Homelessness and Labour Force Participation. Evidence from an Original Data Collection in Italy

HOMELESSNESS and LABOR FORCE PARTICIPATION. Evidence from an Original Data Collection in Milan

Homelessness and Poverty in EuropeParis, September 18th, 2009

Michela Braga

University of Milan

MAIN OBJECTIVES Quantitative and qualitative data collection:

First Census of homeless in Milan

=> count and localization Data collection to understand not only the number of homeless and the

concentration, but also to capture characteristics

=> questionnaire

Are homeless people different from the general population? If yes in which dimension?

Are homeless people rational according to economic theory?

=> case study: labor market behavior

MOTIVATION Information on the number and characteristics of the homeless is

necessary for program planning Quantitative and qualitative data are necessary to quantify

economic resources to reduce homelessness and to prevent it with policies

Baseline survey for further studies => program evaluation

Cross countries analysis: gap between Italian and international research: In US, systematic data collection year by year starting from the

early 80’s In Europe some attempts have been made

…but in a non systematic way No data available in Italy

METHODOLOGY: data collection

All individuals that in the reference night sleep in places not meant for human habitation = street homeless; emergency shelters = sheltered homeless; disused areas/shacks/slums

65 small census blocks Reduce risk of double count (3/4 hours for each block) Simultaneous full census of the whole city

Localization and detection of observable characteristics Costs: monetary, human, time vs Benefits: accuracy, limit under estimates

Sampling procedure: Street: all population Shelter: Random sample proportional to the shelter dimension. Over – sampling

for the small ones and under – sampling for the big ones Disused areas: Stratified random sample according

City administrative division (9 areas) Official area classification (authorized, non authorized, shacks, abandoned

buildings, disused areas, ride men); Dimension: small (n≤30), medium (30<n<100) and big (n≥ 100)

Trade off between accuracy of the data collection and loss of observations

TARGET

COUNT

INTERVIEW

Point in time survey using the S - Night approach (Shelter and Street Night): January 14th 2008

THE HOMELESS POPULATION

408 individuals, 34.5% interviewed 12% refusal rate 21% not found 17% sleeping

1152 individuals, 80% of the sampled interviewed 2% refusal rate 7% not found

2300 adults, 66% of the sample interviewed 33% not found

STREET

SHELTER

DISUSED AREAS

Total adult population: 3863

Final Sample: 941 homeless

Legend: [ blue] =Localization of unsheltered homeless, each dot=1 homeless [ pink ] =Localization of shelters, each dot =10 homeless [ grey] =Localization of slums, each dot =10 homeless

DATA: socio – demographic characteristics

Differently from the general population, the homeless are mainly men (72% vs. 48%) and immigrata (68% vs 5.8%)… but there is a significant variation by sex and nationality in the three sub samples

% Females % Italians

Street 10 56

Shelters 16 40

Disused areas 49 11

Geographical origin in line with general population First generation immigrants => starting period of their

migration project High incidence of divorce (20%) and loss of strong family

ties ( sons, parents)

DATA: age Adults in the central part of their life (average age 39.9)

=> failures in individual life projects (lack/loss job, family relationships, divorces..)…but the total population is spread across all age groups

Younger than general population (42.6) for the high incidence of immigrants. All categories are older than in the general population HL: Italian M=51.1 Foreign M=35 Italian F = 45.6 Foreign F=35.2 GP: Italian M=41.6 Foreign M=30.4 Italian F = 44.5 Foreign F=31.3

Average age is higher among street homeless (49) than among sheltered homeless (43).

Population younger in disused areas (30.7) as in general population (30.9 years)

DATA: education

All sample Italian Foreign Street ShelterDisused

areasGeneral

populationNone 14.45 8.88 17.11 10.71 6.84 25.5 6.8Elementary school 21.68 29.28 18.05 18.45 17.45 28.37 26.4Middle school 33.16 39.47 30.14 34.52 34.43 30.95 31.7High school 25.19 19.41 27.94 30.36 32.78 13.47 27.2University 5.53 2.96 6.75 5.95 8.49 1.72 7.9

Education distribution is in line with the one found in the general population

Higher proportion of people with no education More educated people tend to stay in shelter As in the general population, on average, immigrants are more educated

than native born Native have 8.2 years of education Immigrants have 9.7 years of education…but the higher education level reflects their age structure

DATA: labor market behavior

Labor force participation is higher compared with the general population The 29.3% was employed at the time of the survey. Among unemployed people the

17% worked during the previous month More than half of people are employed in the black market compared with the

12.1% in the general population Only 13% have permanent contract and a significant percentage (20%) has

temporary contract while in the general population the percentages are 65% for permanent and 10% for temporary

Unemployed people are actively looking for a job Reservation wage 827 €

Population non excluded from the labor market but less stable

DATA: income

Low take up rate to social assistance programs and welfare state

Weekly average income 151 €. Higher in disused areas (164€) than on street and in shelters (140 and 145)

=> not lower than the poverty line treshold in Italy (246.5€ for a two person household) but not sufficient to afford everyday expenditures in Milan

ARE HOMELESS PEOPLE RATIONAL AGENTS?

