Directorates
·	Director General's Office
·	Data Development
·	Strategic Evaluations
·	Partnership and Continuous Evaluations
Publications
·	Evaluation Reports
·	Evaluation Briefs
·	Labour Market Development Agreement Series
·	Lessons Learned Series
·	Evaluation Tool Kit Series
·	Evaluation Forum Newsletter
·	From Theory To Practice

3.1 Introduction

The switch from the dependency rate (DR) to the family supplement (FS) involves changes in 1) how eligibility is determined and 2) how the value of the family-related benefits is derived. Under Unemployment Insurance (UI), individuals were eligible to receive the DR if they were eligible for regular benefits, if they had dependants and if their average weekly earnings were less than half the weekly maximum insurable earnings (MIE). Under Employment Insurance (EI), total household income is taken into account in determining eligibility for the FS because one of the objectives is to make the FS more targeted to poor households. Regular EI claimants with children are entitled to receive the FS if total household income is less than $25,921 and the household receives the Child Tax Benefit (CTB). If both spouses are claiming EI benefits at the same time, only one may receive the FS.

In terms of determining the value of the benefits, under UI, claimants who were eligible for family-related benefits were entitled to receive benefits at 60 percent of weekly insurable earnings — referred to as the “dependency rate” — rather than the usual replacement rate (of 55 percent of weekly insurable earnings). Under EI, eligible claimants receive the regular EI benefits at the usual 55 percent replacement rate but, in addition, receive the weekly equivalent of the CTB. Since the CTB increases with number of children, so will the FS. This extra payment increased the effective replacement rate to a maximum of 65 percent of insurable earnings in 1997, and it will gradually increase to 80 percent by January 2000; however, total benefits cannot exceed the maximum weekly benefit of $413. The value of the FS benefits changes as the CTB benefits change, and CTB amounts are calculated at tax time, based on previous year’s income with the changes implemented the following July. Thus, there can be considerable lags in the CTB adjustment if income varies substantially from year to year. Another associated benefit is that individuals receiving the FS are exempted from the new intensity rule, which reduces benefits if the claimant has received more than 20 weeks of regular benefits in the past five years.⁸

3.2 Who Qualifies for the FS? A Descriptive Analysis

To examine the impact of the switch from DR to the FS on who qualifies for family-related additional benefits, we start with a descriptive analysis (see Table 1 for the results). Although the FS was introduced in January 1997, other aspects of the new EI program were initiated on July 1, 1996. Thus, a comparison of UI/DR with EI/FS is “cleanest” if we compare cohorts 1 through 4 (with job separations between July 1995 and June 1996) with cohorts 7 through 10 (with separations between January 1997 and December 1997). In terms of Table 1, this is a comparison of columns 1 and 3. However, an examination of columns 2 and 3 is also interesting, since this enables a comparison of all pre-1997 (and hence pre-FS) cohorts with all post-1997 cohorts.

The switch to the FS has reduced overall access to family-related additional benefits, compared with the situation under UI. For example, the percentage of workers with a job separation who receive DR/FS decreased from 8.7 (in the UI period) to 4.1 (post-January 1997); and if we consider only the workers with a job separation who have at least one child living in the house, then the percentage receiving DR/FS decreased from 21 to 11.6. A comparison of all pre-1997 cohorts (including the transition EI period) yields the same conclusion, though the magnitude of the reduction in family benefit receipt is smaller. Both comparisons yield statistically significant reductions in the percentage of recipients (at the 95 percent level).

The decline in the percentage of workers receiving the FS compared with the percentage receiving the DR reflects a combination of 1) decreased access to the FS (compared with the DR) for those workers receiving regular UI/EI benefits and 2) decreased access to EI regular benefits (compared with UI regular benefits).⁹ With respect to the first point, note, for example, that of workers receiving regular UI/EI benefits and have at least one child less than 18 years old in the house, 33.6 percent received the DR in the UI period compared with 20.8 percent who received the FS. Secondly, however, the decrease in access to the FS (compared with the DR) for all workers with a job separation also stems from the decrease in access to regular EI benefits compared with UI benefits. For example, the percentage of all workers with a job separation who received regular UI/EI benefits decreased from 54.6 to 50.9; for workers with at least one child and who had a job separation, the percentage that received regular UI/EI benefits decreased from 62.6 to 55.7 (see Table 2). The differences in the percentages of the groups under the two policy regimes are statistically significant (at the 95 percent level of confidence) in each of the cases. However, notice that the reduction in regular benefit receipt is substantially larger for recipients with children than for recipients over all.

