Vulnerability to Poverty and Financial Exclusion: The Italian Case

(1)

Vulnerability to Poverty and Financial

Exclusion: The Italian Case

Floris van Willigen florisvw@gmail.com

Department of Economics University of Groningen

(2)

Abstract

Social and economic policy in Italy has focused on giving support to the chronically poor household. Besides poor household Italy has also a significant number of households who are by definition not poor. However, a great share of these non-poor households are at risk of falling in poverty in the event in case idiosyncratic shocks or economic shocks. Such shocks can have serious long term and short term effects on the wellbeing of households in the absence of financial insurance networks. This paper analyzes household vulnerability and financial exclusion drawing upon household finance

(3)

I. Introduction

Keywords: Poverty, Vulnerability, Italy, Financial exclusion

Background/motivation

Poverty is a topic which is mainly associated with developing and rising economies. However, poverty is also present in developed economies and countries. Although poverty may lie under the surface, and for that reason can be hard to pinpoint, individuals can still slide into poverty as it is to be seen in the past. In southern Europe one can observe patterns where individuals slide into poverty. The ‘double dip’ had a devastating effect on functioning of several southern economies and had structural impacts as well. Although some economies have been started to recover again, Southern Europe still faces these structural impacts. This crisis legacy consists of higher inequality, weakened public institutions and infrastructure, social and political dissatisfactions (Zamora-Kapoor & Coller, 2014).

(5)

In such situations financial access is essential for the wellbeing of the described households. Saving accounts or any form of transaction account allow a vulnerable household to smooth its consumption paths when it faces an idiosyncratic shock or a serious change in income stream (Helms, 2010). In the case of Italy this will be especially relevant, since it has one of Europe’s highest rates of financial exclusion (20% vs 17%) (Barboni, Cassar, & Demont, 2017). In the literature it is shown that financial exclusion in Italy for poor households is related to economic and social exclusion (Barboni et al., 2017). In addition, research makes clear that Italia households with financial access reduce the chances of running into financial distress. A poor Italian household which possesses a bank account is well equipped to reduce the chances of financial distress (Brunetti, Giarda, & Torricelli, 2016).

(6)

especially holds for households which are constrained in their resources and finances. Note that in this case research will extend ‘traditional’ analysis of vulnerability to poverty and

sketch the interaction between vulnerability to poverty and financial exclusion. It is a ‘natural’ extension and hopefully this extension captures the amount of risk a

household faces. Nether the less, an approach using VEP and recent data sets will enrich the view on poverty in Europa conspicuously. The aim of this paper is to describe -up to date- vulnerability to poverty of Italian households and to investigate the determinants. To be more specific: Besides looking at sociological, demographic and economic determinants, we will also take a deeper look at the role played by financial exclusion of households. Our research questions can be summarized in the following way:

What are the determinants of the vulnerability to poverty in Italy ?

What is the interaction between financial exclusion and vulnerability to poverty in Italy? Vulnerability to poverty allows us to identify groups of individuals which have greater likelihood of becoming poor. If it is possible to identify the characteristics of these vulnerable individuals, perhaps VEP can serve as a tool for policy making. For instance, it can be very useful in cases of anti-poverty interventions which are aimed to prevent households from falling in poverty (Chaudhuri, Jalan, & Suryahadi, 2002).

II. Related literature

Vulnerability to poverty

There are several definitions concerning Household vulnerability to poverty that I will highlight. You can find two of these definitions below. Firstly, the following authors

(7)

poverty as the probability that a household experiences at least one period of poverty over a given period. A different definition is given by (Chaudhuri et al., 2002), who defines vulnerability to poverty at time t as the probability that a household falls below a certain poverty threshold at time t + 1. So in other words poverty is defined ex-ante and not ex-post, which means that you make prospect about future poverty.

