The impact of floods on firms' performance

Coelli, Federica; Manasse, Paolo Working Paper The impact of floods on firms' performance Quaderni - Working Paper DSE, No. 946 Provided in Cooperation with: University of Bologna, Department of Economics Suggested Citation: Coelli, Federica; Manasse, Paolo (2014) : The impact of floods on firms' performance, Quaderni - Working Paper DSE, No. 946, Alma Mater Studiorum - Università di Bologna, Dipartimento di Scienze Economiche (DSE), Bologna, https://doi.org/10.6092/unibo/amsacta/4025 This Version is available at: https://hdl.handle.net/10419/159785 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/3.0/ ISSN 2282-6483 The impact of floods on firms’ performance Federica Coelli Paolo Manasse Quaderni - Working Paper DSE N°946 The impact of floods on firms’ performance∗ Federica Coelli†Paolo Manasse‡ March 2014 Abstract We estimate the short-run impact of a major flood that hit the region of Veneto in 2010 on firms’ performance. Using firm level data and a difference in difference approach we compare the value added growth of hit firms to the one of a control group of companies that are not exposed to the flood. The results indicate that the value added growth of affected firms is 6.9% higher two years after the flood. We further investigate the role of aid transfers in the aftermath of the disaster event. Considering both the flood and the aid treatment, we construct four mutually exclusive and exhaustive groups. The results indicate that, among firms exposed to the flood, both the ones that benefit from financial aid and the ones that don’t grow faster than the reference groups of firms that neither are exposed to the flood, nor receive financial aid. We also find a 2% additional growth effect that is attributable to the contribution of aid in the recovery phase. Keywords: Natural disasters, flood, firm growth, difference in differences, financial aid JEL classification: D24, Q54, R10, C23 ∗This working paper is based on the master thesis ‘The impact of floods on firms’ performance”; author: Federica Coelli; supervisor: Paolo Manasse. †University of Bologna, Italy. Email: federica.co[email protected]. ‡University of Bologna, Italy. Email: [email protected]. 1 Introduction Natural disasters are not a new phenomenon. Human societies have always been hit by calamities such as floods, hurricanes, storms and earthquakes that resulted in human, social, economic and environmental losses. Some recent catastrophes, such as the December 2004 tsunami, hurricane Katrina in 2005 and the earthquake in Haiti in 2010, brought the issue of the human and material costs of these events to the forefront of public attention and fueled a debate about the influence that a warming climate could have on the frequency and intensity of climateand weather-related disasters. In the last years, indeed, the consensus that climate change could be worsening weather and climate extremes has increased. In its 2012 Special Report the Intergovernmental Panel on Climate Change (IPCC) notes that a changing climate leads to changes in the frequency, intensity, spatial extent, duration, and timing of extreme weather and climate events, and can result in unprecedented extreme weather and climate events (Field, Barros, Stocker, & Dahe, 2012, p. 7). According to the IPCC (Field et al., 2012) floods are the most frequent natural disaster in Europe and data from the International Disaster Database (EM-DAT)1show that floods, immediately followed by storms, caused the highest total and average annual damages among the reported natural disasters between 1990 and 2012 in Europe (Fig.1). In the recent Climate Change, impacts and vulnerability in Europe 2012 report the European Environmental Agency (EEA, 2012) reports that hydro-meteorological events (storms, floods, and landslides) represent 75% of the natural disasters that occurred in Europe since 1980 and account for 64% of the reported damage costs. The report (EEA, 2012) also analyzes the impact of extremeprecipitations-related disasters on society, environment, agriculture, industry and ecosystem services due to both climatic and non-climatic factors. Taking into account both climate change and socio-economic changes, the EEA calculates the Gross Value Added (GVA) affected by floods for the ‘Economy First’2scenario and concludes that for many parts of Europe the GVA affected by floods for the 2050s is larger than for the baseline scenario (EEA, 2012, p. 214). In the last years extreme precipitation events that translated into localized floods have been at the forefront of media and public attention also in Italy. Beyond the human, material and environmental consequences, localized floods are thought to have detrimental effects on the economic fabric of the affected areas, a concern that also motivates aid flows in the aftermath of the disaster. Considering this, the focus of this study is on one major flood that hit the Veneto region in autumn 2010 and it aims to assess the economic impact of such flood on firms’ performance up to two years after the event. Between October 31st and November 2nd 2010 Veneto has been hit by one of the most severe extreme weather events in the last 50 years according to the report published by the Regional Environmental Agency and the authorities of the Veneto Region (ARPAV). The event has been characterized by abundance and persistence of the rainfall rather than by intensity, according 1The EM-DAT database is maintained by the Centre for Research on the Epidemiology of Disasters, based at the Catholic University of Louvain in Belgium. 2The EEA defines the Economy First as a scenario where a globalised and liberalised economy pushes the use of all available energy sources and an intensification of agriculture where profitable. The adoption of new technologies and water-saving consciousness are low and thus, water use increases. Only water ecosystems providing ecological goods and services for economies are preserved and improved (EEA, 2012, p. 214). 2 Figure 1: Average annual damages Source: EM-DAT: The OFDA/CRED International Disaster Database www.emdat.be - Universit´e Catholique de Luvain, Brussels Belgium. to the same report. In total an average of 173mm of rain fell during the period of observation (28/10/2010 - 11/11/2010) with peaks of more than 540mm in the Provinces of Belluno, Padova, Treviso and Vicenza. To give an idea of the magnitude of the extreme weather event Fig. 2 below shows the total amount of rainfall measured by each weather station in the region during the 15 days of observation, although severe precipitations have affected the region mainly for three days, from October 31st to November 2nd 2010. Fig. 3 shows the maximum intensity of daily precipitation3measured by the 193 weather stations compared to the average quantity usually falling on an average rainy day4. The graph clearly shows that in some areas of the region the amount of rain that fell on one day during the flood is more than twenty times the quantity that falls on an average rainy day. Both figures indicate the great variability of the quantity of rainfall in different areas of the region and it is evident that while some areas experienced extreme precipitations, others didn’t. This feature is also shown in Fig. 4, which displays the distribution of the precipitations in the region according to the location of the 193 weather stations. 