[ Pobierz całość w formacie PDF ]
.The implication is that,from a practical standpoint, the coefficients obtained from estimatingthe leaving and employment probabilities separately are not seriously The Determinants of Welfare Exits and Employment 69biased.We therefore do not present coefficient estimates for the bivari-ate probit.Despite the similarities in estimated coefficients, the estimated cor-relation between the unobserved determinants of leaving welfare andof employment (ÃEL) is significantly different from zero, positive, andsubstantial for each site (see Table 3.3).The similarity in the correla-tion estimates across sites, varying from 0.28 to 0.36, is notable.Thepositive estimate of ÃEL indicates that unobserved factors that increasethe probability of leaving welfare also increase the probability of em-ployment.For example, a recipient who is more likely in any quarterto leave welfare for some unmeasured reason (e.g., high motivation) isalso more likely to be employed in that quarter.The parameter coefficients from estimation of the bivariate specifi-cations are used to calculate the expected joint, conditional, and uncon-ditional probabilities presented in Table 3.3.As noted above, althoughemployment and leaving welfare are not modeled as having a causalrelationship (they are modeled simultaneously), the importance of oneoutcome in the determination of the probability of the other outcomecan be calculated through differences in conditional probabilities.The partial derivatives reported at the bottom of Table 3.3 are inter-preted as follows:(a) Among recipients, how does the probability of leaving welfarediffer between those who are employed and those who are not? Whereunmeasured factors induce employment, we may interpret this as indi-cating the extent to which employment is associated with a change inthe probability of leaving.(b) Among recipients, how does the probability of being employeddiffer between those who are leaving welfare and those who are not?Where unmeasured factors induce welfare exit, we may interpret this asindicating the extent to which welfare exit is associated with a changein the probability of employment.The partial derivatives of the leaving probability, line (a), indicatethat an employed recipient is more likely to leave welfare than a non-employed recipient, with the differential between 9 (Chicago) and 17(Houston) percentage points.The implication is that employed recipi-ents have a tremendous advantage over nonemployed recipients in leav-ing welfare.The importance of leaving welfare on the probability of be-ing employed is reported in line (b) of Table 3.3.These figures indicate Table 3.3 Predicted Probabilities and Partial Derivatives, Leaving and EmploymentFortAtlanta Baltimore Chicago Lauderdale Houston Kansas CityError covarianceFEL 0.350 0.356 0.277 0.289 0.360 0.297Unconditional probabilitiesPr[L=1] 0.10 0.11 0.11 0.24 0.18 0.13Pr[E=1] 0.37 0.27 0.38 0.38 0.27 0.44Pr[L=1,E=1] 0.07 0.06 0.06 0.12 0.09 0.08Pr[L=1,E=0] 0.04 0.05 0.05 0.11 0.09 0.05Pr[L=0,E=1] 0.31 0.21 0.32 0.25 0.19 0.36Pr[L=0,E=0] 0.59 0.68 0.57 0.51 0.64 0.51Conditional probabilitiesPr[L=1|E=0] 0.06 0.08 0.08 0.18 0.13 0.09Pr[L=1|E=1] 0.17 0.20 0.17 0.33 0.30 0.19Pr[E=1|L=0] 0.34 0.24 0.36 0.34 0.23 0.41Pr[E=1|L=1] 0.62 0.50 0.57 0.53 0.47 0.6370 King and Mueser Partial derivatives(a)Pr[L=1|E=1] - Pr[L=1|E=0] 0.11 0.12 0.09 0.14 0.17 0.10(As proportion of overall(1.10) (1.09) (0.81) (0.58) (0.94) (0.77)leaving probability)(b)Pr[E=1|L=1] - Pr[E=1|L=0] 0.28 0.26 0.21 0.19 0.24 0.22(As proportion of overall(0.76) (0.96) (0.55) (0.50) (0.89) (0.50)employment probability)NOTE: Based on model with demographic controls estimated for entire sample period.Terms in parentheses are the difference abovedivided by the unconditional probability.For line (a), the divisor is Pr[L=1], and for line (b) the divisor is Pr[E=1].Probabilities reflectthe expected probability for a recipient drawn at random.The Determinants of Welfare Exits and Employment 71 72 King and Mueserthat a recipient leaving welfare has from a 19- to a 28-percentage-pointincrement in probability of being employed relative to a nonleavingrecipient.It is notable that the patterns across sites are similar, eventhough exit probabilities are much higher in Fort Lauderdale and Hous-ton than in the other sites.Certainly, these results show that employment and welfare exits areclosely associated, and that achieving one of these outcomes brings arecipient a long way toward achieving the other.If one simply askshow one obtains the highest probability of exiting welfare and beingemployed, it is clear that, if we observe an exit for the average person,the chance is around 50 percent that the individual will also be em-ployed (Pr[E = 1|L = 1]).On the other hand, if one identifies employedindividuals, the chance that such an individual also leaves welfare in agiven quarter is less than a third, and as low as 17 percent in two sites(Pr[L = 1|E = 1]).Since the welfare exit is the smaller probability event,achieving it moves one farther toward achieving the joint goal.However, this comparison may understate the value of obtainingemployment.Rather than looking at the simple probability, one maywish to apply an adjustment for the overall probability of each outcome.The figures in parentheses in line (a) divide the calculated differentialfor the chance of leaving by the overall probability of leaving [ Pobierz caÅ‚ość w formacie PDF ]