Tables 3a and 3b provide a basic descriptive analysis of who receives family top-ups under the two regimes. Table 3a asks this question for all respondents with children, whereas Table 3b focuses on respondents who have children and who receive regular benefits. Thus, in Table 3a, an individual may not receive the family benefit because he or she was not eligible for regular benefits or because, while eligible for regular benefits, he/she did not qualify for the top-up. In Table 3b, all individuals are regular benefits recipients.

Qualitatively, much the same story about change in access to benefits is told in each table, thus we focus the discussion on Table 3a (with some cross-reference to Table 3b). First, it is striking that receipt of the family top-up has decreased under EI/FS for almost all categories considered (e.g., including province, age, wage, income). Exceptions are men, single parents and those with higher personal annual earnings. (For lower-income households, there is basically no change.)

In terms of annual household income,¹⁰ Table 3a indicates recipiency rates that are nearly identical in both periods for those with household incomes of less than $20,000 (31.8 percent with the DR and 30.2 percent with the FS). Part of what has happened is that fewer individuals from households with low incomes receive regular benefits under EI; however, if an individual from a low-income household manages to qualify for regular benefits, he or she is somewhat more likely to receive family top-up under EI/FS (56.9 percent versus 51.6 percent, Table 3b). In contrast, recipiency rates are much lower for higher-income families (more than $50,000) under the FS (1.7 percent for the FS versus 11.5 percent for the DR, Table 3a). Thus, as intended, the structure of the FS is more targeted to family income, with significantly more lower-income families with children than higher-income families with children receiving benefits. At the same time, if we compare the FS with the DR, it is not true that more lower-income families receive benefits under the new program.

As anticipated, there is a large decline in the percentage of women receiving the FS; for example, under UI, 33.6 percent of women received the DR, compared with only 13.8 percent of women who collected the FS under EI. This contrasts with the experience for men: while less likely to receive family-related top-ups in general, men actually experienced a small increase in access to family-related benefits as a result of the move from the DR to the FS (from 8.2 percent to 9.1 percent) The drop in eligibility for women is probably due to a decline in eligibility among women with spouses who are in the labour force, since there is also a large drop in the percentage of people collecting the FS in households where both adults are in the labour force (the percentage dropped from 21.7 to 6.1). Note also that for single mothers access to family-related benefits increased with the switch from UI to EI. Under UI, 30.3 percent of single mothers, received the DR, whereas under EI, 38.3 percent of single mothers received the FS. On the other hand, 34.4 percent of married mothers received the DR, whereas only 6.5 percent received the FS.

Tables 3a and 3b also illustrate changes in receipt of family top-up by hourly wage¹¹ and by annual earnings.¹² In Table 3a, those with very low hourly wages are much less likely to receive the FS than the DR (the percentage declines from 43.9 to 19 for those with hourly wages of less than $7 per hour). Similarly, those with very low annual earnings (less than $10,000) were more likely to receive family top-up under UI than under EI (40.3 percent versus 8.2 percent).

Table 4 changes the focus of analysis from the probability of getting the DR or FS to the composition of the population of workers who do receive these benefits and how this has changed since the FS was introduced.

FS recipients are more likely to be in the lower-income categories than DR recipients. For example, only 21.6 percent of FS recipients had household incomes greater than $35,000,¹³ while, 46.6 percent of DR recipients had incomes above this level. Again, as intended, the targeting structure of the program has changed.

In terms of annual earnings, there are more relatively “high earning” individuals receiving the FS than the DR (31.8 percent of FS recipients have annual earnings over $20,000 versus 16.5 percent of DR recipients). The basic point is that when cutting off access based on income rather than earnings, some individuals who would not have qualified for the DR (because their personal earnings were too high) now qualify for the FS (because their household income is sufficiently low).