There are different ways to calculate estimates to vulnerability to poverty. In the literature we see usually vulnerability assessments based on panel data (Foster, Greer, econometric, & 1984, n.d.), (Ligon & Schechter, 2003). Most of these approaches require a certain longitudinal richness in the data set. A calculation of our estimates is based on a cross-sectional data set which is adopted from the literature (Chaudhuri et al., 2002). This seems a logical choice since the limited longitude of the data set.

Currently (Amendola et al., n.d.) did some unpublished research, in which they examined the phenomenon of vulnerability to poverty in Italy. In this paper the authors investigated

vulnerability to poverty in Italy and the divergence between north and south. The research was performed on the basis statistics provided by the Italian institute of statistics for the years 1985-2001. The authors found that the group which is vulnerable to poverty is rising in the south of Italy and stable/declining in the north. Furthermore, an interesting finding was that the number of households in chronic poverty declined since 1985. However, the number of vulnerable households started to rise again from 1995.

In Italy the explorative nature of the VEP allows us to pinpoint two different dynamic phenomena: Firstly, chronic poverty, a case wherein households are trapped in poverty and have zero prospect of coming out. Secondly, potential poverty, namely of non-poor

(8)

households remain poor in the future, which allows us to explore the first phenomena. In the second case the VEP incorporates the risk of non-poor households to become poor in the future.

The literature concerning Italy shows that 90% of the households falling below poverty line are chronically poor (Amendola et al., n.d.). This is also described as ‘’hard core’’ poverty, in which a household has little to no hope to escape from poverty. This is also in line with the finding that there exists a low intergenerational mobility in Italy (Piraino, 2007). Secondly, it was estimated that a great share of the non-poor individuals in Italy had a higher than average probability to become poor in the future (Amendola et al., n.d.). Both phenomena make the use of the VEP in Italy relevant. Their vulnerability to poverty analysis is not further extended from 2001 and therefore it would fill a gap in the existing knowledge if the vulnerability to poverty would be estimated with more recent data.

Financial exclusion and poverty (Should be further clarified)

Based on literature, the opinion is possibly right -at least defensible- that financial exclusion plays an important role in the vulnerability to poverty of Italian households. A field

experiment among migrants and low-income household reveals that financial exclusion is associated with poverty and social exclusion. The following authors found in field

(9)

It remains hard to identify only one single explanation for financial exclusion. Despite the ambiguity about possible causes, the literature is clearer in its description of the impact of the access to finance. Financial access is and remains of the highest importance for the poorest households, especially when such households face an unpredictable shock (Ambuehl, Bernheim, Lusardi, & Douglas Bernheim, 2015), (Lusardi, 2008). The lack of financial access for households is not only a problem for the developing world, it also remains a problem in the ‘developed’/high-income countries. Despite the coverage and penetration of the financial sector in the developed world, the rates of financially excluded households are surprisingly high.

In the European context, Italy is a great example. Note that Italy has one of the highest rates of financial exclusion in Europe (Barboni et al., 2017). A major aim of this paper is to

provide an investigation between the risk of households to fall in future poverty and financial exclusion. This goes beyond established link by Barboni (2017). In the paper published by Barboni, the authors established a clear link between the household already in poverty and the restricted financial access of these households. In our paper we will consider a multitude of financial services (ranges from saving account to holding bond/shares) as indicators to answer the question if a household is financially excluded or not. These indicators will be interacted with vulnerability to poverty. In this way the paper will depart from the literature already in place and look further than interaction between households already in poverty and the financial exclusion of these households.

III. Data-description

Data choices

(10)

covering information on balance sheets of household it also covers information on consumption and income. A great advantage of this dataset is that all the surveys are standardized over the European Union. This allows us to compare the data which are available for us.

The fieldwork for the survey was done for the first wave in 2010-2011 and for the second wave between 2013-2015. The results of the third wave are expected to be released

somewhere in 2019. Preferably the vulnerability to poverty would be estimated with panel data of several years. However, taking into consideration that specific survey is performed only twice, in this paper a cross-sectional analysis has been performed, relying on data of the second wave (2013-2015).