3Maximum daily precipitation measures the intensity of precipitation on a daily basis during the observed period. It is calculated as the total quantity of rainfall fallen on the day when it rained most. 4The quantity of precipitations on each of the flooding days, compared to the average quantity on an average rainy day, can be seen in the Appendix. 3 Figure 2: Total precipitations registered by each weather station. The figure shows the quantity of precipitations (in mm) registered by 193 weather stations located in Veneto Region in the observed period (28/10/2010 - 11/11/2010). Each value is a sum over the period. Source: graph made using data from ARPAV. Figure 3: Maximum precipitations during the flood. The figure shows the maximum precipitations (in mm) registered by each weather station in one day. The ‘mean by weather station’ (green line) represents the average quantity of rainfall fallen on a raining day in October and November during the observed period (2003-2012) excluding 2010. The value has been calculated by taking an average between October and November for each weather station. The data are available only for 163 stations. The ‘overall mean’ (red line) represents an average of the quantity of precipitations on a raining day in October and November calculated as an average among 163 weather stations during the observed period excluding 2010. The data are available only for 163 weather stations. The average value has been extended to the remaining 30 stations. Source: graph made using data from ARPAV. 4 Figure 4: Distribution of precipitations during the flood. The figure shows the distribution of precipitations in the Region according to the location of the 193 weather stations in Veneto Region. Only the proper flooding days (October 31st , November 1st and November 2nd) are considered in this figure. Precipitations are measured in mm. Source: graph made using data from ARPAV. A natural question that arises when looking at these figures is whether the flood had a negative impact on the economic activity of firms located in areas that were hit by the flood and if there is any significant difference between these companies and those that haven’t been directly exposed to it. Like most of studies and due to data availability the analysis focuses on the short-term effects (up to two years after the event). However, contrary to the majority of studies in the literature on the impact of natural disasters, this analysis uses firm-level data of a panel of 2065 manufacturing companies during the period 2003-2012. Given the focus on a small area of the country, characterized by firms with similar characteristics and operating in a similar context, the natural approach to perform this analysis is a difference in differences (DID) methodology. Following Leiter, Oberhofer, and Raschky (2009) we also take into account different levels of exposure and vulnerability of capital stock including the assets composition in the estimation. Finally, recognizing that disaster and recovery aid could represent a confounding factor in our analysis, we explicitly control for the presence of financial aid in the aftermath of the flood and distinguish between firms that received aid flows and those that didn’t. The results indicate that firms affected by the flood experience faster value added growth two years after the natural disaster, with a growth rate that is 6.9% higher than the one of firms that haven’t been directly hit by the flood. We also found a 2% additional growth effect that is attributable to the presence of aid flows. The paper is organized as follows: Section 1 reviews the existing literature on the economic impact of natural disasters. Section 2 describes the data and the main variables. Section 3 discusses the empirical strategy (model and econometric approach). Section 4 presents the results. The last section concludes. 5 1 Literature review The focus of this literature review is on the economic impacts of natural disasters (in particular weather-related disasters), although it should be noticed that the consequences of natural catastrophes go far beyond the economic losses. Indeed, natural disasters comprise profound human, social, environmental and political consequences whose analysis, however, is beyond the scope of this study. Many scholars distinguish between direct and indirect impacts. The former refer to the physical destruction caused to property and human beings (National Research Council, 1999). This category includes: damages to homes and their contents, damages to firms’ capital and lost production, damages to infrastructure, people killed or injured, environmental degradation and emergency response (Kousky, 2013). Indirect damages, that some authors prefer to define as higher-order impacts (Rose, 2004), result from the consequences of physical destruction (National Research Council, 1999). They include business interruption for those businesses that didn’t sustain direct damages, multiplier effects from reductions in demand and supply, adoption of costly measures as well as mortality, injury and environmental degradation (Kousky, 2013). In practice measuring indirect impacts is difficult. Higher-order losses cannot be readily verified as direct ones and modeling them requires sophisticated economic models. Furthermore their size varies considerably depending on the resiliency of the economy and the path of recovery. Finally caution is required since these effects can be manipulated for political purposes (e.g. inflating multipliers) (Rose, 2004). These challenges make the majority of scholars opt for another approach: instead of trying to estimate direct and indirect costs, they evaluate the impact of natural disasters on macroeconomic indicators, primarily GDP and GDP growth. Macroeconomic variables are often used as a proxy of both kinds of impacts, relying on the fact that direct and indirect impacts could be large enough to have macroeconomic consequences. There are only few studies in the literature that adopt a microeconomic approach and use disaggregated data. Within the literature on the economic impacts of natural disasters it is also possible to distinguish between studies looking at the shortto medium-term, defined as one to five years post disaster, and studies adopting a long-run perspective, beyond five years, with the bulk of the empirical evidence focusing on the short-run. 1.1 Macroeconomic studies Macroeconomic theory does not have a unique answer to whether and how natural disasters affect economic growth: depending on the theoretical framework, different conclusions can be reached. Neo-classical growth models predict that the (physical and/or human) capital destruction may boost short-term economic growth. Endogenous growth models, instead, lead to different and less clear-cut conclusions: AK growth models with constant returns do not predict any change in the growth rate after a negative capital shock; models based on increasing returns predict that the capital destruction leads to a lower growth rate and within the Schumpeterian creative destruction’s framework, negative shocks due to natural calamities may provide opportunities to upgrade obsolete capital and adopt new technologies, thus leading to 6 higher growth. Therefore