The Dynamics of Welfare Participation in Newfoundland: 1986-1998 - April 1999

2. Sampling Procedure and Basic Data Description

Previous

Contents

Next

The data used in this study are drawn from the 100% Social Assistance Recipients file. The file contains as many as 3,288,593 records on spells that occurred between January 1986 and June 1998. The records are sorted with respect to the Social Insurance Number (SIN) of each recipient and each record contains information on a single month of a specific spell. Once the records are converted into spells, we end up with as many as 102,829 individuals who had a claim during that period.¹

Naturally, some individuals have severe work impediments and their stay on welfare can be considered nearly permanent. For obvious reasons, these individuals must be excluded from the sample. We use the "Employment Status" variable to determine work eligibility. Thus, anyone suffering from blindness (Code 23), mental illness (Code 24), mental retardation (Code 25), physical illness (Code 26) or "social disability" (Code 27) is excluded from the sample. These criteria reduce our sample size from 102,829 to 85,669 individuals. Thus, 16.7% of welfare recipients must be considered ineligible for work. This proportion is slightly below that observed in Quebec or British Columbia [See Lacroix(1999)].

Table 1 provides basic information on the sample used in this study. The sample sizes of various socio-demographic groups are presented in the second column. The third column shows the total number of spells for each group and the last column reports the corresponding average number of spells. Single parents and families with children constitute the main household type (46% of total). Individuals with secondary schooling and those living either in Central or Eastern Newfoundland are by far the most important groups (72.2% and 71.6% of total, respectively). As shown in the last column, there is considerable recidivism in the data. Individuals in our sample have experienced on average 2.9 spells between 1986 and 1998. Single women aged 30 and over, single parents and families with children return to welfare sooner than others. Similarly, individuals with primary schooling or those living in Labrador have higher return rates than average.

Table 1 - Sample Statistics

	# Individuals	# Spells	Spells/Individual
Household type
Single Men 18-24	16,615	38,599	2.32
Single Men 25-30	3,968	9,921	2.50
Single Men 30 +	6,964	18,290	2.63
Single Women 18-24	6,262	12,339	1.97
Single Women 25-30	1,361	3,265	2.40
Single Women 30 +	3,908	13,009	3.33
Single Parents	23,843	75,356	3.16
Families no Children	3,187	9,016	2.83
Families with Children	19,561	68,702	3.51
Total	85,669	248,497	2.90
Level of schooling
Primary	16,634	58,559	3.52
Secondary	61,872	175,436	2.84
Post-Secondary	3,358	6,905	2.06
University	3,805	7,597	2.00
Total	85,669	248,497	2.90
Region of residence
East	33,417	94,970	2.84
Central	27,954	80,874	2.89
West	19,830	57,359	2.89
Labrador	4,468	15,295	3.42
Total	85,669	248,498	2.90

To look further into the extent of recidivism, Figure 1 reports the number of individuals who experienced various numbers of spells over the 1986-1998 period. Of the 85,669 individuals in our sample, 29,169 had only 1 spell, 20,310 had 2 spells, 11,668 had 3 spells, etc. Recall that at least 2 months without receipt of benefits are necessary to identify return spells. So it is very unlikely that such recidivism is simply a statistical artefact or the result of coding errors.

In studies of welfare dynamics in Canada it has been found that entry into and exit from welfare usually depict strong seasonal patterns [See Fortin and Lacroix (1997)]. Figure 2 plots the monthly distribution of fresh starts for the whole period. There are four striking features in this figure. First, it is clear that entry into welfare follows a seasonal pattern. Within each single year entry rates follow a similar pattern. Second, January and March witness more entries than any other month. This holds for nearly each year in the figure. Third, aggregate business cycles appear to have a definite impact on the level of entries. For instance, entry rates were at their lowest in the year 1989. Fourth, the months of January 1986, March 1991 and January 1992 witnessed an unusually high level of entries into welfare. These sharp increases coincide with the very poor economic conditions that prevailed at the beginning of 1986 and 1992 as measured, say, by the provincial unemployment rate. On the other hand, the period between 1990 and 1994 witnessed a sustained high level of entry into welfare.

Just as entries into welfare are closely linked to the overall economic conditions, exits from welfare should similarly follow a procyclical pattern and possibly a seasonal pattern. Figure 3 plots the distribution of exits from welfare for the whole period. For each single year depicted in the figure exits rates are at their highest in February and decline gradually until December. It is readily apparent from the figure that the exit rates are both strongly related to general economic conditions (level of exits) as well as seasonal fluctuations in the business cycle (shape of monthly exits).

• ¹In this study a welfare spell is defined as an uninterrupted sequence of months during which benefits are claimed. A single month without receipt is not considered an interruption. At least two months are necessary. This is a common assumption made when working with administrative data [see Barrett and Cragg (1998)].

Previous

Contents

Next