The raw dataset was roughly divided in two samples. The first sample compiled all the demographic information on all members of the household. In addition to demographic information, the HFCS provides a second sample containing economic information at household(head) level, including net wealth, net income and consumption. To obtain a consistent merged dataset, we removed the sample observation related to those members of the household who were not head. This leaves 8156 (head of) households for the cross-sectional analysis. Table 2 reports the descriptive statics for the second wave. Table 1 reports a brief description of the variable use in the empirical investigation.

Table 1. Variable description

Variable Description

Lnconsumbasic Logarithm of the compiled total consumption in the household

Age Age of the household head

age squared Square of the age of the house hold head

Male Sex of the head of the household

single/never married Factor variable equals to 1 for all respondents reporting to be single and never married, 0 otherwise

Married Factor variable equals to 2 for all respondents reporting to married

Widowed Factor variable equals to 3 for all respondents reporting to be married and widowed, 0 otherwise

(11)

no EU-immigrant Dummy variable equals to 1 if household head is born outside Italy and outside Europe Union, 0 otherwise

EU-immigrant Dummy variable equals to 1 if household head is born outside Italy and inside Europe Union, 0 otherwise

household size Number of household members

low secondary Factor variable equals to 2 if the respondent highest completed education is lower education or second stage of basic education

secondary school Factor variable equals to 3 if the hh highest completed education is upper secondary school

Graduate Factor variable equals to 5 if the hh highest completed education is first stage tertiary education

Employed Factor variable equals to 1 if labour status of the household head is employed

Self-employed Factor variable equals to 2 if labour status of household head is self-employed

Unemployed Factor variable equals to 3 if labour status of household head is unemployed

Retired Factor variable equals to 4 if labour status of household head is retired

atypical Factor variables equals to 5 if labour status of household head is not clearly defined

temporary contract Dummy variable equals to 1 if head of household is employed on the basis of a temporary contract, 0 otherwise

log income Logarithm of the income of the head household, which takes into account gross income

log wealth Logarithm of the wealth of the household, which takes into account, Net wealth exclusive pension wealth

house ownership Dummy variable equals to 1 if main residence of the household is owned or partly owned, and 0 otherwise

FEone Dummy variable equal to 1 if households do not possess money in checking or savings accounts, money market funds, certificates of deposit, government savings bonds, and 0 otherwise

FEtwo Dummy variable equal to 1 if households do not possess over a credit line/overdraft facility or credit card, and 0 otherwise.

FEthree Dummy variable equal to one if households do not possess publicly traded shares, corporate or government bonds, and investment trusts, and 0 otherwise.

Economic Controls

In accordance with recent literature (C. Giannetti, Madia, & Moretti, 2014), this study defines an economic control based on the type of employment. This is done by estimating 4 exclusive job categories (employed, self-employed, unemployed, retired). In addition to this a second indicator (temporary) is defined who identifies those employees, who -despite being

(12)

Further, we define indicators which control for the financial position of the household. One of the variables take into account the certain amount of net wealth acquired in the household, by taking the logarithm of wealth (Paiella, 2007). Moreover, controls are constructed, based on logarithm of household income. In taking logarithm of both variables the literature has been followed and this appeared helpful in dealing with outliers in wealth or income (M. Giannetti, 2011).

Demographic controls

Demographic information is gathered on household gender (‘male’), level of education (basic education, secondary school, graduate), household size. Furthermore, controls for the marital-status of household are added (Brunetti et al., 2016). In addition, dummies are created, intended to control the country of origin of the household head. In the literature it is founded that an immigrant status may be associated with higher poverty rates and higher rates of financial exclusion (Barboni et al., 2017). Hence, two dummies are constructed: one for EU immigrants and another for non-EU immigrants.

