Page 39 - IDEA Study 3 2015 Duchodova studie

national budget. The mandatory premium pension is the funded part of the earnings- related old-age pension, based on individual accounts in mutual funds. Swedish pensioners (older than 65 years) with low incomes have a guaranteed minimal pension, financed by the government’s budget. The second tier is a compulsory occupational pension covered by nationwide collective bargaining agreements for certain occupations, or voluntary pension plans organised by employers without collective agreements. The statutory retirement age in Sweden is flexible, at between 61 and 70 years. Retirees choose their age of retirement and a premium is awarded for postponed retirement (early retirees at 61 years receive 72% of the average pension, late retirees at 70 years receive 157% of the average pension). The value of pension income is derived on the basis of the life expectancy of Swedish men and women. Regression Model We estimated a linear probability model4 of employment (Ei=1 if person i is employed and 0 otherwise) as (1) Explanatory variables are represented by three row vectors DEMOGRAPHIC, HEALTH, and AGE. The descriptors contained in the vector DEMOGRAPHICi are indicator types of variables of educational attainment (lower, middle and higher education), last occupation (professional, services, manual), family status (living single or with a spouse/partner) and number of children raised (none, one, two, more than two). Individual health situation is captured by indicator variables in the vector HEALTHi for the variables described in chapter 3 and in the Appendix (chronic diseases, ADL, cognitive abilities, self-perceived health, grip strength, depression). The impact of age is captured by a linear spline function with kinks set at ages 55, 60, 65, 70, 75. We estimate model (1) using the SHARE sample of women in the Czech Republic and in the comparison country Sweden in wave 2013. Using estimates of the Swedish coefficients A, B1, B2, and B3, we computed fitted values for the sample of Czech women and men (separately) using their actual (Czech) personal, health and age characteristics. In this way we simulate the Czech women's hypothetical employment probability (with their real individual characteristics) in the Swedish environment. Graph 5.1 depicts three age profiles of average employment probability (share of employed persons in the population): the real employment probability profile in the Czech Republic (denoted as Real CZ and marked by “+”), the real employment probability profile in Sweden (denoted as Real SE and marked by “o”), and the simulated employment probability profile for the Czech sample of women using Swedish coefficients (denoted as Simulated CZ and not marked). 4 As an alternative we used a Probit model, and the key patterns in the resulting simulation were not affected. 37