the question on whether disaster events have positive, negative or no consequences is an empirical one. The majority of macroeconomic studies regresses a macroeconomic variable, primarily GDP and GDP growth, on measures of disaster occurrence and intensity, generally using damages and fatalities as proxies for disaster’s magnitude. These studies usually consider the disaster measure as exogenous, an approach that could be undermined by some problems of reverse causation since the impact of a natural disaster could also depend on the economic conditions of a country. For instance, the levels of income and development can determine the magnitude of the impact of natural calamities. Furthermore, political factors could influence macroeconomic conditions as well as the effects of natural disasters and the quality of institutions might shape both economic performances and the consequences of natural calamities. Most studies use data from the EM-DAT database maintained by CRED. This is primarily due to the fact that the dataset is publically available, but it also ensures consistency with respect to the data used and the definition of natural disaster among different studies. Finally, some of these papers adopt a cross-country approach, while others focus on a single country. 1.1.1 Multi-country studies The bulk of empirical macroeconomic studies adopt a short-run perspective, while only few papers focus explicitly on the long-run effects of natural disasters. Short-run focus The first, recent attempt to describe the short-run macroeconomic dynamics following natural disasters is Albala-Bertrand (1993). Using a sample of 28 disasters in 26 countries between 1960 and 1979 this study estimates the impact of natural calamities on GDP, GDP growth and rate of inflation by means of a simple before-and-after analysis. The results indicate that disasters do not impact GDP and may have a slightly positive impact on GDP growth; no effect is found on the rate of inflation. The author concludes that natural disasters are a problem of development rather than for development. In an unpublished study Caselli and Malhotra (2004) fail to reject the hypothesis that losses of labor and capital stock have no effect on short-term economic growth. Using a dataset of 172 countries for events between 1974 and 1996 they estimate an equation using the difference in the log of output as dependent variable and include several controls and country and year fixed effects. When they include a dummy variable for the occurrence of natural disasters they don’t find any significant effect. Rasmussen (2004), in a statistical comparison among the Eastern Caribbean Currency Unions countries (ECCU), identifies a median reduction of the real GDP growth of 2.2 percentage points in the year of the event. Raddatz (2007) quantifies the impact of a broad set of external shocks (natural disasters, price fluctuations, the role of the international economy) on the output of 40 low-income countries over the period 1965-1997. The author uses a panel vector auto-regression approach and assumes the shocks, including disaster occurrence, are exogenous. He finds that climatic disasters result in a 2% decline in real GDP one year after the event. However, overall external shocks explain only a small fraction of output variance (11%) and climatic disasters are only the third 7 organizational structure. From this database we selected data on performance for the manufacturing firms located in Veneto. The dataset obtained for the empirical analysis is a panel of 2084 companies observed over a period of 10 years, from 2003 to 2012. The AIDA and ARPAV databases are merged using the information about the geographic location of firms and weather stations. More precisely, using latitude and longitude, each firm in the sample is matched to the nearest weather station among the 193 positioned in the region10. The average and the median distance between each firm in the sample and the matched weather station are 4.17km and 2.16km respectively, while the minimum and the maximum distance are 0.13km and 12.60km. This allows to determine the quantity of precipitation received by each firm during the observed period (28/10/2010 - 11/11/2010) with a high degree of precision. Furthermore, this information is crucial to define the treatment and the control group, using a more objective criterion than the amount of reported losses or government reimbursements. 2.1 Precipitations Table 1 reports mean, median, standard deviation, minimum and maximum precipitations for each of the 15 observed days, for the cumulative and for the maximum daily intensity. The Table 1: Precipitations Mean Median SD Min Max precipitations 28/10 .9820513 0 12.21945 0 257 precipitations 29/10 .1620513 0 1.608493 0 32 precipitations 30/10 .1592821 0 1.423357 0 18.2 precipitations 31/10 48.45005 36.6 34.59999 0 218.9 precipitations 1/11 52.91585 41.7 36.36454 0 236.4 precipitations 2/11 31.53528 28 16.59654 0 90.4 precipitations 3/11 .3746154 .2 .8062778 0 8.6 precipitations 4/11 .1043077 0 .1281831 0 .6 precipitations 5/11 .148718 .2 .1518685 0 .8 precipitations 6/11 .1292308 0 .3249769 0 2 precipitations 7/11 7.166257 6.2 4.908876 0 38.2 precipitations 8/11 9.364359 9.4 3.415285 0 25.4 precipitations 9/11 11.43385 11.6 4.018876 0 25 precipitations 10/11 9.769128 8.6 5.795597 0 36.2 precipitations 11/11 .0133333 0 .0499016 0 .2 precipit (flood days) 132.9012 100.4 82.55859 0 479.4 total precipitations 28/10-11/11 172.7084 136 93.79491 0 591 maximum daily precipitations 56.27046 43.3 37.81218 0 257 Notes: Precipitations are measured in mm. Precipit. (flood days) indicates the total quantity of rain fallen during the three flooding days (October 31st - November 2nd). Maximum daily precipitation measures the intensity of precipitation on a daily basis. It is calculated as the total quantity of rainfall fallen on the day when it rained most. table shows that the extreme precipitations have been concentrated during three days, from October 31st to November 2nd. The increase in the quantity of rainfall during these days is enough to lead to a considerable rise in the mean and median precipitations. The increase in 10For 16 companies in the sample the data on latitude and longitude are not available. In this case the match is not possible and these observation are dropped. 