Homeless people are thought to be no rational agents (from an economic point of view) as a result of their housing condition, drug/alcohol use, physic and psychic disorders

Determine which variables affect homeless people's labor market behavior

Test if they are in line with the underlying theoretical framework of utility

maximization and labor-leisure choice

EMPIRICAL ANALYSIS

yi= β0+ β1 Xi +μi

yi = binary variable defining individual labour market status (in vs out labour force), employment status (employed/unemployed), source of income (legal/illegal)

Xi = exogenous explanatory variablesμi = error term

RESULTS (I): Labor market participation

Variables affecting labour market participation in line with the utility maximization and labour-leisure choice framework Traditional income effect Education ↑ probability to be active Gender gap Awareness and degree of information ↑ probability to

be active

Labor force participation (1) (2) (3) (4)Female -0.0737*** -0.0718*** -0.0738*** -0.0600***

[0.0156] [0.0166] [0.0168] [0.0177]Age 0.0281*** 0.0285*** 0.0287*** 0.0275***

[0.0033] [0.0030] [0.0031] [0.0033]Age (squared) -0.0004*** -0.0004*** -0.0004*** -0.0004***

[0.0000] [0.0000] [0.0000] [0.0000]Primary Edu.Level 0.0846*** 0.0848*** 0.0861*** 0.0978***

[0.0173] [0.0195] [0.0195] [0.0213]Middle Edu. Level 0.1414*** 0.1418*** 0.1439*** 0.1555***

[0.0058] [0.0170] [0.0174] [0.0208]Secondary Edu. Level 0.0718** 0.0741 0.0782 0.0866

[0.0358] [0.0559] [0.0544] [0.0574]Universitary Edu. Level -0.0303 -0.0285 -0.0251 -0.0233

[0.1173] [0.1494] [0.1506] [0.1530]Received money from family -0.1214*** -0.1224*** -0.1232*** -0.1121***

[0.0315] [0.0279] [0.0275] [0.0234]Non-financial help -0.1533*** -0.1575*** -0.1649*** -0.1624***

[0.0371] [0.0075] [0.0058] [0.0069]Essential inkind help 0.3051*** 0.3124*** 0.3208*** 0.3181***

[0.0442] [0.0631] [0.0651] [0.0763]Prison before -0.0822* -0.0557* -0.0215 -0.0173

[0.0468] [0.0329] [0.0377] [0.0438]Shelter 0.0491*** 0.0450***

[0.0044] [0.0071]Disused area 0.2037*** 0.1920***

[0.0491] [0.0418]Romanian 0.0510*** 0.0506** 0.0461** 0.0448**

[0.0180] [0.0201] [0.0231] [0.0219]Other Europe -0.0442* -0.0496** -0.0455** -0.0477***

[0.0261] [0.0213] [0.0188] [0.0136]African 0.1351*** 0.1313*** 0.1333*** 0.1305***

[0.0056] [0.0051] [0.0070] [0.0084]Asian/American and other 0.1511*** 0.1466*** 0.1465*** 0.1457***

[0.0153] [0.0164] [0.0163] [0.0165]Non labor income -0.1762*** -0.1688*** -0.1710*** -0.1711***

[0.0076] [0.0050] [0.0038] [0.0053]Sick in the past month -0.0532* -0.0548* -0.0523

[0.0297] [0.0291] [0.0352]Wrong month -0.1069** -0.1075** -0.0997*

[0.0545] [0.0543] [0.0559]Shelter 0.0431*** 0.0578***

[0.0075] [0.0097]Authorized disused area 0.1898*** 0.2128***

[0.0401] [0.0417]Non authorized disused area 0.1347*** 0.1536***

[0.0301] [0.0287]Read new spaper 0.0776***

[0.0089]Information 0.0094 0.0095

[0.0100] [0.0102]

