Job search and Employment Insurance

In the job search model, an unemployed worker who receives wage offers one at a time accepts the first offer at or above an asking wage $w_A$ (the reservation wage of the search model, written $\bar w$ in some texts). The asking wage is set where the marginal benefit of searching one more week equals its marginal cost. Unemployment insurance lowers the marginal cost of search, so it raises the asking wage, the expected wage of the job finally taken and the length of unemployment spells.

The intuition

A laid-off machinist receives an offer of $900 a week. Taking it ends the search. Turning it down costs a week of pay, minus any benefit she collects, but keeps open the chance of a $1,200 offer next week. The higher she sets her standard, the better the job she ends up with and the longer she waits for it. The asking wage is the standard at which one more week of waiting stops paying.

The model

Assumptions:

• The worker knows the wage offer distribution, the range of wages employers might offer and how likely each is, but not which employer offers what.
• Each week an offer arrives with probability $\lambda$. Offers are independent draws from the distribution.
• While unemployed she receives a benefit of $b$ a week.
• She accepts any offer at or above her asking wage $w_A$ and rejects any offer below it.

This is sequential search: the worker decides offer by offer. In nonsequential search she commits in advance to sampling a fixed number of employers and takes the best offer among them; each extra employer raises the expected best offer by less than the one before, so she samples until that gain equals the cost of one more application. Sequential search does at least as well, because a worker who draws a good offer early can stop.

Let $p$ be the share of offers at or above $w_A$. Each week the search ends with probability $\lambda p$, so

$$\text{expected weeks of search} = \frac{1}{\lambda p},$$

and the expected wage of the job taken is the average offer at or above $w_A$. Raising $w_A$ lowers $p$, which lengthens the search and raises the expected wage.

Asking wage, w_A (per week)

$1,000

Chance an offer is acceptable

50%

Chance a week of search ends in a job

10.0%

Expected weeks of search

10.0 weeks

Expected wage of the job taken

$1,250

Open in Graph Lab

Figure 1. Offers uniform between $500 and $1,500 a week, one arriving in a given week with probability 0.2. An asking wage of $1,000 accepts half of all offers, giving an expected search of 10 weeks and an expected wage of $1,250; a higher asking wage raises both.

Setting the asking wage

Compare accepting an offer $w$ now with searching one more week.

• The marginal cost of the extra week is the wage given up minus the benefit collected: $MC(w) = w - b$. It rises one for one with $w$.
• The marginal benefit is the expected gain from the extra week. With probability $\lambda$ a new offer $w'$ arrives; if it beats $w$, the worker gains $w' - w$ a week for as long as the better job lasts, $H$ weeks (discounted to today). So $MB(w) = \lambda H \, E[\max(w' - w, 0)]$. It falls as $w$ rises, because a better offer becomes less likely and is better by less.

The asking wage is the offer at which the two are equal:

$$MB(w_A) = MC(w_A) = w_A - b.$$

For any offer below $w_A$, another week of search is worth more than it costs, so the worker rejects it. For any offer above $w_A$, she accepts.

Chance of an offer each week, λ

0.20

Asking wage, w_A

$1,153

Chance an offer is acceptable

35%

Expected weeks of search

14.4 weeks

Expected wage of the job taken

$1,326

Open in Graph Lab

Figure 2. The marginal cost of another week of search, $w - b$ with a benefit of $550, and its marginal benefit with offers uniform between $500 and $1,500. They cross at an asking wage of $1,153; fewer offers shift the marginal benefit curve down, lowering the asking wage but lengthening the expected search.

Where the marginal benefit curve in Figures 2 and 3 comes from

With offers uniform on $[a, c]$, the expected gain from one draw when the worker holds $w$ is

$$E[\max(w' - w, 0)] = \int_w^c \frac{w' - w}{c - a}\, dw' = \frac{(c - w)^2}{2(c - a)}, \qquad a \le w \le c.$$

With $a = 500$, $c = 1{,}500$, $\lambda = 0.2$ and $H = 50$, $MB(w) = 10 \times (1{,}500 - w)^2 / 2{,}000 = (1{,}500 - w)^2 / 200$. Setting it equal to $w - 550$ and solving the quadratic gives $w_A = 1{,}152.8$. This is the reservation wage condition of the standard sequential search model.

What moves the asking wage:

• A higher benefit lowers the marginal cost line by the amount of the benefit and raises $w_A$.
• More frequent offers or longer-lasting jobs raise the marginal benefit curve and raise $w_A$. In a boom the asking wage rises and the search still gets shorter.
• Impatience or urgent needs for cash lower the value of future gains, shift the marginal benefit curve down and lower $w_A$.
• Direct search costs, such as travel and applications, raise the marginal cost line and lower $w_A$.