14 the variability of rainfall during the flooding days can also be noticed, suggesting that only some areas have been hit by the flood. Table 2 reports summary statistics for the total precipitations and maximum daily intensity in each province. As shown in the table, Belluno, Vicenza, Treviso and Padova are the provinces Table 2: Precipitations by Province Mean Median SD Min Max Belluno total precipitations 28/10-11/11 27.44778 0 89.48925 0 591 maximum daily precipitations 11.09037 0 36.52741 0 257 Padova total precipitations 28/10-11/11 11.31248 0 38.29337 0 543.4 maximum daily precipitations 2.971836 0 10.29553 0 143.6 Rovigo total precipitations 28/10-11/11 4.963375 0 14.92984 0 63.2 maximum daily precipitations 1.209054 0 3.737368 0 22.4 Treviso total precipitations 28/10-11/11 19.41069 0 65.17197 0 543.4 maximum daily precipitations 7.133831 0 24.68617 0 164.2 Venezia total precipitations 28/10-11/11 8.859114 0 27.66879 0 124 maximum daily precipitations 2.202785 0 6.876166 0 33.6 Verona total precipitations 28/10-11/11 11.11059 0 35.87179 0 340.8 maximum daily precipitations 3.142821 0 10.45234 0 119.2 Vicenza total precipitations 28/10-11/11 23.1024 0 75.31425 0 554 maximum daily precipitations 7.525674 0 25.15913 0 236.4 Total total precipitations 28/10-11/11 16.81384 0 58.9694 0 591 maximum daily precipitations 5.478153 0 20.43013 0 257 Notes: Precipitations are measured in mm. Maximum daily precipitation measures the intensity of precipitation on a daily basis. It is calculated as the total quantity of rainfall fallen on the day when it rained most. where it rained most. However, the large standard deviations in these provinces indicate that extreme precipitations didn’t strike homogeneously, suggesting that only some areas have been severely affected. The large variability of precipitations both between and within provinces in the region suggests that there is scope for an analysis of the floods effects in the affected areas and indicates a difference in difference methodology as the most natural approach. 2.2 Firms Table 3 shows the number of manufacturing firms located in each province of Veneto. Vicenza and Treviso have the highest concentration of firms, with 54.3% of the regional firms located in these two provinces. Table 4 displays the number of firms located in flooded and non-flooded areas. Slightly more 15 Table 3: Number of firms by province Province Number of firms Percent Belluno 519.00 2.59 Padova 3437.00 17.16 Rovigo 486.00 2.43 Treviso 5090.00 25.41 Venezia 1580.00 7.89 Verona 3134.00 15.65 Vicenza 5784.00 28.88 Total 20030.00 100.00 than 25% of firms in the region operate in areas that have been exposed to the flood, according to the criterion used to distinguish between flooded and non-flooded areas11. Table 4: Number of firms in flooded and non-flooded areas Treatment Number of firms Percent No-Flood 14933.00 74.55 Flood 5097.00 25.45 Total 20030.00 100.00 Table 5 reports the number of affected and non-affected firms within each province. Vicenza and Treviso are the areas where the highest number of companies has been exposed to the event; here the flood has directly hit 3500 and 1135 companies respectively, corresponding to 60.51% and 22.3% of the provinces’ firms. Moreover, as Table 3 shows, with 28.9% and 25.4% respectively, these provinces have the highest concentration of manufacturing firms. On the contrary, in the Belluno province the flood has affected the majority of firms, with 73.6% of firms located in flooded areas, but only 2.6% of the regional firms are in this province. Table 6 shows the number of observations in the sample before and after the flood. Table 5: Number of firms in flooded and non-flooded areas by province Province No-Flood Flood Total No-Flood Flood Total No. of firms No. of firms No. of firms Percent Percent Percent Belluno 137.00 382.00 519.00 26.40 73.60 100.00 Padova 3397.00 40.00 3437.00 98.84 1.16 100.00 Rovigo 486.00 0.00 486.00 100.00 0.00 100.00 Treviso 3955.00 1135.00 5090.00 77.70 22.30 100.00 Venezia 1580.00 0.00 1580.00 100.00 0.00 100.00 Verona 3094.00 40.00 3134.00 98.72 1.28 100.00 Vicenza 2284.00 3500.00 5784.00 39.49 60.51 100.00 Total 14933.00 5097.00 20030.00 74.55 25.45 100.00 Table 7 displays the number of observations in the treatment and in the control group before and after the natural disaster. 11See empirical strategy. 16 Table 6: Number of observations before and after the flood Pre/Post-Flood Number of obsr Percent Pre-Flood 16103.0 80.4 Post-Flood 3927.0 19.6 Total 20030.0 100.0 Notes: the pre-flood period comprises the years from 2003 to 2010. The post-flood period refers to 2011 and 2012. Table 7: Number of observations before and after the flood by treatment Treatment/Period Number of obs Percent 0/0 12001.00 59.92 0/1 2932.00 14.64 1/0 4102.00 20.48 1/1 995.00 4.97 Total 20030.00 100.00 Notes: for the variable Treatment/Period the first number refers to the treatment or control group, the second one to the time period before or after the flood: 0/0 denotes an observation belonging to the control group before the flood; 0/1 indicates to an observation in the same group after the event; 1/0 is an observation belonging to the treatment group before the flood; 1/1 refers to an observation in the same group after the flood. 2.3 Main variables Table 8 reports mean, median, standard deviation, minimum and maximum of the main variables, distinguishing between the groups of affected and non-affected firms and between the period before and after the disaster event. Considering the pre-flood period, firms located in the non-flooded areas are on average slightly larger both in terms of total assets12 and in terms of employees. However, the median firms in the two groups are more similar to each other with respect to both variables. Looking more specifically at the productive assets, Table 8 shows that before the event the sum of tangible13 and intangible14 assets (Assets) is on average slightly larger for the control group, although this small difference disappears considering the median assets. In terms of tangible fixed asset the average firm operating in a non-flooded area is slightly larger. On the contrary intangible assets are on average higher for affected firms. However, in both cases the difference is negligible when looking at a company with median characteristics. The share of intangible assets, calculated as the ratio between intangible fixed assets and the sum of tangible and intangible fixed assets, is similar in both groups, both in average and median terms. Summary statistics of value added indicate that companies in areas hit by the flood are slightly more productive on average, although the difference shrinks when the median productivity is considered. In terms of value added per employee firms in non-flooded areas have better performances on average but the difference is very small. The median values are slightly larger for 12Total assets are given by total receivables due to shareholders, total fixed assets (total tangible fixed assets, total intangible fixed assets and total financial fixed assets), total current assets (total inventories, total receivables, total financial assets and total liquid funds) and total accrued income and prepaid expenses. 13Tangible fixed assets comprise: land and buildings; plant and machinery; industrial and commercial equipment; other assets; addition in progress and advances and depreciation provision. 14Intangible fixed assets comprise: start-up and expansion costs; research and development expenses; industrial patents and intellectual property rights; concessions, licenses, trademarks and similar rights; goodwill; additions in progress and advances; others and amortization provisions. 