Measurement of Financial exclusion

The survey also provides information about financial exclusion of households, which is a key to answer the research question. In accordance with the literature three indicators are

constructed on the different aspects of financial exclusion (Barboni et al., 2017)(Devlin, n.d.)(Mylonidis, Chletsos, & Barbagianni, 2019). The first variable (FEone) captures the aspect of exclusion from saving and related products. According the literature this indicator displays the strict definition of financial inclusion. Households lack access necessary in handling everyday transaction in developed countries (Mylonidis et al., 2019). In the second indicator (FEtwo) the aspect of exclusion from credit is captured. Lastly, in the third indicator (FEthree) exclusion from financial markets is captured. The latter two indicators are

(13)

Descriptive statistics

Table 2 reports the descriptive statistics of variable data, used in the analysis. The majority of household heads is a male (64%) who is married (57.5%) with an average age of 61 years. On average an Italian household counts 2.3 members. Only 2.8 % of the respondence is not born in the European Union. In comparison 3.2% of immigrated household heads is born in the E.U. which makes a total share of household heads which are not born in Italy at 6.0% in the sample. In relation to the highest completed education level of the household, a great share of the respondent completed their secondary school, which is compulsory in Italy. Around 12% percent of household heads completed their University degree. Taking into account the average age which lies above the 60, it comes not as a surprise that a great share of the household heads is retired (38.7%). In the remaining 1 out of 3 household heads is actively employed and 5% of this employment happens on temporary basis. Around 7.1% percent of the household do not possess a checking or saving account. A slightly higher percentage (12.5%) of the households do not possess any form of credit line/overdraft facility. Finally, in 8 of 10 cases the household is not actively involved in a financial market, for example such as a share market or bond market.

Table 2. Summary statistics

Variable Obs Mean Std.Dev. Min Max

Lnconsumbasic 8156 7.401 .5 5.011 9.292

Age 8156 61.139 15.093 17 85

age squared 8156 3965.769 1812.982 289 7225

Male 8156 .644 .479 0 1

Single never married 8156 .142 .349 0 1

(14)

self-employed 8156 .106 .308 0 1 unemployed 8156 .028 .166 0 1 retired 8156 .387 .487 0 1 temporary contract 8156 .049 .215 0 1 log income 8037 10.112 .978 .334 13.553 log wealth 7895 11.503 1.803 3.401 16.498 houseownership 8156 .908 .288 0 1 FEone 8156 .071 .257 0 1 FEtwo 8156 .125 .331 0 1 FEthree 8156 .828 .378 0 1

IV. Empirical model

Empirical Methods

The tool which is used to perform this research is the vulnerability to poverty measurement. Methodological we follow (Chaudhuri et al., 2002) whose method is suitable for a cross-sectional data set. So they start with assuming stochastic process in basic consumption where they correlate a basic consumption basket with a bundle of household characteristics. This process is denoted below in 𝑋_ℎ ′ as representative of the household characteristics, 𝑒_ℎ is an idiosyncratic error factor centered around zero and 𝑐_ℎ is the consumption per capita. Note that household characteristics will be tailored to the country specifics which means that for example that in Italian case we include wealth indicator

𝑙𝑛𝑐_ℎ= 𝑋_ℎ ′_{𝛽 + 𝑒} ℎ

If the stochastic process is generated this allows us to estimate the variance of this specific consumption function. This estimated variance can then be regressed with respect to the family characteristics on function which is denoted below.

(15)

On the basis of three-stage FGLS estimation, the following estimates for 𝛽̂ and 𝜃̂. can be found.(Am, 1977.) Now, by finding 𝛽̂ it is possible to do a probability assessment if this specific household has a chance for becoming poor. Notice that finding vulnerability of a specific household is only possible, if it is assumed that 𝑙𝑛𝑐ℎ is normally distributed, where Φ

denotes the cumulative density function of the assumed normal distribution. The logarithmic poverty line 𝑙𝑛𝑍 is pre determined and it is suggested 60 % of the equivalent median consumption. (“European Commission (2010), Europe 2020. A strategy for Smart, Sustainable and Inclusive Growth)