How the asking wage changes over a spell

With constant $b$ and $\lambda$ the asking wage is constant. Over a real spell three forces push it down: benefits run out (Canadian regular benefits normally last 14 to 45 weeks), savings run down, and offers may arrive less often as skills fade or employers read a long spell as a bad sign. When benefits end, $b$ falls to zero, the marginal cost line shifts up by the lost benefit and the asking wage drops, so the model predicts that job-finding rises as exhaustion approaches. Surveying unemployed workers in New Jersey weekly, Krueger and Mueller found that self-reported reservation wages fell only slowly, by 0.05 to 0.14 percent a week, mostly among older workers and those with savings.¹

Employment Insurance in Canada

Rules for EI regular benefits in 2026:

• Eligibility. A claimant must have lost the job through no fault of their own, have been without work and pay for at least 7 consecutive days, have worked enough insurable hours in the past 52 weeks, be ready and able to work each day, and be actively looking for work.² People who leave a job voluntarily without just cause, or lose it through their own misconduct, may not qualify.
• Entrance requirement. The hours needed depend on the unemployment rate in the claimant’s EI economic region: 700 hours where it is 6 percent or less, 35 fewer hours for each additional percentage point, down to 420 hours where it is above 13 percent.² The regional rates are three-month moving averages of seasonally adjusted LFS rates.³
• Benefit rate. 55 percent of average weekly insurable earnings over the best 14 to 22 weeks (fewer weeks where unemployment is higher), up to $729 a week, because insurable earnings are capped at $68,900 a year. A family supplement can raise the rate to as much as 80 percent for low-income families with children.⁴
• Duration. 14 to 45 weeks, longer with more hours worked and in regions with higher unemployment.⁴
• Temporary measures. The usual one-week unpaid waiting period is waived for claims starting between March 30, 2025 and October 9, 2027, and long-tenured workers can receive up to 20 extra weeks, to a maximum of 65, on claims starting between June 15, 2025 and June 12, 2027.⁵
• Financing. Premiums on insurable earnings: $1.63 per $100 for employees outside Quebec in 2026, with employers paying 1.4 times the employee premium (payroll taxes). The rate does not depend on an employer’s layoff record.⁶

Unemployment rate by province, 2025

Annual average, population aged 15 and over

Canada6.8%

Newfoundland and Labrador10.1%

Prince Edward Island8.0%

Nova Scotia6.6%

New Brunswick7.1%

Quebec5.6%

Ontario7.7%

Manitoba5.9%

Saskatchewan5.2%

Alberta7.2%

British Columbia6.2%

Unemployment rate by province, 2025. EI uses finer economic regions, but at these rates a claimant would need 525 insurable hours to qualify in a region at Newfoundland and Labrador's 10.1 percent and 700 in one at Saskatchewan's 5.2 percent.Source: Statistics Canada, Table 14-10-0327-01, Labour force characteristics by gender and detailed age group, annual. Retrieved 2026-10-23.

Show the data and notes

Annual average of monthly Labour Force Survey estimates, both genders. The Labour Force Survey covers the 10 provinces; it excludes the territories, people living on reserves, full-time members of the Canadian Armed Forces and people living in institutions.

Category	Unemployment rate
Canada	6.8%
Newfoundland and Labrador	10.1%
Prince Edward Island	8.0%
Nova Scotia	6.6%
New Brunswick	7.1%
Quebec	5.6%
Ontario	7.7%
Manitoba	5.9%
Saskatchewan	5.2%
Alberta	7.2%
British Columbia	6.2%

Predicted effects of unemployment insurance

On search. For a worker who qualifies, a higher benefit lowers the marginal cost of search and raises the asking wage. The worker accepts fewer offers, searches longer and takes a better-paid job on average.

Lean benefit, b_L (per week)

$200

Generous benefit, b_G (per week)

$650

Asking wage, lean (A)

$1,080

Asking wage, generous (B)

$1,176

Expected weeks of search, lean

11.9 weeks

Expected weeks of search, generous

15.4 weeks

Open in Graph Lab

Figure 3. Two workers with the same offers, one with a lean benefit of $200 a week (A) and one with a generous benefit of $650 (B). The generous benefit lowers the marginal cost line by $450, raising the asking wage from $1,080 to $1,176 and the expected search from 11.9 to 15.4 weeks.