17 non-affected companies. Inspection of maximum values of total assets, assets and value added also shows that there are Table 8: Summary statistics of the main variables Value Added Observations Mean Median SD Min Max 0/0 12001 4748.667 2147.323 14776.67 .001 1013700 0/1 2932 4773.553 2113.94 11633.3 .001 215513.1 1/0 4102 4297.594 1980.458 9168.966 .9842107 164020.1 1/1 995 4288.486 1883.663 9125.132 .001 138720.8 Value Added per employee Observations Mean Median SD Min Max 0/0 12001 56.6238 49.27266 32.95921 .0005 318.8578 0/1 2932 54.963 48.59666 32.18068 .0000175 311.1376 1/0 4102 56.03332 45.90213 36.95243 .3457883 315.6109 1/1 995 54.59068 47.11374 33.40911 .001 255.1976 Total Assets Observations Mean Median SD Min Max 0/0 12001 18976.82 7863.719 51776.86 1.042528 2882645 0/1 2932 21205.91 8733.313 51853.36 20.74036 718554.6 1/0 4102 16991.23 7748.512 40459.45 4.17011 773372.5 1/1 995 19020.05 8395.337 49638.1 9.750721 838585.1 Assets Observations Mean Median SD Min Max 0/0 12001 4697.55 1496.821 13120.88 0 236026.2 0/1 2932 5364.527 1902.008 13414.12 0 287949.4 1/0 4102 4473.274 1478.418 13189.07 0 290376.3 1/1 995 4999.918 1903.163 13449.95 0 262336.7 Tangible fixed Assets Observations Mean Median SD Min Max 0/0 12001 4056.557 1308.548 10976.28 0 233782.6 0/1 2932 4725.935 1642.332 10870.6 0 177505 1/0 4102 3604.147 1297.493 6669.398 0 117540.1 1/1 995 4272.846 1744.896 7936.434 0 108890.5 18 Intangible fixed Assets Observations Mean Median SD Min Max 0/0 12001 640.9938 46.83421 5460.209 0 207259.7 0/1 2932 638.5918 46.89125 6064.582 0 245500.1 1/0 4102 869.1263 46.95687 9352.768 0 269653 1/1 995 727.0718 44.18599 6874.282 0 153446.2 Employees Observations Mean Median SD Min Max 0/0 12001 78.85668 43 192.1752 1 6299 0/1 2932 79.80014 42 201.0776 1 6162 1/0 4102 70.55583 41 103.158 1 2044 1/1 995 68.95678 41 98.8198 1 853 Share of Intangible Assets Observations Mean Median SD Min Max 0/0 11979 .1217324 .0346667 .1913689 0 1 0/1 2921 .1204735 .0277778 .1992515 0 1 1/0 4087 .1178565 .0327869 .188799 0 1 1/1 985 .1077548 .0238298 .1861278 0 1 Notes: Value added, value added per employee, total assets, assets, tangible fixed assets and intangible fixed assets are measured in thousands of Euro. Total assets are given by total receivables due to shareholders, total fixed assets (total tangible fixed assets, total intangible fixed assets and total financial fixed assets), total current assets (total inventories, total receivables, total financial assets and total liquid funds) and total accrued income and prepaid expenses. The variable Assets is the sum of tangible and intangible fixed assets. The share on intangible assets (SIA) is the ratio between total intangible fixed assets and Assets. Treatment/Period is defined as follows: the first number refers to the treatment or control group, the second one to the time period before or after the flood. 0/0 denotes an observation belonging to the control group before the flood; 0/1 indicates to an observation in the same group after the event; 1/0 is an observation belonging to the treatment group before the flood; 1/1 refers to an observation in the same group after the flood. some very large companies in the sample, especially in the group of non-affected firms. The summary statistics discussed above show that treated and non-treated firms have similar assets and performance characteristics in the period before the event and thus the flood treatment can be considered as randomly assigned to companies. Comparing the pre-and-post-flood period, Table 8 shows that value added and value added per employee are on average lower after the flood, both for affected and non-affected firms. The same is true when median values are considered. The average firm reduced its number of employees in the post-flood period in both groups of companies. In median terms, instead, the number of employees decreases only for non-treated companies while it stays constant for affected firms. A very slight decrease can also be noticed in the share of intangible assets in both groups. Unlike the former variables, total assets, assets and tangible fixed assets show a different pattern when comparing pre-and-post-flood period: for both groups of companies and both in average and median terms, total assets are higher after the flood than before. Intangible assets, instead, are lower after the flood, both for the treated and the control group. To further investigate the average differences between affected and non-affected firms before and after the occurrence of the flood, a mean difference test has been performed for the main 19 Table 9: Mean difference test for the main variables in the preand post-flood periods Mean difference test (pre-flood) mean diff two-sided p one-sided pl one-sided pu mean control mean treat VA 222.9553 .5485833 .7257083 .2742917 4676.169 4453.213 VAempl -.7137525 .5949733 .2974867 .7025133 55.22527 55.93902 TA 1679.618 .3076885 .8461558 .1538442 20082.02 18402.41 Assets 154.0296 .7681441 .6159279 .3840721 5421.03 5267.001 TtangFA 396.5686 .2167994 .8916003 .1083997 4777.235 4380.666 TintFA -242.539 .4393105 .2196553 .7803447 643.7955 886.3345 SIA .0008743 .9017071 .5491464 .4508536 .1155676 .1146933 employees 8.908197 .0613505 .9693247 .0306753 80.37559 71.46739 Mean difference test (post-flood) mean diff two-sided p one-sided pl one-sided pu mean control mean treat VA 485.0674 .1783971 .9108015 .0891985 4773.553 4288.486 VAempl .3723243 .75921 .620395 .379605 54.963 54.59068 TA 2185.862 .2355402 .8822299 .1177701 21205.91 19020.05 Assets 364.6094 .4597822 .7701089 .2298911 5364.527 4999.918 TtangFA 453.0893 .1593728 .9203136 .0796864 4725.935 4272.846 TintFA -88.47994 .7180696 .3590348 .6409652 638.5918 727.0718 SIA .0127186 .0687175 .9656412 .0343588 .1204735 .1077548 employees 10.84335 .0256883 .9871559 .0128441 79.80014 68.95678 Notes: Value added (VA), value added per employee (VAempl), total assets (TA), assets (Assets), tangible fixed assets (TtangFA) and intangible fixed assets (TintFA) are measured in thousands of Euro. Total assets are given by total receivables due to shareholders, total fixed assets (total tangible fixed assets, total intangible fixed assets and total financial fixed assets), total current assets (total inventories, total receivables, total financial assets and total liquid funds) and total accrued income and prepaid expenses. The variable Assets is the sum of tangible and intangible fixed assets. The share on intangible assets (SIA) is the ratio between total intangible fixed assets and Assets. From left to right, column titles denote mean difference, two-sided p value, lower one-sided p value, upper one-sided p value, mean in control and mean in treatment group. In this table two years are considered for both the pre-flood and the post-flood periods: the pre-flood period comprises the years 2008 and 2009; the post-flood period includes 2011 and 2012. The year in which the flood occurred, 2010, has not been considered, due to the fact that the flood struck at the beginning of November and the its effects are observed from 2011 onwards. Figure 5: Time trends of the main variables (a) Average value added (b) Median value added 20 (c) Average value added per employee (d) Median value added per employee (e) Average value added growth (f) Median value added growth (g) Average total assets (h) Median total assets (i) Average assets (j) Median assets 21 (k) Average tangible fixed assets (l) Median tangible fixed assets (m) Average number of employees (n) Median number of employees Notes: Value added, value added per employee, total assets, assets and tangible assets are measured in thousands of Euro. The variable Assets is the sum of total tangible and intangible fixed assets. variables. In order to have a balanced comparison, the same number of years has been considered for the preand post-period15. Results are reported in Table 9. Before the flood no significant