Under either regime, women constitute the majority of benefit recipients. However, the gender composition of family benefit receipt has changed markedly since the introduction of the FS. Under UI, 80.8 percent of DR recipients were women, whereas under EI, only 62.6 percent are women. Couples with both spouses in the labour force make up a smaller fraction of the family top-up recipient pool post-1997 (34.0 percent versus 66.1 percent under the DR). Consistent with results from Table 3, there is an increase in the percentage of family top-up recipients who are single mothers (40.1 percent for the FS versus 16.7 percent for the DR). There is a reduction in the percentage of recipients who are married mothers (64.1 percent for the DR versus 22.6 percent for the FS).

3.3 Multivariate Analysis of the Probability of Benefit Receipt

We turn next to consider how the switch to EI affects who qualifies for family-related benefits, after controlling for the characteristics of workers with dependents and who experience a job separation. We estimate a probit model of the probability of receiving the DR (pre-January 1997) or the FS (post-January 1997) where the key explanatory variable (post-1997) is a dummy variable indicating whether the individual’s job separation occurred after January 1997, in which case the FS rather than the DR would be the relevant policy (here, having a job separation pre-January 1997 is the base case). A set of seasonal, age and education dummies are also included, along with the regional unemployment rate,¹⁴ the hourly wage at the time of the sample job separation, household equivalent income¹⁵ (at the time of the Record of Employment (ROE)), and a household income dummy for having monthly income greater than $9,000 (in the four-week period preceding the Canadian Out of Employment Panel (COEP) interview). Notice that by including regional unemployment and seasonal dummies, we control for changing labour market conditions over the period of the study.

A second model is estimated in which the gender, wage and income variables are interacted with the dummy variable representing the policy switch in January 1997¹⁶ to test our central hypotheses: 1) that the FS should be more targeted to family income but that, 2) more women than men may lose access to benefits as a result of the family income-test. Both models are estimated for all workers who have dependents and who experience a job separation; results are presented in Table 5.

The main result from the probit analysis is that the post-January 1997 variable is negative and significant, indicating that the probability of receiving the FS is lower than the probability of receiving the DR (see Table 5, column 1).

The first key issue addressed in this paper is that of targeting: is the FS more targeted to low-income families than the DR? Under either regime, families with higher incomes are less likely to receive family top-ups (the sign on household equivalent income is statistically significant and negative). In addition, the household equivalent income variable interacted with the post-January 1997 variable is statistically significant and negative (Table 5, column 2); this indicates that higher-income households are even less likely to receive the FS than the DR and that the FS is more targeted to household income than the DR.

We then consider whether women lose disproportionately as a result of the switch to the FS. Specifically, we examine whether women with similar eligibility based upon their own work and earnings have reduced access to family-related benefits, relative to men, as a result of using household income rather than individual earnings as the basis for determining eligibility for the family-related top-up. The “female” variable interacted with the post-January 1997 dummy variable is negative and statistically significant (see Table 5, column 2), indicating that women’s access to the family-related top-up benefit is reduced, compared with men’s, as a result of the policy change to EI.

Since women are, relative to men, less likely to receive the FS (after the January 1997 policy), we explore which women are most affected by estimating a probit model that replaces the “female” variable with “single mother” and “married mother” (with the base category being all men).¹⁷ The results shown in Table 5, column 3, are similar to those of column 1. For example, the probabilities of receiving benefits are greater for single mothers and married mothers than for the base case of fathers; both the single mother and married mother variables are positive and significant. This result is consistent with the result reported in column 1 that the probability of receiving family-related benefits is higher for women. When the two mother variables are interacted with the post-January 1997 variable (see Table 5, column 4), the interaction is statistically significant and negative for married mothers, indicating that married mothers, compared with fathers, have reduced access to the FS, after controlling for household income. The interaction term for single mothers is not statistically significant, indicating that once we have controlled for other relevant characteristics, single mothers experience no change in the probability of receiving family top-up after the program change.