RESULTS (II): Employment

Factors affecting the probabily to be employed are in line with those of the general population Gender gap in favour of males More educated people have a relative advantage Traditional income effect Awareness and degree of information ↑ probability to

be employed Previous convictions not correlated with the probability

to be employed

(1) (2) (3) (4) (5)Female -0.1373*** -0.0715*** -0.0569*** -0.0562*** -0.0555**

[0.0418] [0.0249] [0.0181] [0.0177] [0.0241]Age 0.0168*** 0,0022 -0,0013 -0,0014 -0,0012

[0.0065] [0.0094] [0.0079] [0.0078] [0.0074]Age (squared) -0.0002** 0 0 0 0

[0.0001] [0.0001] [0.0001] [0.0001] [0.0001]Primary Edu.Level -0.3215*** -0.1486*** -0.1625*** -0.1619*** -0.1647***

[0.0110] [0.0111] [0.0281] [0.0288] [0.0410]Middle Edu. Level -0.3037*** -0.1636*** -0.1856*** -0.1851*** -0.1894***

[0.0242] [0.0266] [0.0060] [0.0055] [0.0054]Secondary Edu. Level -0.2876*** -0.1396*** -0.1581*** -0.1591*** -0.1622***

[0.0221] [0.0274] [0.0183] [0.0185] [0.0176]Universitary Edu. Level -0.2196*** -0,0896 -0,1032 -0,1026 -0,1053

[0.0618] [0.0983] [0.0789] [0.0786] [0.0731]No family Relations -0.0551* -0,0218 -0,0253 -0,0274 -0,0276

[0.0291] [0.0464] [0.0368] [0.0357] [0.0389]Faith 0,0034 -0,0115 0,0118 0,0121 0,0109

[0.0598] [0.0573] [0.0630] [0.0626] [0.0610]Received money from family -0.1893*** -0.0499*** -0.0589** -0.0581** -0.0610***

[0.0226] [0.0183] [0.0236] [0.0236] [0.0220]Received money from friends -0.1442*** -0.0378* -0.0330* -0.0344* -0,0325

[0.0250] [0.0201] [0.0191] [0.0205] [0.0208]Financial help from close relatives -0,0011 0,0007 0,0008 0,0009 0,0009

[0.0012] [0.0016] [0.0017] [0.0017] [0.0017]Non-financial help 0.1695*** 0,093 0,0741 0,0779 0,0776

[0.0444] [0.0728] [0.1315] [0.1294] [0.1236]Essential inkind help -0.2380*** -0,1521 -0,1306 -0,1354 -0,134

[0.0669] [0.0972] [0.1563] [0.1557] [0.1536]Additional inkind help -0,0115 -0,0142 -0,0069 -0,007 -0,0068

[0.0399] [0.0419] [0.0383] [0.0382] [0.0401]Prison before -0,0512 0.0368** 0,0027 0,0034 0,0033

[0.0438] [0.0179] [0.0396] [0.0397] [0.0491]Prison after -0,0499 -0,0195 -0,0102 -0,0111 -0,0096

[0.0483] [0.0382] [0.0441] [0.0436] [0.0434]Shelter 0.0344*** -0,0033 -0,0113

[0.0097] [0.0102] [0.0110]Disused area 0.1451*** 0.2029*** 0.1867***

[0.0552] [0.0500] [0.0418]Romanian 0,0077 0,0211 0,0295 0,0336 0,0371

[0.0767] [0.0841] [0.0753] [0.0785] [0.0853]Other Europe 0,0965 0,0555 0,078 0,0764 0,0779

[0.1184] [0.1219] [0.1129] [0.1129] [0.1109]African -0.1774** -0.1450*** -0.1316*** -0.1344*** -0.1334***

[0.0852] [0.0361] [0.0309] [0.0305] [0.0296]Asian/American and other 0.1233*** 0.1761*** 0.1798*** 0.1788*** 0.1799***

[0.0319] [0.0388] [0.0468] [0.0465] [0.0499]Duration 0.0158*** 0.0142** 0.0153** 0.0145**

[0.0059] [0.0060] [0.0063] [0.0067]In and out 0.1250*** 0.1210*** 0.1202*** 0.1200***

[0.0235] [0.0222] [0.0214] [0.0220]Non labor income -0.5243*** -0.5259*** -0.5258*** -0.5251***

[0.0287] [0.0301] [0.0304] [0.0291]Sick in the past month -0.0643** -0.0642** -0.0646**