differences can be identified between firms exposed to the flood and the control group of companies in the region. The only exception is the number of employees, for which there is a 6% significant difference between the mean values in the two groups of firms. This difference remains, with a stronger significance, in the post-flood period. After the flood, value added is on average lower for firms located in affected areas; this difference is significant at a 9% significance level. The tangible assets are also lower on average for the treated group (at 8% significance level). Significant differences in the aftermath of the flood can also be identified in the share of intangible assets, with affected firms having on average a lower share. The mean difference test confirms that before the occurrence of the flood, firms located in flooded and non-flooded areas had similar characteristics, so that the treatment can be considered as randomly assigned to companies. After the flood has occurred some statistically significant differences can be noticed, which gives reasons to perform a more formal analysis on the impact of the flood in the region. The graphs in Figure 5 provide a graphical representation of differences between the treatment 15The pre-flood period comprises the years 2008 and 2009; the post-flood period includes 2011 and 2012. The year in which the flood occurred, 2010, has not been considered, due to the fact that the flood struck at the beginning of November and the its effects are observed from 2011 onwards. 22 and the control group of firms, showing the evolution of input factors and value added for the treated and the control group in the sample period 2003-2012. For each variable both the average and the median values are plotted16. The graphical trend analysis shows that the variables considered have similar patterns in the pre-flood period and thus there are no apparent reasons to suspect that some firms have worse or better performances in the aftermath of the event for causes different from the randomness of the flood. This is a further confirmation that firms have been randomly assigned to the flood treatment. 3 Empirical strategy 3.1 Estimation procedure In the empirical analysis we compare the value added growth of companies exposed to the 2010 flood to the one of a control group of unaffected firms up two years after the event. We distinguish between affected and non-affected companies on the basis of their location in flooded and non-flooded areas. The criterion to decide whether an area has been hit by the flood or not is the quantity of rainfall17. The distribution of the precipitation during the observed period and during the three flooding days suggests a cutoff point at around 70% of the distribution when the entire period is considered and at around 75% of the distribution when the analysis is restricted to the proper flooding days (see Fig. 9 and Fig. 4). A firm is considered as affected by the flood if it is located in an area belonging to the highest 30% of the entire period distribution or to the top 25% of the fooding days distribution. According to the distribution the threshold values corresponding to the 70th and 75th percentile are 208.4mm of rainfall fallen during the observed period and 181.6mm fallen during the three flooding days. These threshold values correspond to a quantity of rain which is more than eleven times the amount that falls on an average rainy day18, at least one and a half times the median value of the precipitations’ distribution for the entire observed period and at least twice the median quantity fallen during the three flooding days in the region. The impact of the 2010 flood on firms’ performance is estimated following a difference in differences (DID) approach. Firms located in flooded areas represent the treatment group, companies operating in areas where extreme precipitations didn’t strike constitute the control group. The trend analysis of the main variables and the mean difference tests discussed above confirm that there are no reasons to suspect that differences in firms’ performance after the natural disas16As Table 8 shows there are some very big firms in the sample, therefore it is important to consider both average and median characteristics. 17We recognize that the quantity of rainfall alone is not the most accurate measure to determine if an area is flooded or not. Indeed, beyond intense and/or long lasting precipitations other factors, such as snow/ice melt, the water levels in the rivers and the soil status and characteristics, are also important (Field et al., 2012, p. 175). However, we think that a physical measure of precipitations is a more objective criterion than the use of damage reports. As Yang (2008) and Rose (2004) point out, the prospect of aid may create incentives to inflate losses causing the estimation of the outcome of interest to be biased. Therefore, taking into account our data availability, we preferred to use the quantity of rainfall as the best available criterion to distinguish between flooded and non-flooded areas. 1810 times when only the three flooding days are considered. 23 group of firms29, while firms hit by the flood that didn’t benefit from the financial aid grow at a 5.5% higher that the reference group. This result indicates that, even without financial aid, firms are still able to recover and grow faster than unaffected companies, but that there is a 2% additional effect on growth that is due to the aid. However, we fail to reject the hypothesis that the estimated coefficient of the dummy denoting hit companies that benefited from aid flows (G3 2012) and the one of those that didn’t (G1 2012) are statistically different. Finally, receiving free financial aid without having been exposed to the flood doesn’t have any significant impact on the value added growth of this group of companies (G2). Both in 2011 and in 2012 the estimated coefficient of G2 is found to be not significant, nor important in magnitude. This analysis showed that the role of aid flows in the aftermath of natural disasters is an important factor to consider. Future research should include financial aid in the estimation of the economic impact of natural disasters, although acquiring information about national and international aid flows is often difficult. More research needs to be done in order to improve our knowledge of the true usefulness of emergency and recovery aid. Conclusion In this empirical analysis we estimated the impact of a major flood that hit the region of Veneto in 2010 on firms’ performance up to two years after the event. Using information about geographical locations based on latitude and longitude, we matched each company in the region to the nearest weather station and distinguished between affected and non-affected firms on the basis of the quantity of precipitations received by each firm in the sample. Using firm level data and a difference in differences approach we compared the value added growth of the treatment group to the one of a control group of firms that haven’t been hit by the flood. Our results indicate that the flood had a positive impact on the treated firms, whose value added growth was 6.9% higher two years after the flood struck. The estimation also shows that the positive impact was not experienced in the immediate