Setting the conditions for a longitudinal study Rosalba Manna - - PowerPoint PPT Presentation

Research Overview and Preliminary Analysis on CBD and SAMIP: Setting the conditions for a longitudinal study

Rosalba Manna Research Fellow in Statistics, University ”Parthenope” of Naples, Italy Statistician at the National Institute for Educational Research, Florence, Italy and Rocco Palumbo Research Fellow in Organization Studies, University of Salerno, Italy

Who we are

Rocco Palumbo Research Fellow in Organization Studies

• Dept. Management and Innovation Systems,

University of Salerno Lecturer of Human Resource Management and Organizational behavior at the School of Medicine, University “Federico II” of Naples Scientific coordinator of the HLS-IT and FLS-IT projects

Who we are

Rosalba Manna Research Fellow in Statistics

• Dept. Management and Quantitative Studies,

University “Parthenope” of Naples Data analyst and statistician at the National Institute for Educational Innovation and Research (INDIRE), Florence Fellow statistician at the Italian Institute of Statistics (ISTAT) Rome, Italy

Aims of our visiting at SOFI

• 1. To measure the effect of child benefits on income inequality in two group
• f countries using Child Benefit Dataset (CBD) of Social Policy Indicators

Database (SPIN). Income data provided by EUROSTAT were also used. The following two groups of Countries were contemplated:

• Italy, France, Austria and Belgium
• Sweden, Denmark, Norway and Finland

The time span ranged:

• from 1995 to 2010

Aims of our visiting at SOFI

• 2. To investigate the relationship between different forms of social

assistances – as assessed by the SAMID interim dataset – and a set of socio- economic outcomes, including net earnings, material deprivation and crude divorce rates. The following European Countries were contemplated: Austria, Belgium, Denmark, Finland, France, Germany, Ireland, Italy, Lichtenstein, Spain, Portugal, and UK. The time span ranged:

• From 1990 to 2013 for most of Countries
• From 1990 to 2009 for Italy
• From 1996 to 2013 for Portugal

Decomposition methods

A regression based decomposition technique was implemented, following the Pyatt (1976) approach The analysis exploited the potential of panel data, with reference to the pooling of observations on a cross-section of countries over several time periods.

Pyatt`s decomposition

Pyatt`s decomposition allows to compute Gini index within and between group inequality decomposition. It provides the Gini coefficient for the whole population, for each subgroup specified. Any group is a categorical variable (in our case: country) which determines the subgroups in which the population is grouped. We choose this method to capture the effects both across countries (from panel data estimation with fixed effects) and within different countries.

Overview decomposition methods: Traditional approaches

• INCOME SOURCE:

This method decomposes the Gini coefficient by income source, using the approach described in Lerman and Yitzhaki (1985) and Stark, Taylor, and Yitzhaki (1986), which allows the calculation of the impact that a marginal change in a particular income source has on inequality.

• POPULATION SUBGROUPS:

This method estimates a range of inequality and related indices commonly used by economist to calculate inequality measures by subgroups (gender gap, educational levels, etc…).

Regression-based decomposition methods

SHAPLEY VALUE AND FIELDS APPROACH:

The regression-based method allows to quantify the contribution of a set of factors to inequality, while taking into account the correlations among them. Shapley Value and Fields methods allow to measure in micro and macro analysis the relative contributions of any factors to inequality. OAXACA-BLINDER (1973) This method performs a decomposition of the mean outcome differential of linear and nonlinear regression models, as described by Bauer and Sinning (2008). Daymond and Andrisani (1984) proposed an extension of the B-O decomposition.

Income generating model

• 1. First step:
• Specification and estimation of an income generating function
• The log of income is regressed on some explanatory variables:
• Employment-based child benefit
• Total child benefit (Child tax allowance+Child tax

credit+Child tax rebate)

• Etc.

Model Selection

• Second step:

A model for longitudinal data with fixed effects is selected:

Where:

• Yi is the vector Y = (Y1, Y2, …, YT) of the dependent variable;
• The observations of the explicative variables (i.e. the regressors) are included in the matrix X, whose

dimensions are (T * K);

• The parameters to be estimated are K and they are part of the vector:
• The factor αi is the individual effect.

i i i i

      i X Y

Model Assumptions

• The panel data model with fixed effects allows to capture the heterogeneity between

countries in several time periods;

• The effects are correlated with the regressors;
• This hypothesis does not allow to include time-invariant covariates, but in our case this is

not a problem, because the analysis is conducted on countries, and not on individuals.

Longitudinal Data Models

• LINEAR REGRESSION
• LINEAR REGRESSION WITH AR
• BINARY OUTCOMES
• ORDINAL OUTCOMES
• COUNT OUTCOMES
• ENDOGENOUS COVARIATE
• Etc...

Longitudinal Data Models

• LINEAR REGRESSION
• LINEAR REGRESSION WITH AR BINARY AUCOMES
• ORDINAL OUTCOMES
• COUNT OUTCOMES
• CENSORED OUTCOMES
• SURVIVAL MODELS
• ENDOGENOUS COVARIATE

FE RE PA BE

Linear Regression

• RE
• BE
• MLE
• AR

Binary Outcomes

If we assume a normal distribution for the RE , where The panel-level likelihood is given by

BINARY OUTCOMES

• Logistic Regression (FE,RE, PA)
• Probit Regression (RE, PA)

ORDINAL OUTCOMES

• Logistic (RE)
• Probit (RE)

COUNT OUTCOMES

• Poisson regression (FE, RE, PA)
• Negative Binomial regression (FE, RE, PA)

Alternative Data Models

DYNAMIC PANEL DATA

• Arellano/Bond estimation
• Arellano/Bover e Blundell/Bond

ENDOGENOUS COVARIATE

• Instrumental variables regression (FE; RE;

FD, BE)

• Hausman Taylor (RE)

Alternative Data Models

SPIN

• As previously anticipated, data were collected from the SPIN

database; they were handled to perform comparative and longitudinal research, in association with the EUROSTAT database.