[0.0289] [0.0291] [0.0297]Wrong month -0.0738* -0.0717* -0,0744

[0.0384] [0.0399] [0.0500]Wrong year -0,0005 -0,0018 0,0003

[0.0275] [0.0279] [0.0224]Drug use 0.1069*** 0.1049*** 0.1067***

[0.0274] [0.0281] [0.0229]

Determinants of being employed

RESULTS (III): sources of income

Rationality hypothesis seems to hold also for what concerns individual income sources (legal/illegal) No gender gap nor nationality gap No age effect More educated people are less prone to act illegally to

obtain income Traditional income effect Previous convictions not correlated with current illegal

behaviour Drug use correlated with illegal behaviour

Illegal activities (1) (2) (3) (4)Female -0,0099 -0.0117*** -0.0115*** -0.0127***

[0.0065] [0.0039] [0.0039] [0.0032]Age 0.0029*** 0.0027*** 0.0027*** 0.0027***

[0.0006] [0.0004] [0.0004] [0.0005]Age (squared) -0.0000*** -0.0000*** -0.0000*** -0.0000***

[0.0000] [0.0000] [0.0000] [0.0000]Formal job -0.0348*** -0.0323*** -0.0321*** -0.0316***

[0.0022] [0.0033] [0.0033] [0.0034]Primary Edu.Level 0.8592*** 0.8199*** 0.8193*** 0.7981***

[0.0252] [0.0322] [0.0313] [0.0384]Middle Edu. Level 0.6479*** 0.6022*** 0.5994*** 0.5766***

[0.0808] [0.0878] [0.0867] [0.0934]Secondary Edu. Level 0.7151*** 0.6546*** 0.6504*** 0.6297***

[0.0411] [0.0516] [0.0537] [0.0640]Universitary Edu. Level 0.9252*** 0.9042*** 0.9042*** 0.8986***

[0.0638] [0.0744] [0.0763] [0.0857]No family Relations -0.0095*** -0.0097*** -0.0099*** -0.0100***

[0.0031] [0.0001] [0.0002] [0.0003]Received money from friends -0.0090*** -0.0054* -0.0055* -0.0045**

[0.0004] [0.0030] [0.0028] [0.0022]Financial help from close relatives -0,0001 -0,0001 -0,0001 -0,0001

[0.0002] [0.0002] [0.0002] [0.0002]Non-financial help -0,0029 0,0012 0,0021 0,0015

[0.0302] [0.0246] [0.0238] [0.0236]Essential inkind help -0,0083 -0,0103 -0,011 -0,0105

[0.0329] [0.0309] [0.0314] [0.0294]Prison before 0,0073 -0,0105 -0,0104 -0,0098

[0.0301] [0.0143] [0.0141] [0.0140]Shelter -0.0138*** -0.0107***

[0.0015] [0.0018]Disused area -0,0007 0,0076

[0.0070] [0.0104]Romanian 0,0049 0,0077 0,009 0,0089

[0.0084] [0.0072] [0.0078] [0.0079]Other Europe 0,0127 0,0114 0,0115 0,0117

[0.0144] [0.0105] [0.0105] [0.0105]African 0,0063 0,0082 0,0072 0,0075

[0.0112] [0.0096] [0.0088] [0.0089]Asian/American and other -0,0108 -0,0083 -0,0084 -0,0088

[0.0085] [0.0089] [0.0085] [0.0088]Duration 0.0032*** 0.0031*** 0.0035*** 0.0030***

[0.0010] [0.0007] [0.0010] [0.0011]Drug use 0.0170*** 0.0166*** 0.0165***

[0.0017] [0.0013] [0.0013]Legal problems 0,0177 0,0178 0,0179

[0.0263] [0.0264] [0.0252]Shelter -0.0104*** -0.0119***

[0.0023] [0.0030]Authorized disused area 0,0038 0,0008

[0.0083] [0.0086]Non authorized disused area 0,0136 0,0096

[0.0148] [0.0144]Observations 882 856 856 856Pseudo R-squared 0,1884 0,2003 0,2011 0,2039

CONCLUSION Homeless population similar in many dimensions to the Italian general

population

Variables affecting homeless people's labor market behavior are in line with the underlying theoretical framework of utility maximization and labor-leisure choice Rationality hypothesis satisfied

Correlation vs. causality? Necessary to solve endogeneity problems In kind help = > variation in charity services within the city

=> journal articles on homelessness, news on television

Duration => weather conditions (average temperature, rainfall) from the first arrive in street