1. Are People Smarter than We Give Them Credit for
Are People Smarter than We Give Them Credit for? The case of Mandatory Pensions in Israel:
Unintended (and predictable) Consequences of Policy
Adi Brender
Research Department Bank of Israel
Seminar on: Aging, Retirement and Pensions: Trends, Challenges and Policy
Leonardo Hotel Ashkelon: March 2018
The opinions and analysis presented in this lecture do not necessarily reflect those of the Bank of Israel

2. Background story: Old-age poverty
High old-age poverty rate – 22 percent – much higher than the OECD average.
Low pension eligibility among low-income retirees and low savings among low-wage employees.
Policy-makers mandated pension savings.
However: the ratio of old-age poverty to total poverty in Israel is similar to the OECD average.
Hence the problem lies in the overall income distribution – not in the pension system.
The forced savings hurt poor working families.

3. Outline
Ex-ante analysis (based on pre-policy data) showed that a mandatory pension arrangement in Israel will have negative implications for the targeted population
Impede life-time income smoothing and reduce benefits for low-income workers?
Reduce lifetime benefits of the working poor relative to non-working households.
These negative effects depend on the employer contributions’ incidence and the employment effect?
Behavior during the arrangement’s first year - when enforcement was lax - as an indication for preferences.
Use ex-post medium-term data to estimate the "tax" incidence and employment effect.
הצגתי ב-2009 מאמר על המצב שלפני ההסדר וההשלכות הצפויות של ההסדר.

4. The Policy’s Rationale
People do not save enough for pension
Short-sightedness
Psychological aversion to deal with being old
Complexity of decision
Too late to fix when the consequences are realized
Cannot call upon past employers
Fiscal cost
Prevent “taking advantage” of the system
However:
If people are rational, honest and optimizing, intervention may be harmful to welfare.

5. What do we test and what we do not
Examine responses at the margin, given the pre-existing social security and tax benefits.
Do not argue that no intervention is warranted.
Argue that people responded rationally to the “old” system and to the change.
This seems to be a case of “too much” savings for the affected population.

6. The Mandatory Pension Arrangement
Contribution rates according to the arrangement
Beginning from
Employer Contributions
Employee Contributions
Total Contribution
1.67
0.83
2.50
3.34
1.66
5.00
7.50
6.67
3.33
10.00
8.34
4.16
12.50
15.00
12.00
5.50
17.50
The duty applies to income up to the average wage (70% of all employees earn less),
for people who worked at least 6 months. Those who had previous coverage – from day 1.
Exempt individuals – in 2008: those who worked less than 9 months for the same employer.

7. Pension Savings Incentives in Israel
Tax incentives at the contribution and withdrawal stages – relevant to those above the tax threshold.
Yields on accumulated savings are exempt from capital gains taxation – relevant for those who want to save.
Social security replacement rates are high for low income employees – but depend on family characteristics.
Social security income guarantee is offset against pension benefits.

8. Implications for the Incentives to Save
Characteristics Affecting the Decision to Begin to Contribute
Characteristic
Cause for behavioral effect
Effect
Low wage
Sufficient replacement rate through old-age allowances and income guarantee
(-)
No tax benefits on withdrawal
Wage below the tax threshold
No tax benefits at the time of contribution
Married woman
Insufficient replacement rate through old-age allowances as spouse is likely to work, even if he does not work currently.
(+)
Working spouse
Insufficient replacement rate through old-age allowances. No offset of allowances against pension.
Spouse contributing to pension
Additional contribution is unlikely to be offset against the old-age income supplement
Female
Even if currently unmarried, expected to have a working spouse later in her career.
Children
Consumption smoothing, liquidity.
Older age - no past contribution
Insufficient accumulation to overcome the pension offset against the old-age income guarantee
Arab
Unlikely to have a working spouse, especially if currently single or has a non-working spouse
Small Employer
Worse terms in pension funds

9. The Targeted Population (Brender, 2010)
About 38% of all employees did not contribute to pension in 2007: their types are consistent with the incentives

10. Characteristics of the relevant population
75% of those who reach retirement live with a spouse.
Little change in relative income positions after age 30-35; current wages from this age are a good proxy for lifetime income.
Wages of low income people grow slower.
Spouses incomes are positively correlated.
90% have children during their life, 80% more than 1 child.

11. Characteristics of typical households
30%-35% 6%
20%

12. Consequences of the Arrangement for the Target Population
Impedes income smoothing:
Poverty during working/parenting years
Excessive replacement rate
Reduces lifetime income:
Offset of income guarantee
Employer contribution incidence?
State benefits (tax + Soc. Sec.) at all income levels were similar before the arrangement was adopted.

15. 2008 as a window to observe preferences
Lack of clear enforcement mechanism
51% began to contribute, compared to 17% in 2007
By 2012 more than 80% contributed
Mobile target population (between employers)
Large sample – 80,000 non-contributors
Panel data: behavior before and after

16. Three Tests
Consistency of the decision to begin contributing with the a-priori incentives
Did the impact of these variables decline compared to pre-arrangement years?
Match between factors affecting beginning to contribute and contribution rates
Helps to sort obedience from preferences
Did wages fall to compensate for employer contributions?