𝑉̂ = 𝜈̂_ℎ _ℎ = 𝑃𝑟(𝑙𝑛𝑐_ℎ < 𝑙𝑛𝑍) = Φ(𝑙𝑛𝑧ℎ −𝑋ℎ ′𝛽̂

√ 𝑋ℎ ′𝜃̂

)

Based on predicted vulnerability and the vulnerability threshold a dummy variable is established , representing the vulnerability of each individual household. Based on the results of (Zhang & Wan, 2009) , vulnerability threshold is set on the stringent line of 50%. If the threshold is set at 50 % , this basically means that those households with a chance of more than 50% falling into future poverty are the vulnerable ones. The higher you set this bar the lower chance of a household to be identified as a vulnerable one in the future. In the case a household is vulnerable this is indicated by dummy variable equal to 1 and if not vulnerable this is indicated by 0.

𝑉𝑢𝑙ℎ = {

1, 𝑉̂ > 0.5_ℎ 0, 𝑉̂ < 0.5_ℎ

Now it is possible to look at the marginal effect of the household determinants on the predicted future vulnerability for poverty of a household. Based on following regression are estimated. 𝑉𝑢𝑙_ℎ = 𝛼_ℎ+𝐻ℎ_ℎ+ 𝑒_ℎ (1)

(16)

𝑉𝑢𝑙_ℎ = 𝛼_ℎ+𝐻ℎ_ℎ+ 𝐹𝐸𝑡𝑤𝑜_ℎ+ 𝑒_ℎ (3) 𝑉𝑢𝑙_ℎ = 𝛼_ℎ+𝐻ℎ_ℎ+ 𝐹𝐸𝑡ℎ𝑟𝑒𝑒_ℎ+ 𝑒_ℎ (4)

Results and Discussion

Table : 3-Stageleastquares : Cross-sectional report

(1) (2) (3) (4)

VEP FEone FEtwo FEthree

age 0.005*** 0.005*** 0.004*** 0.004*** (0.001) (0.001) (0.001) (0.001) age squared -0.000*** -0.000*** -0.000*** -0.000*** (0.000) (0.000) (0.000) (0.000) male -0.004 -0.004 -0.005 -0.004 (0.005) (0.005) (0.005) (0.005)

(17)

FEone 0.061*** (0.018) FEtwo 0.043*** (0.004) FEthree -0.045*** (0.003) _cons 1.151*** 1.127*** 1.181*** 1.251*** (0.051) (0.051) (0.051) (0.052) Obs. 7808 7808 7808 7808 R-squared 0.360 0.363 0.363 0.365

Standard errors are in parenthesis

*** p<0.01, ** p<0.05, * p<0.1

Cross sectional results: Vulnerability to poverty

As shortly denoted, what are the special characteristics of households which are susceptible for poverty in the future? The results from the demographic and income variables are in some aspects in line with the literature. The older households are less vulnerable. However, note that this effect is diminishing over time. There is no direct evidence that male headed households are less likely to be vulnerable to poverty. With respect to the baseline of single and never married household heads, the individuals are less vulnerable if the respondent is married. In the case of a divorced or widowed head, the probability of falling in poverty rises, compared to an individual who is single and never married. Being immigrant makes you more vulnerable, however in the case the individual is born outside the European Union the effect on the VEP is larger.

From an economic perspective, VEP is decreasing in the logarithm of income and the

(18)

countering effect. This increases the vulnerability to poverty for these individuals at a temporary payroll. Finally, retired individuals are less likely to fall in poverty which is consistent with the finding that older household heads are less vulnerable.

Vulnerability to poverty and financial exclusion

Table 3 reports the results for the variables, which concern the financial exclusion indicators. One can notice that there are no dramatic changes in the coefficient of the baseline regression (1) and the regression (2) (3) (4). The first dependent variable, where the households are excluded from essential financial services, increases significantly the vulnerability of

households to future poverty. Other households, which do not have bank account or checking account, are more likely to fall in poverty. Also there exists a second case, wherein non-usage of credit line/overdraft facilities leads to significant increase in vulnerability to future

poverty. Finally, holding shares or bonds as a household increases the likelihood of being vulnerable in the future. We see namely an opposite picture when a household is active on share and bond markets. This activity makes households in many cases more vulnerable to poverty on a significant basis.