3.4 Has the FS Increased the Level of Benefits Received? A Descriptive Analysis

While the switch from UI to EI has reduced access to family-related top-up benefits, we now consider whether the replacement of the DR by the FS has increased the value of family-related top-up benefits for those who receive them, as this was a stated intention of the program reform. The mean weekly DR is $13.37 compared with an FS of $28.10, which is a large and statistically significant increase (see Table 1). After the switch from UI to EI, the average weekly value of regular benefits received is also higher for workers receiving the family-related top-up ($166.79 versus $201.75, Table 2). This is very likely a result of the change in the composition of workers who actually receive benefits — that is, a larger percentage of high-earnings individuals receive the FS than receive the DR (see Table 4).¹⁸

Table 6 shows the mean weekly value of the DR and FS for those receiving these benefits, by population sub-group. For recipients, the switch to the FS results in a doubling of the mean weekly value of family-related benefits for each of the different groups considered including province, gender, age, household composition, workforce status, wage and household income. However, surprisingly, the mean weekly value of the FS does not decrease as household income increases. For example, the mean weekly value of the FS is $27.39 for households with incomes of less than $20,000 annually, compared with $29.93 for households with an incomes of $50,000 or more. As noted previously (see footnote 13), it seems odd that any household with an income of $50,000 or more should be receiving benefits under the FS. To understand this result, it is important to recall that eligibility depends upon last year’s income, so that a household with low income last year but higher income this year could receive benefits.¹⁹ Thus, the structure of the FS is more targeted than the DR in terms of receipt of benefits but not in terms of benefit payments. However, the FS does pay more money to lower-income families who receive benefits.

For recipients, the level of benefits has increased by roughly the same amount for both men and women, and for married mothers versus fathers or single mothers. Hence, the potential within-household equity issue is more a question of who gets benefits rather than how much is received.

3.5 Multivariate Analysis of Level of Benefits Received

We turn next to a multivariate analysis of whether, compared with the DR, 1) the value of the FS is higher, 2) the value of the FS offers more income protection to low-income households than high income households and 3) the value of the FS is relatively lower for women than for men. A model of the value of the weekly family-related benefit received is estimated using Ordinary Least Squares, controlling for age, season, provincial unemployment rate, education, gender, own wage and household income. The model is estimated for all workers receiving family-related benefits in the two periods, pre- and post-January 1997, and the results are presented in Table 7.

To assess whether, for families receiving benefits, the level of FS benefits is higher than that of DR benefits, the key explanatory variable is the post-January 1997 dummy variable. As can be seen from Table 7 (column 1), the post-January 1997 variable is positive and statistically significant, indicating that the value of the weekly FS is greater than that of the DR, after controlling for other factors such as household income and own wage.

To assess whether the level of benefits is more sensitive to household income under the FS than under the DR, the post-January 1997 dummy variable is interacted with household equivalent income. As can be seen from Table 7 (column 2), the interaction of the post-January 1997 and household income variables is not statistically significant; this indicates that, after the switch to EI, the value of family-related benefits is no greater for low-income households who receive benefits. As noted earlier, the improvement in targeting is evident in terms of who gets benefits, but not in terms of how much they receive.

Notice that the female variable is negative and statistically significant, indicating that women, relative to men, receive a lower weekly DR/FS. To test whether the move to a family income-tested benefit has had an additional adverse affect on the level of benefits received by women relative to men, we also included an interaction between the post-January 1997 and female dummy variables. This term is negative and statistically significant, indicating that women receive a lower value of family top-up relative to men, after the switch to EI, controlling for household income and personal hourly wage. Thus, relative to men, women’s access to family-related benefits and level of benefits is lower after the program change.

We again separate “women” into married and single mothers and present results for this specification in Table 7, column 3. The single mothers and married mothers variables are negative and statistically significant, see Table 7 (column 3), indicating that these two groups, relative to all fathers, receive a lower value of benefits. This result is consistent with the finding reported in column 1 that the female variable is negative and statistically significant. When the “mother” variables are interacted with the post-January 1997 variable, the married mother interaction term is statistically significant and negative. Thus, married mothers are particularly likely to receive a lower value of family benefits in the post-1997 period, presumably because many of them are married to men with earnings sufficiently high to make the household ineligible for family-related top-ups. The single mother interaction is not statistically significant.

3.6 Combining Incidence and Level of Benefits

Results thus far have separated issues of who receives FS/DR benefits from the level of benefits received. The conclusion reached is that, in general, fewer job separators with children receive benefits with the FS, but for those who do receive benefits these are higher than with the DR. Table 8 attempts to combine these issues by reporting the average level of benefits for all individuals who have children and who experienced a job separation, regardless of whether they received UI/EI or the DR/FS. This provides an estimate of expenditure per person to compare across regimes.²⁰

Key points to notice are, first, that overall, the expected value of family benefit top-up for job separators with children has increased slightly from $2.81 to $3.26 (this is a statistically significant change, though 45 cents per week does not sound like very much of a difference; see Table 1).