• It covers 34 countries from 1930 to 2013 and includes the following

modules:

• OUTWB (Out of Work Benefit)
• CBD (Child Benefit Database)
• PLB (Parental Leave Benefit)
• SAMIP (Social Assistance and Minimum Income Protection)
• SCIP (Social Citizenship Indicator Program)
• SIED (Social Insurance Entitlements Dataset)
• SPEAD (Social Policy in East Asia Dataset)
• CCD (Child Care Dataset)

CB dataset

• The current version of CBD includes detailed information about

the generosity of child benefit in 18 countries between 1960 and 2010.

• We conveniently selected data on 8 countries from 1995 to 2010.
• This choice was dictated by availability of data and it is inspired by

the purpose of analyzing 2 group of countries in function of their geographical, social and cultural proximity.

• The first group includes: Sweden, Finland, Norway and Denmark
• The second group includes: Italy, France, Austria and Belgium

Variables CBD

• Dependent variable: income expressed in € and transformed in log, in order to

allow the appropriate estimation of the model;

• Benefits levels are calculated for a two-parent model family with two children

aged 2 and 7;

• All benefit in CBD are expressed in national currency and converted in € to allows

comparisons.

All Countries

AU, BE, FR, IT DK, FI,NO,SE

AU, BE, FR, IT DK, FI,NO,SE

AU, BE, FR, IT DK, FI,NO,SE

Social Benefit (%)

Consist of transfers, in cash or in kind, by social protection schemes to households and individuals to relieve them of the burden of a defined set of risk or needs. The risk are: sickness/healthcare, disability, old age, survivors, children, unemployment, housing, social exclusion. In this work, we consider social benefit to children.

Total Child Benefits trend

GINI* before the crisis (2007)

*The Gini coefficient measures the extent to which the distribution of income within a country deviates from a perfectly equal distribution. A coefficient of 0 expresses perfect equality where everyone has the same income, while a coefficient of 100 expresses full inequality where

• nly one person has all the

income.

GINI during the crisis (2008)

GINI during the crisis (2009)

People at risk of poverty* or social exclusion** (%), 2007

*At risk-of-poverty are persons with an equivalised disposable income below the risk-of- poverty threshold, which is set at 60 % of the national median equivalised disposable income (after social transfers) ** Severely materially deprived persons have living conditions severely constrained by a lack

• f resources.

People at risk of poverty or social exclusion (%) , 2008

People at risk of poverty or social exclusion (%), 2009

SAMIP interim dataset

• The current version of SAMIP includes detailed

information about the generosity of means-tested benefits in 36 countries year-by-year for the period 1990-2013.

• We conveniently selected data on 15 countries from

1990 to 2013.

• All Western European Countries, excluding

Switzerland and Iceland, were involved in the analysis.

• We tried to investigate the relationship between the

generosity of social assistance and a set of social

• utcomes.
• We contemplated a model family composed of two

children aged 7 and 14.

Main findings

• We found that the generosity of social assistance had

a positive and significant effect on actual consumption across Western European Countries.

• Interestingly, in Scandinavian Countries Child and

House supplement were especially likely to determine an increase of actual consumption.

• Alternatively, in British Countries Social Assistance

and Child supplements had a relevant and positive influence on actual consumption.

• Obviously, further developments are needed to shed

light on this issue and to pinpoint its implications.

Main findings

• Surprisingly, a negative relationship linked the

generosity of social assistance in Western European Countries and the individual access to health-related services.

• In spite of this consideration, in Scandinavian

countries the generosity of Minimum Income Protection policies had a positive effect on the individual access to health services.

• From this point of view, the implications of different

forms of social assistance should be carefully investigated to illuminate the interplay between access to care and social policies.

Main findings

• While minimum income protection seemed to

encourage births across Western European Countries, house supplements showed a negative relationship with birth rates.

• This was particularly true in continental European

Countries and in Italy. Probably, this circumstance was dictated by the higher socio-economic inequality which affects these kinds of Countries.

• Interestingly, refundable tax credits were likely to

curb divorce rates across European Countries.

Main findings

• Refundable tax credits, social assistance generosity

and child supplements were found to have a significant and negative influence on emigration.

• Alternatively, we did not find significant evidence on

the relationship between social assistance generosity and immigration levels.

Main findings

• Social assistance generosity was positively and

significantly associated with net earnings across Western European Countries; however, further analysis is required to assess the ability of social assistance policy to fill socio-economic inequalities.

• Interestingly, child supplements were found to

be positively related to income in Continental European Countries, while house supplements had a positive relation with earnings in both Scandinavian and British Countries.

• Lastly, yet importantly, in Continental European

Countries house supplements were likely to reduce material deprivation rates.

What’s next

• To refine our estimation model and to apply it to all the

countries included in SPIN;

• To build a robust and consistent conceptual framework,

aimed at contextualizing the evidence obtained from empirical research;

• To identify additional secondary sources – beyond Eurostat

databases – to increase the breadth and the depth of our study;

• To design a mixed research strategy intended to merge the

“macro level” analysis allowed by SPIN with a more depth investigation at single Country levels;

• And, lastly, yet importantly, to delve into the relationship

between social policies and health-related conditions of citizens, envisioning several scenarios for future health and social policies.

We hope to see you soon in Italy …or in Stockholm, as well! Rosalba Manna

rosalba.manna@uniparthenope.it

Rocco Palumbo

rpalumbo@unisa.it