V. Robustness

Model specifications

(19)

In addition to this, literature shows that the log-normal function is superior in modeling the poor household. Given the fact that in this paper we are mainly interested in poverty, the log normality of consumption is a reasonable assumption. To control for the case of

heteroscedasticity in the error terms, the three- stage FGLS stage least square approach is run with robust standard errors. This will solve the problem of the heteroscedasticity in the errors. Endogeneity

One of the variables which is capturing financial exclusion might be endogenously

determined in the model. In other words, there may exist potential correlation between the idiosyncratic component of the error term and financial exclusion. So, in this case there may exist potential correlation between financial exclusion indicators and the idiosyncratic components of the error term. This increases the probability of being vulnerable to future poverty. To solve this issue, one should identify that the determinants of financial exclusion are not correlated with error terms of equation (2) (3) (4). To our judgement finding an instrument which is completely unrelated to the determinants of vulnerability to poverty and highly correlated with financial exclusion, is unrealistic. We thus rely on a different approach in which we use the three indicators on financial exclusion. The main idea is as follows: firstly a condition is placed on one of the indicators (FEone=1), secondly one of the remaining indicators of financial exclusion (FEtwo, FEthree) will be regressed on the vulnerability to poverty indicator. By running such a regression one can break the link between the idiosyncratic component of the error term and the determinants of financial exclusion. Speaking in terms of the variables, a household will be already financially included and this breaks the reverse causal link between VEP and FE.

(20)

Table 4 : VEP and FE indicators: Household being financially included

(1) (2) (3) (4) (5) (6)

FEtwo=0 FEthree=0 FEone=0 FEthree=0 FEone=0 FEtwo=0

FEone 0.056*** (0.018) FEtwo 0.038*** 0.006** (0.004) (0.003) FEthree -0.039*** -0.042*** (0.003) (0.004) _cons 1.240*** 0.272** 1.090*** 0.279*** 1.152*** 1.350*** (0.055) (0.105) (0.050) (0.107) (0.051) (0.056) Obs. 6797 1405 7400 1405 7400 6797 R-squared 0.376 0.214 0.344 0.215 0.346 0.376

Standard errors are in parenthesis

*** p<0.01, ** p<0.05, * p<0.1

The results of financially included households are denoted in table 4 and are regressed with standard robust errors. In the 1th regression, where the household has access to a credit line/overdraft facility, highlights a positive statically significant relationship between the VEP and financial exclusion (FEone). In addition, the 2th regression shows that in a household which has access to a bank account, there also exists a positive relationship between the VEP and financial exclusion (FEtwo). The 4th regression gives additional

evidence that in the case a household has access to the financial market, also in this case there exists a weaker positive significant relationship between the VEP and being excluded from a credit line/overdraft facility. Finally, we come to a household which has credit access or bank account access. In that case the results in regression (5) and (6) suggest that a household which is active on the financial markets is more likely to be vulnerable to poverty. The general interpretations of the coefficients suggest that there are no endogeneity problems related to variables concerning financial exclusion.

(21)

active on financial markets increases the chances of households to become more vulnerable to future poverty.

Limitations and further Directions

All the estimations and results are based on cross-sectional analysis. As already denoted this cross-sectional approach leads to some strong underpinning assumptions, which are likely to be violated. Thus, in order to check if the main results will hold over time, further research in a panel data setting is needed. Another great limitation is the absence of regional variables in the empirical analysis. Since this is an European data-set, information on a regional level is not included. However, as denoted in the literature, there is a great divergence between north and south Italy (Amendola et al., n.d.). Therefore a regional variable should be included as dummy in the empirical analysis. In future research the data should contain information on a regional level. This would sketch a more detailed and complete picture.

