589 -.5348∗∗∗
-.5658∗∗∗
.0930 .0468
(.2483) (.2450) (.2148) (.2350) (.2311) (.2426) (.2457) (.2588)
Robust standard errors in parenthesis. Controls for sector (4-digit), incorporation year and district dummy included in
OLS (OLS regression) and PSM (Propensity Score Matching). Significance at: 1% ∗∗∗; 5% ∗∗; 10% ∗.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 22/29
23. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Average impact of the earthquake
Industrial district effect
Average impact of the earthquake: main results
Short-term (6-7 months) average impact of earthquake on activity
and efficiency:
statistically significant decrease of turnover, production value, value
added (and ROS);
increase in debt-sales ratios.
No evidence of longer term effects (18 months), but for the slightly
higher debt-sales ratio.
Biased estimates in PSM in levels for systematic differences in
pre-treatment levels between firms hit/not hit by the earthquake not
accounted by the controls.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 23/29
24. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Average impact of the earthquake
Industrial district effect
The effect of industrial districts: FE estimator
Dependent variable Regressor
Years: 2010/11-
2012 2013
ln(turnover) District dummy (d) · Time dummy (t) .0026 (.0104) -.0056 (.0136)
Earthquake dummy (e) · Time dummy (t) -.0083 (.0175) .0186 (.0197)
e · d · t -.0358 (.0241) -.0304 (.0286)
e · t + e · d · t -.0441∗∗∗
(.0166) -.0118 (.0207)
ln(tangibles) District dummy (d) · Time dummy (t) .0167 (.0106) .0155 (.0141)
Earthquake dummy (e) · Time dummy (t) .0048 (.0154) .0322 (.0234)
e · d · t -.0130 (.0226) -.0107 (.0324)
e · t + e · d · t -.0082 (.0166) .0214 (.0223)
debt/sales District dummy (d) · Time dummy (t) .0735 (.2424) .0583 (.2945)
Earthquake dummy (e) · Time dummy (t) .1706 (.3697) .2849 (.4767)
e · d · t 1.229∗∗
(.5424) 1.477∗∗
(.6715)
e · t + e · d · t 1.400∗∗∗
(.3968) 1.761∗∗∗
(.4802)
ln(value-added) District dummy (d) · Time dummy (t) .0044 (.0114) .0099 (.0134)
Earthquake dummy (e) · Time dummy (t) -.0253 (.0185) .0046 (.0211)
e · d · t -.0401 (.0261) -.0291 (.0296)
e · t + e · d · t -.0655∗∗∗
(.0184) -.0245 (.0208)
Robust standard errors in parenthesis. Significance at: 1% ∗∗∗; 5% ∗∗; 10% ∗.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 24/29
25. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Average impact of the earthquake
Industrial district effect
The effect of industrial districts: FE estimator
Dependent variable Regressor
Years: 2010/11-
2012 2013
ln(production) District dummy (d) · Time dummy (t) .0019 (.0101) .0013 (.0128)
Earthquake dummy (e) · Time dummy (t) -.0116 (.0149) .0204 (.0188)
e · d · t -.0451∗∗
(.0210) -.0359 (.0268)
e · t + e · d · t -.0566∗∗∗
(.0148) -.0155 (.0198)
ROE District dummy (d) · Time dummy (t) .2851 (.4725) 1.046∗∗
(.4912)
Earthquake dummy (e) · Time dummy (t) -.6541 (.7976) 1.038 (.7755)
e · d · t -.1384 (1.103) -1.979∗
(1.111)
e · t + e · d · t -.7925 (.7616) -.9412 (.7955)
ROS District dummy (d) · Time dummy (t) .0037 (.1922) .0920 (.2112)
Earthquake dummy (e) · Time dummy (t) -.4211 (.3218) .3739 (.3458)
e · d · t -.3526 (.4426) -.5904 (.4693)
e · t + e · d · t -.7737∗∗
(.3039) -.2166 (.3172)
Robust standard errors in parenthesis. Significance at: 1% ∗∗∗; 5% ∗∗; 10% ∗.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 25/29
26. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Average impact of the earthquake
Industrial district effect
The effect of industrial districts: FD
Dependent variable Regressor
Years: 2010/11-
2012 2013
∆ ln(turnover) District dummy (d) -.0039 (.0109) -.0149 (.0138)
Earthquake dummy (e) -.0094 (.0154) .0206 (.0200)
e · d -.0254 (.0225) -.0109 (.0289)
e + e · d -.0348∗∗
(.0166) .0100 (.0211)
∆ ln(tangibles) District dummy (d) .0103 (.0107) .0049 (.0142)
Earthquake dummy (e) .0055 (.0155) .0334 (.0230)
e · d -.0087 (.0228) -.0065 (.0322)
e + e · d -.0033 (.0170) .0269 (.0229)
∆ debt/sales District dummy (d) .0025 (.2484) -.0100 (.3038)
Earthquake dummy (e) .0711 (.3773) .1863 (.4548)
e · d 1.336∗∗
(.5538) 1.668∗∗
(.6683)
e + e · d 1.407∗∗∗
(.4103) 1.855∗∗∗
(.4982)
∆ ln(value added) District dummy (d) -.0026 (.0115) .0016 (.0136)
Earthquake dummy (e) -.0263 (.0185) .0042 (.0212)
e · d -.0366 (.0262) -.0236 (.0299)
e + e · d -.0629∗∗∗
(.0190) -.0194 (.0216)
Robust standard errors in parenthesis. Significance at: 1% ∗∗∗; 5% ∗∗; 10% ∗.
Regressions include sector (4-digit) dummies and year of incorporation.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 26/29
27. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Average impact of the earthquake
Industrial district effect
The effect of industrial districts: FD
Dependent variable Regressor
Years: 2010/11-
2012 2013
∆ ln(production) District dummy (d) -.0020 (.0102) -.0054 (.0129)
Earthquake dummy (e) -.0097 (.0150) .0230 (.0189)
e · d -.0358∗
(.0213) -.0183 (.0270)
e + e · d -.0455∗∗∗
(.0153) .0047 (.0196)
∆ ROE District dummy (d) -.3928 (.4786) .9712∗
(.4959)
Earthquake dummy (e) -.5294 (.8098) 1.139 (.7810)
e · d .0980 (1.123) -1.554 (1.123)
e + e · d -.4314 (.7963) -.4155 (.8241)
∆ ROS District dummy (d) -.0418 (.1965) -.0097∗∗
(.2172)
Earthquake dummy (e) -.3433 (.3264) .3435 (.3516)
e · d -.3730 (.4508) -.4895 (.4797)
e + e · d -.7163∗∗
(.3192) -.1460 (.3354)
Robust standard errors in parenthesis. Significance at: 1% ∗∗∗; 5% ∗∗; 10% ∗.
Regressions include sector (4-digit) dummies and year of incorporation.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 27/29
28. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Average impact of the earthquake
Industrial district effect
The effect of industrial districts: main results
Negative short-term impact of the earthquake on the activity and
the efficiency (production, turnover, value added, and ROS) slightly
higher for firms located in industrial districts.
Impact of earthquake on firm indebtedness (debt/sales):
positive and longer;
larger for firms in IDs.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 28/29
29. Aim and theoretical framework
Data and empirical methodology
Estimation results
Closing remarks
Closing remarks
Our analysis confirms previous findings that major supply shocks
have limited and temporary (negative) effects on surviving
companies.
Indebtedness appears the tool through which firms preserve their
activities.
The location of firms within industrial districts weakens their
response to localized exogenous shocks, but only in the short term.
Caution with “causal interpretation”:
not truly random shock if localization is very high in many industries;
limited capacity to control for unobservables (for the “ID effect” in
particular);
ID channels not disentangled.
Cainelli, Fracasso, Vittucci Marzetti Natural disasters and firm resilience in Italian industrial districts 29/29