17. Factors affecting the decision to begin contributing: 2007-8 compared to 2006-7z
Marginal effect
between 2006
and 2007
between 2007
and 2008
Individual characteristics: *
2.99
0.0120
18.40
0.1159
Gender (0 – men, 1 – women)
-11.71
-21.44
Resides in an Arab town (binary variable)
2.95
0.0031
8.08
0.0128
Age
-5.12
-9.58
Age squared
-1.56
-1.25
Married man (binary variable)
4.47
0.0242
8.90
0.0719
Married woman (binary variable)
-3.36
-4.02
Number of children aged 0-3
-5.72
-6.04
Number of children aged 4-8
-5.34
-7.79
Number of children aged 9-18
Change in incentives for young workers: the age effect is decreasing from 35. It decreased at all relevant ages prior to the arrangement.

18. Income and Employer Characteristics
Marginal effect
between 2006
and 2007
between 2007
and 2008
Income and employment characteristics: *
0.0188
0.0046
Annual income (in 10,000s of NIS)
0.0777
0.1254
Annual income >48,000 – tax threshold (binary variable)
Spouse characteristics:
0.0851
0.1032
Does spouse work? (binary variable)
0.0409
0.0996
Does spouse contribute to pension savings? (binary variable)
Employer characteristics:
0.0000
0.0001
Size of employer (number of employees)
Up to 15 workers (binary variable)
15-30 workers (binary variable)
30-50 workers (binary variable)
***
Switched employer between the two years (binary variable) **
0.0272
Switched employer*worked less than 9 months (binary variable)
0.0590
0.1234
Employed in the public sector (binary var.)

19. Contribution Rates
More than 60% of those who began saving in 2008 did so at the minimum mandated rate of 0.83%, compared to only 18% in 2007.
Beginning to contribute was associated with the factors that make pension savings desirable
If these factors reflect obedience, they should not correlate with saving above the minimum required rate
If they reflect preferences, they should also be correlated with contributing above the minimum.

20. Factors affecting workers starting to save in 2008 to contribute at above-minimum rates
Individual and income characteristics z
Marginal effect
between
2007 and 2008
Individual characteristics: *
-6.27
Gender (0 – men, 1 – women)
5.74
0.0118
Age
-6.56
Age squared
3.13
0.0294
Married woman (binary variable)
-3.76
Number of children aged 0-3
-2.71
Number of children aged 4-8
***
-1.85
Number of children aged 9-18
Income and employment characteristics:
27.76
0.0462
Annual income (in 10,000s NIS)
9.43
0.0715
Annual income >48,000 (binary variable)
-15.88
Wage in 2007 (in 10,000s NIS)
The age effect begins to decrease at 34

21. Spouse and Employer Characteristicsz
Marginal effect
between
2007 and 2008
Spouse characteristics: *
8.75
0.0757
Does spouse work? (binary variable)
5.46
0.0026
Annual income of spouse (in 10,000s NIS)
Employer characteristics:
17.07
0.0001
Size of employer (number of employees)
5.73
Up to 15 employees (binary variable)
6.39
15-30 employees (binary variable)
4.39
30-50 employees (binary variable)
4.80
0.0551
Switched employer between the two years (binary variable)
6.29
0.1223
Switched employer*worked less than 9 months (binary variable)
9.43
0.0820
Employed in the public sector (binary variable)
Those who were exempt and began contributing anyway did so at higher rates

22. Wage and Employment Effects – Medium-term
A key argument of those favoring the mandatory pensions is that most of the burden is imposed on employers.
Given a legal minimum wage, weak employees – the target group - are protected from wage cuts.
Hence, income smoothing is not substantially influenced.
Employment is not very sensitive to labor cost at the low end of the wage distribution due to a high non-tradable proportion.

23. The analysis
Examine the change in wages and employment between 2007 and 2012 (when the share of contributors was equalized) of those who worked in 2007.
Compare the changes between contributors and non-contributors in 2007 – controlling for observables.
Since there may be unobservable differences correlated with the economic cycle: compare the dif-in-dif to a parallel cyclical period in Israel.
Employment effects: hard to detect due to low labor supply elasticity ( ) of the relevant group.

25. Key Results
(Conditional) Wages of those who did not contribute increased between by 7.5% less than the wages of the employees that did contribute.
During the gap was 3%.
The difference of 4-4.5% is equivalent to the employer contribution (net of severance pay fund) at 2012.
The results are robust to changes of period, income and age ranges, ethnicity and gender.
Occupation and industry composition do not account for the difference.
Negative employment effect is not statistically discernible, perhaps due to its small size.
Tests of industry and occupation – which are not reported – did not show a significant difference.

26. Conclusion
Ex-ante analysis suggested that the mandatory pension arrangement harms the target group.
Ex-post analysis finds a behavior that is consistent with employees trying to avoid contributing when it was possible – in line with incentives.
Both the decision to contribute and contribution rates suggest rational optimization by employees – in line with objective simulations.

27. Conclusion – cont.
There was essentially no relief via employer contributions; almost the entire incidence fell on the employees.
Given low labor supply elasticity: most of the burden fell on wages – not employment.
These results were predictable based on ex-ante analysis and data that were available in real-time.

28. If it ain’t broken don’t fix it