VI. Concluding Remarks

This short paper shows that being vulnerable to poverty is positively related to having no saving access/credit access. The model used to predict the vulnerability to poverty assessments is estimated by means of a three-step FGSL procedure. We used a cross-sectional household survey provided by the Banca di Italia, which compiled information on more than 8000 households over the period 2013-2015.

(22)

In the analysis of the poverty dynamics in Italy and the role of financial exclusion, there are three main findings. The results shows; (i) that access to credit devices can help to prevent households from slipping into poverty when such households face idiosyncratic shocks, (ii) access to saving or deposit device reduces the vulnerability to poverty (iii) while access to financial markets in the Italian context not necessarily leads to reduction in proneness to poverty.

(23)

VII. References

Amemiya, T. (1977) ”The maximum likelihood estimator and the non-linear three stage least squares estimator in the general nonlinear simultaneous equation model,”

Econometrica, 45, 955-968.

Ambuehl, S., Bernheim, B. D., Lusardi, A., & Douglas Bernheim, B. (2015). GFLEC Working Paper Series. The Effect of Financial Education on the Quality of Decision Making.

Amendola, N., Rossi, M., Vecchi, G., & Vergata, T. (n.d.). Vulnerability to Poverty in Italy. Barboni, G., Cassar, A., & Demont, T. (2017). Financial exclusion in developed countries: a

field experiment among migrants and low-income people in Italy. In Journal of Behavioral Economics for Policy (Vol. 1).

Barr, M. S., & Blank, R. M. (n.d.). Findings Access to Financial Services, Savings, and Assets Among the Poor Highlights from the forthcoming edited volume, Insufficient Funds.

Brunetti, M., Giarda, E., & Torricelli, C. (2016). Is Financial Fragility a Matter of Illiquidity? An Appraisal for Italian Households. Review of Income and Wealth.

Chaudhuri, S., Jalan, J., & Suryahadi, A. (2002). Assessing household vulnerability to poverty from cross-sectional data: A methodology and estimates from Indonesia.

Coppola, L., & Laurea, D. Di. (2016). Dynamics of persistent poverty in Italy at the beginning of the crisis.

Devlin, J. F. (n.d.). A Detailed Study of Financial Exclusion in the UK.

Foster, J., Greer, J (1984). A class of decomposable poverty measures. Journal of econometric, E. T.-E.

Giannetti, C., Madia, M., & Moretti, L. (2014). Job insecurity and financial distress. Giannetti, M. (2011). Liquidity constraints and occupational choice. Finance Research

Letters, 8(1), 37–44.

Healy, A., & Mansuri, G. (2005). Will it Rain Again Tomorrow? Vulnerability to Poverty and Serial Correlation in Income Shocks in Rural Pakistan.

Helms, B. (n.d.). Access for All Building Inclusive Financial Systems.

Ligon, E., & Schechter, L. (2003). Measuring vulnerability. Economic Journal, 113(486), C95–C102.

Lusardi, A. (2008). Household Saving Behavior: The Role of Financial Literacy, Information, and Financial Education Programs.

Mylonidis, N., Chletsos, M., & Barbagianni, V. (2019). Financial exclusion in the USA: Looking beyond demographics. Journal of Financial Stability, 40, 144–158. Paiella, M. (2007). Does wealth affect consumption? Evidence for Italy. Journal of

Macroeconomics, 29(1), 189–205.

Piraino, P. (2007). Comparable estimates of intergenerational income mobility in Italy. B.E. Journal of Economic Analysis and Policy, 7(2).

(24)

Pritchett, L., Suryahadi, A., & Sumarto, S. (2000). Quantifying Vulnerability to A proposed measure to assess Poverty: Indonesia

Terraneo, M. (2018). Households’ financial vulnerability in Southern Europe. Journal of Economic Studies.

Zamora-Kapoor, A., & Coller, X. (2014, November 27). The Effects of the Crisis: Why Southern Europe? American Behavioral Scientist, 58(12), 1511–1516.

Vulnerability to Poverty and Financial Exclusion: The Italian Case