aftermath of the disaster, but two years after the event. We also found that the assets structure had a positive impact on the value added growth of firms exposed to the flood: a positive change in the share of intangible assets increased value added growth of treated companies. We further investigated the role of aid transfers after the flood for the faster growth of firms exposed to extreme precipitations. Using the information provided by the commissarial orders with which the Italian government established the municipalities eligible for emergency and recovery aid, we identified the firms that benefited from financial flows and constructed four mutually exclusive and exhaustive groups, which allowed us to distinguish between firms that received financial aid and those that didn’t, both in the treatment and in the control group. Using again a difference in difference procedure, our results indicate that, among firms exposed to the flood, both the ones that benefited from financial aid and the ones that didn’t grow faster 29In the estimated model, the reference group is group 4, indicating those companies that neither experienced extreme precipitations nor received financial aid. 30 than the reference group of firms that neither have been exposed to the flood, nor received financial aid. We also found that there is a 2% additional growth effect that is attributable to the contribution of aid in the recovery phase. This analysis showed that possible negative effects of the flood were short-lived and already two years after the event the companies exposed to the flood experienced a recovery. Furthermore, the financial aid after the natural disaster contributed to this outcome, with an additional growth effect experienced by firms that benefited from aid flows. In this analysis we explicitly addressed the role of aid flows in the recovery phase after a natural disaster and answered the question of what would have been the pattern of the value added growth of affected firms after the disaster in the absence of aid flows in our study case. However, the role of national and international aid flows in the aftermath of natural catastrophes has been rarely addressed so far and it is in need of further investigation. 31 References Albala-Bertrand, J. M. (1993). Political Economy of Large Natural Disasters: Clarendon Press, Oxford. Anttila-Hughes, J. K., & Hsiang, S. M. (2011). Destruction, disinvestment, and death: Economic and human losses following environmental disaster. Working paper. Columbia University. Arellano, M. (1987). PRACTITIONERS CORNER: Computing Robust Standard Errors for Within-groups Estimators*. Oxford Bulletin of Economics and Statistics, 49(4), 431-434. doi: 10.1111/j.1468-0084.1987.mp49004006.x ARPAV. Scheda evento ”pluvio”: 31 ottobre - 2 novembre: ARPAV Regione Veneto. Barone, G., & Mocetti, S. (2014). Natural disasters, growth and institutions: a tale of two earthquakes. Temi di discussione Working Paper 949. Banca d’Italia. Barrios, T., Diamond, R., Imbens, G. W., & Kolesr, M. (2012). Clustering, Spatial Correlations, and Randomization Inference. Journal of the American Statistical Association, 107(498), 578-591. doi: 10.1080/01621459.2012.682524 Bertrand, M., Duflo, E., & Mullainathan, S. (2004). How much should we trust differencesin-differences estimates? The Quarterly Journal of Economics, 119(1), 249-275. Cameron, A. C., Gelbach, J. B., & Miller, D. L. (2008). Bootstrap-Based Improvements for Inference with Clustered Errors. Review of Economics and Statistics, 90(3), 414-427. doi: 10.1162/rest.90.3.414 Cameron, A. C., & Miller, D. L. (2010). Robust inference with clustered data. Handbook of empirical economics and finance, 1-28. Caselli, M., & Malhotra, P. (2004). Natural disasters and growth: from thought experiment to natural experiment. Unpublished paper. Cavallo, E., Galiani, S., Noy, I., & Pantano, J. (2010). Catastrophic natural disasters and economic growth. Crespo Cuaresma, J., Hlouskova, J., & Obersteiner, M. (2008). Natural disasters as creative destruction? Evidence from developing countries. Economic Inquiry, 46(2), 214-226. Cu˜nado, J., & Ferreira, S. (2011). The Macroeconomic Impacts of Natural Disasters: New Evidence from Floods. Paper presented at the Agricultural and Applied Economics Associations 2011 AAEA & NAREA Joint Annual Meeting. Pittsburg, PA. Philadelphia, PA: Center for Risk Management and Decision Processes, The Wharton School, University of Pennsylvania. EEA. (2012). Climate change, impacts and vulnerability in Europe 2012 Report No.12/2012 (pp. 300). Copenhagen, Denmark: European Environment Agency. Field, C. B., Barros, V., Stocker, T. F., & Dahe, Q. (2012). Managing the Risks of Extreme 32 Events and Disasters to Advance Climate Change Adaptation: Special Report of the Intergovernmental Panel on Climate Change: Cambridge University Press. Fomby, T., Ikeda, Y., & Loayza, N. V. (2013). The growth aftermath of natural disasters. Journal of Applied Econometrics, 28(3), 412-434. doi: 10.1002/jae.1273 Hammes, D., & Vu, T. B. (2010). Dustbowls and high water, the economic impact of natural disasters in China. Asia-Pacific Journal of Social Sciences(1), 122-132. Hochrainer, S. (2009). Assessing the macroeconomic impacts of natural disasters: are there any?. Policy Research Working Paper 4968. The World Bank, Sustainable Development Network Vice Presidency. Global Facility for Disaster Reduction and Recovery Unit, Washington DC. Kousky, C. (2013). Informing climate adaptation: A review of the economic costs of natural disasters. Energy Economics(0). doi: http://dx.doi.org/10.1016/j.eneco.2013.09.029 Leiter, A. M., Oberhofer, H., & Raschky, P. A. (2009). Creative Disasters? Flooding Effects on Capital, Labour and Productivity Within European Firms. Environmental and Resource Economics, 43(3), 333-350. doi: 10.1007/s10640-009-9273-9 Liang, K.-Y., & Zeger, S. L. (1986). Longitudinal Data Analysis Using Generalized Linear Models. Biometrika, 73(1), 13-22. doi: 10.2307/2336267 Loayza, N. V., Olaberra, E., Rigolini, J., & Christiaensen, L. (2009). Natural disasters and growth: going beyond the averages. Policy Research Working Paper 4980. The World Bank, East Asia and Pacific Social Protection Unit & Development Research Group Retrieved from http://econ.worldbank.org National Research Council. (1999). The Impacts of Natural Disasters:A Framework for Loss Estimation: The National Academies Press. Noy, I. (2009). The macroeconomic consequences of disasters. Journal of Development Economics, 88(2), 221-231. doi: http://dx.doi.org/10.1016/j.jdeveco.2008.02.005 Noy, I., & Vu, T. B. (2010). The economics of natural disasters in a developing country: The case of Vietnam. Journal of Asian Economics, 21(4), 345-354. doi: http://dx.doi.org/10.1016/ j.asieco.2010.03.002 Raddatz, C. (2007). Are external shocks responsible for the instability of output in lowincome countries? Journal of Development Economics, 84(1), 155-187. doi: http://dx.doi.org/ 10.1016/j.jdeveco.2006.11.001 Raddatz, C. (2009). The Wrath of God Macroeconomic Costs of Natural Disasters. Policy Research Working Paper 5039. The World Bank, Development Research Group Macroeconomics and Growth Team. Rasmussen, T. N. (2004). Macroeconomic implications of natural disasters in the Caribbean. IMF Working Paper WP/04/224. International Monetary Fund. 33 Rose, A. (2004). Economic Principles, Issues, and Research Priorities in Hazard Loss Estimation. In Y. Okuyama & S. Chang (Eds.), Modeling Spatial and Economic Impacts of Disasters (pp. 13-36): Springer Berlin Heidelberg. Skidmore, M., & Toya, H. (2002). Do natural disasters promote long-run growth? Economic Inquiry, 40(4), 664-687. doi: 10.1093/ei/40.4.664 Strobl, E. (2010). The Economic Growth Impact of Hurricanes: Evidence from U.S. Coastal Counties. Review of Economics and Statistics, 93(2), 575-589. doi: 10.1162/REST a 00082 White, H. (1984). Asymptotic Theory for Econometricians: San Diego: Academic Press. Yang, D. (2008). Coping with disaster: The impact of hurricanes on international financial flows, 1970-2002. The BE Journal of Economic Analysis & Policy, 8(1). 