zanotowane.pl doc.pisz.pl pdf.pisz.pl trzylatki.xlx.pl
.The implication is that,from a practical standpoint, the coefficients obtained from estimatingthe leaving and employment probabilities separately are not seriously The Determinants of Welfare Exits and Employment 69biased.We therefore do not present coefficient estimates for the bivari-ate probit.Despite the similarities in estimated coefficients, the estimated cor-relation between the unobserved determinants of leaving welfare andof employment (ÃEL) is significantly different from zero, positive, andsubstantial for each site (see Table 3.3).The similarity in the correla-tion estimates across sites, varying from 0.28 to 0.36, is notable.Thepositive estimate of ÃEL indicates that unobserved factors that increasethe probability of leaving welfare also increase the probability of em-ployment.For example, a recipient who is more likely in any quarterto leave welfare for some unmeasured reason (e.g., high motivation) isalso more likely to be employed in that quarter.The parameter coefficients from estimation of the bivariate specifi-cations are used to calculate the expected joint, conditional, and uncon-ditional probabilities presented in Table 3.3.As noted above, althoughemployment and leaving welfare are not modeled as having a causalrelationship (they are modeled simultaneously), the importance of oneoutcome in the determination of the probability of the other outcomecan be calculated through differences in conditional probabilities.The partial derivatives reported at the bottom of Table 3.3 are inter-preted as follows:(a) Among recipients, how does the probability of leaving welfarediffer between those who are employed and those who are not? Whereunmeasured factors induce employment, we may interpret this as indi-cating the extent to which employment is associated with a change inthe probability of leaving.(b) Among recipients, how does the probability of being employeddiffer between those who are leaving welfare and those who are not?Where unmeasured factors induce welfare exit, we may interpret this asindicating the extent to which welfare exit is associated with a changein the probability of employment.The partial derivatives of the leaving probability, line (a), indicatethat an employed recipient is more likely to leave welfare than a non-employed recipient, with the differential between 9 (Chicago) and 17(Houston) percentage points.The implication is that employed recipi-ents have a tremendous advantage over nonemployed recipients in leav-ing welfare.The importance of leaving welfare on the probability of be-ing employed is reported in line (b) of Table 3.3.These figures indicate Table 3.3 Predicted Probabilities and Partial Derivatives, Leaving and EmploymentFortAtlanta Baltimore Chicago Lauderdale Houston Kansas CityError covarianceFEL 0.350 0.356 0.277 0.289 0.360 0.297Unconditional probabilitiesPr[L=1] 0.10 0.11 0.11 0.24 0.18 0.13Pr[E=1] 0.37 0.27 0.38 0.38 0.27 0.44Pr[L=1,E=1] 0.07 0.06 0.06 0.12 0.09 0.08Pr[L=1,E=0] 0.04 0.05 0.05 0.11 0.09 0.05Pr[L=0,E=1] 0.31 0.21 0.32 0.25 0.19 0.36Pr[L=0,E=0] 0.59 0.68 0.57 0.51 0.64 0.51Conditional probabilitiesPr[L=1|E=0] 0.06 0.08 0.08 0.18 0.13 0.09Pr[L=1|E=1] 0.17 0.20 0.17 0.33 0.30 0.19Pr[E=1|L=0] 0.34 0.24 0.36 0.34 0.23 0.41Pr[E=1|L=1] 0.62 0.50 0.57 0.53 0.47 0.6370 King and Mueser Partial derivatives(a)Pr[L=1|E=1] - Pr[L=1|E=0] 0.11 0.12 0.09 0.14 0.17 0.10(As proportion of overall(1.10) (1.09) (0.81) (0.58) (0.94) (0.77)leaving probability)(b)Pr[E=1|L=1] - Pr[E=1|L=0] 0.28 0.26 0.21 0.19 0.24 0.22(As proportion of overall(0.76) (0.96) (0.55) (0.50) (0.89) (0.50)employment probability)NOTE: Based on model with demographic controls estimated for entire sample period.Terms in parentheses are the difference abovedivided by the unconditional probability.For line (a), the divisor is Pr[L=1], and for line (b) the divisor is Pr[E=1].Probabilities reflectthe expected probability for a recipient drawn at random.The Determinants of Welfare Exits and Employment 71 72 King and Mueserthat a recipient leaving welfare has from a 19- to a 28-percentage-pointincrement in probability of being employed relative to a nonleavingrecipient.It is notable that the patterns across sites are similar, eventhough exit probabilities are much higher in Fort Lauderdale and Hous-ton than in the other sites.Certainly, these results show that employment and welfare exits areclosely associated, and that achieving one of these outcomes brings arecipient a long way toward achieving the other.If one simply askshow one obtains the highest probability of exiting welfare and beingemployed, it is clear that, if we observe an exit for the average person,the chance is around 50 percent that the individual will also be em-ployed (Pr[E = 1|L = 1]).On the other hand, if one identifies employedindividuals, the chance that such an individual also leaves welfare in agiven quarter is less than a third, and as low as 17 percent in two sites(Pr[L = 1|E = 1]).Since the welfare exit is the smaller probability event,achieving it moves one farther toward achieving the joint goal.However, this comparison may understate the value of obtainingemployment.Rather than looking at the simple probability, one maywish to apply an adjustment for the overall probability of each outcome.The figures in parentheses in line (a) divide the calculated differentialfor the chance of leaving by the overall probability of leaving [ Pobierz caÅ‚ość w formacie PDF ]