Second, the expected value of benefits for low-income families (i.e., less than $20,000) has increased more noticeably (from $4.20 to $8.28). Thus, FS is directing more money to lower-income families. On the other hand, on average less is spent per woman ($3.68 versus $4.31), while more is spent per man ($2.80 versus $1.29). If we distinguish single mothers from married mothers, we see that considerably more is spent on single mothers ($10.52 versus $3.79), but less is spent on married mothers ($1.64 versus $4.44).

Table 9 provides a multivariate analysis of the same issue. In this case, since we are including individuals with zero benefits, Ordinary Least Squares is no longer the appropriate estimation technique, so we use Tobit analysis instead. In specification 1, we find that our key “post-January 1997” variable is not statistically significant in an estimation of the expected value of family-related top-up. That is, controlling for other relevant factors, the FS does not increase average expected per-person expenditures in a specification that models the impact of the policy change through a single dummy variable.

When we allow for interactions with “female” and “household equivalent income,” we find that expected per-person expenditures are more targeted to income with the FS than with the DR (i.e., the household equivalent income interaction term is statistically significant and negative), but that expected payments to women fall significantly with the FS compared with the DR. If we further distinguish single mothers from married mothers, we find no statistically significant effect of the policy change on expected payments to single mothers, but a sizable reduction in expected payments to married mothers.

3.7 What Is the Impact of the FS on Standard of Living?

As discussed above, the switch from UI to EI results, first, in fewer individuals receiving family-related benefits and, second, a higher value of family-related benefits for those who receive them. We now examine the impact of this policy reform on the standard of living of low-income families with unemployment who receive the family-related benefit. To do this we focus upon 1) the incidence of low-income with and without benefits and 2) the depth of low-income with and without benefits. Since the FS is intended only to improve the standard of living of poor families, and not to lift their incomes out of poverty, we would not necessarily expect to see any impact of the FS upon the incidence of poverty; however, we would expect to see at least a small reduction in the depth of poverty. Table 10 shows the incidence and depth of monthly²¹ poverty in the UI and EI periods, for several income definitions and population groups. The poverty line applied in the calculations is a one-half median equivalent income definition.²² The first income variable, “Income before the ROE,” is household income before the spell of unemployment begins and is used to reflect the extent of poverty just before the spell of unemployment happened. For example, before experiencing unemployment, 15.5 percent of all households with dependant children were poor prior to January 1997, compared with 16.2 percent in the post-January 1997 period.

We consider whether the policy reform has decreased the incidence of poverty among those households that receive the family-related benefit while they are still experiencing unemployment, acknowledging that fewer families now receive benefits (regular or top-up). The second income variable, “Income before the interview,” is household income at the time of the interview and reflects the household income of individuals who are still unemployed, as well as that of individuals who have started a job, at the time of the interview. Separating out those individuals who are still unemployed at the time of the interview gives an estimate of the incidence of poverty among those individuals who are still experiencing unemployment. For example, as shown in Table 10, for unemployed workers who receive the family-related benefit, 38.3 percent are poor during the UI period compared with 60.1 percent during the EI period. Notice that the increase in the rate of poverty reflects the change in the composition of households that receive family-related benefits.

Family-related benefits are not large enough to lift households out of poverty in either of the policy periods, nor was it expected that this program alone could achieve such a goal. For those households receiving family-related benefits and still experiencing unemployment, the rates of poverty when total household income includes or excludes the family benefit are not significantly different. For example, in the UI period, 38.3 percent of households receiving the DR are poor compared with 40.4 percent who are poor when the value of the DR benefits are excluded, and this is not statistically significant. In the EI period, 60.0 percent of households receiving the FS are poor compared with 68.2 percent who are poor when the value of the FS is excluded from household income. While this is a larger reduction, it is still not a statistically significant difference.

It is more reasonable to suppose that the FS should significantly alleviate the degree of financial hardship experienced by poor households. That is, the program is not intended to eliminate poverty, but since it is more targeted to low-income households, we might expect low-income families to have incomes that are higher in a statistically discernible way. However, this does not appear to be the case. For example, for those workers who receive the FS and who are still unemployed, the average equivalized monthly poverty gap was $201 compared with $221 if the value of the FS benefits are excluded, but this is not a statistically significant difference.