34 Appendix Robustness checks The results of model 3 are robust to a change in the number of years considered, to the exclusion of 2010, the year of the flood, and to the estimation of the same model using OLS instead of fixed effects procedure. The estimated coefficient of the dummy denoting firms exposed to the flood that received financial aid (G3 2012) ranges between 7.3% and 9% in 2012. For companies affected by the flood that didn’t benefit from financial aid, the estimated coefficient in 2012 (G1 2012) ranges between 5.5% and 6.9%. Table 12: Estimates of flood effect on value added growth - Robustness checks (1) (2) (3) (4) (5) (6) VARIABLES VA growth se VA growth se VA growth se DlAssets 0.0534** (0.0223) 0.0565*** (0.0194) 0.117*** (0.0272) Dlempl 0.347*** (0.0551) 0.351*** (0.0663) 0.451*** (0.0407) DShare -0.168 (0.120) -0.0351 (0.0676) -0.116* (0.0667) G1 2011 -0.0558 (0.0392) -0.0398 (0.0379) -0.0575 (0.0353) G1 2012 0.0548** (0.0273) 0.0694** (0.0305) 0.0596** (0.0261) G2 2011 0.0269 (0.0253) 0.0302 (0.0261) 0.0117 (0.0199) G2 2012 0.0145 (0.0242) 0.0145 (0.0279) 0.00511 (0.0221) G3 2011 -0.0183 (0.0376) -0.00975 (0.0356) -0.0266 (0.0324) G3 2012 0.0863** (0.0395) 0.0908** (0.0409) 0.0735** (0.0366) G1 2011 DShare 1.097** (0.456) 0.886* (0.466) 1.115*** (0.356) G1 2012 DShare 0.286 (0.302) 0.153 (0.310) 0.203 (0.238) G2 2011 DShare 0.230 (0.495) -0.0500 (0.420) 0.0414 (0.387) G2 2012 DShare 0.441** (0.221) 0.417** (0.205) 0.333** (0.163) G3 2011 DShare 0.517 (0.607) 0.328 (0.657) 0.524 (0.482) G3 2012 DShare 0.684 (0.740) 0.592 (0.737) 0.341 (0.625) Constant -0.0635*** (0.00894) -0.0637*** (0.00759) 0.0117*** (0.00368) Observations 9,674 7,770 17,724 Number of firms 2,026 2,025 Adjusted R-squared 0.093 0.085 0.134 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Notes: model 1 uses fixed effects for the period 2008-2012; model 2 uses fixed effects for the period 2008-2012, excluding 2010; model three uses OLS for the entire period considered (2003-2012). The three estimated models also contain year-time dummies and interaction variables between treatment dummy (G1, G2 G3), year-time dummies and the input variables (DlAssets and Dlempl). In the three models, G1 denotes firms that experienced extreme precipitations but didn’t receive aids; G2 indicates firms that haven’t been hit by extreme precipitations but received financial aid; G3 are firms hit by extreme precipitations that received financial aid after the flood; G4 is the reference group and denotes firms that haven’t been hit by extreme precipitation and didn’t receive aid flows. In each model cluster-robust standard errors are used; the cluster unit is comune. The flood 35 Figure 6: Precipitations during the first flooding day. The figure shows the quantity of precipitation (in mm) registered by each weather station in the first flooding day. The ‘mean by weather station’ (green line) represents the average quantity of rainfall fallen on a raining day in October and November during the observed period (20032012) excluding 2010. The value has been calculated by taking an average between October and November for each weather station. The data are available only for 163 stations. The ‘overall mean’ (red line) represents an average of the quantity of precipitations on a raining day in October and November calculated as an average among 163 weather stations during the observed period excluding 2010. The data are available only for 163 weather stations. The average value has been extended to the remaining 30 stations. Source: graph made using data from ARPAV. Figure 7: Precipitations during the second flooding day. The figure shows the quantity of precipitation (in mm) registered by each weather station in the second flooding day. The ‘mean by weather station’ (green line) represents the average quantity of rainfall fallen on a raining day in October and November during the observed period (2003-2012) excluding 2010. The value has been calculated by taking an average between October and November for each weather station. The data are available only for 163 stations. The ‘overall mean’ (red line) represents an average of the quantity of precipitations on a raining day in October and November calculated as an average among 163 weather stations during the observed period excluding 2010. The data are available only for 163 weather stations. The average value has been extended to the remaining 30 stations. Source: graph made using data from ARPAV. 36 Figure 8: Precipitations during the third flooding day. The figure shows the quantity of precipitation (in mm) registered by each weather station in the last flooding day. The ‘mean by weather station’ (green line) represents the average quantity of rainfall fallen on a raining day in October and November during the observed period (20032012) excluding 2010. The value has been calculated by taking an average between October and November for each weather station. The data are available only for 163 stations. The ‘overall mean’ (red line) represents an average of the quantity of precipitations on a raining day in October and November calculated as an average among 163 weather stations during the observed period excluding 2010. The data are available only for 163 weather stations. The average value has been extended to the remaining 30 stations. Source: graph made using data from ARPAV. Figure 9: Distribution of precipitations. The figure shows the distribution of precipitations in the Region according to the location of the 193 weather stations in Veneto Region. The entire observation period (October 28th −November 11th) is considered in this figure. Source: graph made using data from ARPAV. 37 Acknowledgments My first and sincere appreciation goes to my supervisor, Paolo Manasse, for his continuous assistance and support in in all stages of this thesis, for his valuable advices and suggestions and for his attention to detail. I would like to thank professor Maria Elena Bontempi and Sarah Grace See for their help in the data preparation in the early stage of this thesis; Leiter Andrea M., Oberhofer Harald and Raschky Paul A., whose research inspired me, for their methodological advices. I must also acknowledge the Regional Environmental Agency of Veneto (ARPAV) for providing me with precious insight into the flood event and for the useful clarifications about their data. Finally, I would like to express my deep gratitude to my family for teaching me the value of education, for giving me the possibility to study and for always trusting my education choices. 38 