Finally, Table 11 reports on the availability of liquid assets for families with children and who do and do not receive top-ups. It is clear that recipient families have lower assets than other families and that, with the reduction in the number of individuals receiving top-ups, these top-ups are now going to families with very little in the way of funds. About 60 percent of individuals who have children and who experience job separations would not have sufficient assets to enable their families to survive for even one week at a poverty-level standard of living. For DR recipients, that figure increases to 71.0 percent and for FS recipients, to 84.2 percent.

Footnotes

8	This is a potentially important benefit for FS recipients; however, it is really too early to assess the implications since all claimants were granted a “clean slate” of past unemployment history at July 1, 1996. Thus, for the period of our data (ending with job separations at December 1997), no-one would yet have a five-year history. A rough calculation suggests about 20 percent of FS recipients would have been affected by the intensity rule but were exempted. However, this percentage is likely to increase over time.
9	In this descriptive analysis, no control is made for potentially changing economic conditions over the period of study. We are able to control for changing unemployment in the multivariate work.
10	The COEP survey asks respondents their total household income in the four-week period preceding the interview. This is to include income from all sources (e.g., earned income, pensions, interest, dividends, rents, net farm or business profits, and government benefits such as welfare, family allowance and unemployment insurance payments). They are also asked if this is an increase or decrease from their total household income in the four weeks preceding the sample job separation. If there was a change, they are asked how much. We use this information to calculate household income in the four weeks preceding the sample ROE and then multiply this figure by 13 to obtain an estimate of annual income.
11	Hourly wage is obtained from the administrative “employer” file. It is the hourly wage reported for the sample job (i.e., the job from which the individual experienced a separation and so ended up in the survey).
12	Our estimate of annual paid wages is derived from the COEP survey. For a very small number of individuals, self-employment earnings will be included.
13	While it sounds odd that any households with income at this level are receiving the FS, there are two explanations for this result. First, recall that the COEP data only provide income information for a four-week period, which we have annualized. Thus, we assume that they continue to receive for all 12 months the income that they report for the one month. Of course, this need not be the case — the month we observe may be a particularly good or bad one. A second reason for this finding is that the FS is tied to the CTB, and thus entitlement is based upon last year’s income. Given the substantial adjustment lags associated with the CTB, it is entirely possible that a family that had a lower income last year will still receive benefits this year when their income is higher and vice versa — families whose income drops will not receive benefits until after taxes have been filed and all adjustments made.
14	This varies by province and by year (1995 through 1997).
15	Household income is adjusted for economies of scale available to individuals who live together using the OECD equivalence scale.
16	We tested interactions with all other explanatory variables but found the ones that we focus upon here to be most important.
17	The number of single fathers was insufficient for separate analysis.
18	Another factor may be that FS recipients are not subject to the “intensity rule.” As noted earlier, a rough estimate suggests about 20 percent of regular benefit claimants were subject to the intensity rule; about 20 percent of FS recipients would have been subject to this penalty but were exempted.
19	We should again point out that the only way to estimate annual income is to extrapolate from a four-week estimate provided in the data.
20	We might think of this as the “expected value” of family top-up benefits — the probability of getting benefits times the amount you will receive if you get them.
21	The COEP survey does not have ideal measures of income for an analysis of poverty, which is typically studied on an annual basis. As noted earlier, respondents are asked about their total household income in the four-week period preceding the interview. They are also asked if this differs from total household income in the four-week period preceding the ROE job separation, and by how much. For individuals who received UI/EI and the DR/FS in the four weeks preceding the interview, we can compute the impact of these benefits on their experience of “monthly” poverty. For individuals who were not receiving UI/EI at the time their household income was assessed, we have to estimate the likely effect of UI/EI on poverty. The procedure used to do this is 1) subtract earnings from the ROE job from household income, 2) add in EI/UI and 3) assume no other change in family income (e.g., spouse obtaining employment; individual working for a few hours). We think it best to estimate “monthly” poverty by dividing an annual poverty line by 12, so that we can minimize any extrapolations required.
22	Equivalent income is calculated using the Organization for Economic Corporation and Development (OECD) equivalence scale. Since we only have income for four-week period, we divide the annual poverty line, calculated as 50 percent of median equivalent income using Survey of Consumer Finance data, by 12.