The longer search has two readings. Under moral hazard, insured workers search less intensively or hold out longer than they would if they bore the full cost, a distortion. Under the liquidity reading, benefits let workers without savings avoid grabbing the first offer out of necessity, which can produce better matches. Chetty estimated that about 60 percent of the effect of US unemployment benefits on spell length is a liquidity effect.⁷

EI also shapes behaviour before a spell starts. Christofides and McKenna found that a significant number of Canadian jobs in the late 1980s ended just when they reached the duration needed to qualify for a claim.⁸ After the large 1971 expansion of Canadian unemployment insurance, workers became more likely to claim once they had used the system, so its effects persisted (Lemieux and MacLeod).⁹

On layoffs. A firm facing a temporary drop in demand can cut wages or hours, or lay workers off and recall them later. If laid-off workers collect benefits financed by premiums that do not depend on the firm’s own layoffs, other employers and workers pay part of the cost of each layoff, which subsidizes temporary layoffs (Feldstein).¹⁰ Experience rating ties an employer’s unemployment insurance tax to its layoff history. US states experience-rate their taxes, but each state’s schedule runs between a minimum and a maximum rate, which limits how far a firm’s tax can rise with its layoffs.¹¹ Studies using variation across US states and industries link weaker experience rating to more temporary layoff unemployment, especially in recessions (Topel; Card and Levine).¹² Canada’s EI premiums are not experience-rated.

Key idea

The asking wage equates the marginal benefit and the marginal cost of another week of search. Unemployment insurance lowers the cost, so it raises the asking wage and lengthens spells; without experience rating, it also subsidizes temporary layoffs.

Worked example

Offers are uniform between $600 and $1,400 a week, one arrives in any week with probability 0.25, and the worker’s asking wage is $1,000.

1. Share of acceptable offers. $p = (1{,}400 - 1{,}000) / (1{,}400 - 600) = 400 / 800 = 0.5$.

2. Weekly chance the search ends. $\lambda p = 0.25 \times 0.5 = 0.125$.

3. Expected weeks of search. $1 / 0.125 = 8$ weeks.

4. Expected wage of the job taken. Accepted offers are uniform between $1,000 and $1,400, so the expected wage is $(1{,}000 + 1{,}400)/2 = 1{,}200$ dollars a week.

5. A more generous benefit raises the asking wage to $1,100. Now $p = 300/800 = 0.375$, the weekly exit chance is $0.25 \times 0.375 = 0.094$, the expected search is $1/0.094 = 10.7$ weeks and the expected wage is $(1{,}100 + 1{,}400)/2 = 1{,}250$ dollars. About 2.7 more weeks of search buy $50 a week more in expected pay.

6. Her EI benefit. If her best weeks averaged $1,000, she receives $0.55 \times 1{,}000 = 550$ dollars a week. Had they averaged $1,500, 55 percent would be $825, above the cap, so she would receive $729, a replacement rate of $729/1{,}500 = 48.6$ percent.

Evidence

Benefit levels and exhaustion. Using individual unemployment spells in the United States, Meyer found that higher benefits strongly reduce the probability of leaving unemployment, and that the exit rate jumps just before benefits lapse; when benefits are extended, the jump moves to the new exhaustion date.¹³ Katz and Meyer estimated that each extra week of potential benefits lengthens the average spell of recipients by 0.16 to 0.20 weeks.¹⁴

What the spike measures. With administrative data for Austrian job losers, Card, Chetty and Weber found that the exit rate from registered unemployment rises by over 200 percent when benefits expire, but the re-employment rate rises by only 20 percent. Many people leave the unemployment register at exhaustion without a job, so spells measured by benefit receipt overstate how much workers time their return to work.¹⁵

Paying people to find jobs faster. In a randomized experiment in Illinois, claimants offered a $500 bonus for finding a job within eleven weeks spent more than one week less on insured unemployment, on average.¹⁶ Search responds to its payoff.

Common mistakes

Traps on exams

• Getting the direction wrong. Unemployment insurance raises the asking wage; it does not lower it. It also raises the expected wage of the job taken, not only the length of the spell.
• Confusing the asking wage with the wage received. The asking wage is the lowest acceptable offer. The expected wage of the job taken is the average acceptable offer, which is higher.
• Assuming fewer offers raise the asking wage. Fewer offers lower the marginal benefit of search and lower the asking wage, yet spells still lengthen because acceptable offers arrive less often.
• Forgetting the cap. EI pays 55 percent of average insurable earnings only up to $729 a week in 2026, so higher earners get a lower replacement rate.

On the exam

Put the wage offer on the horizontal axis and dollars per week of search on the vertical. Draw the marginal benefit curve falling and the marginal cost line $w - b$ rising, and label their crossing $w_A$. For a benefit increase, shift the marginal cost line down by the change in $b$ and show the new, higher $w_A$. State the trade-off in one sentence: a longer expected search for a better expected job.

Check yourself

1.

Unemployment insurance benefits rise. According to the job search model, what happens to an insured worker's asking wage, the expected length of her search and the expected wage of the job she takes?

1All three rise2The asking wage falls because she is less desperate, so the search gets shorter3The asking wage and the length of search rise, but the expected wage of the job taken is unchanged4Nothing changes, because benefits replace income without changing the costs of search

Check answer

2.

An unemployed worker receives a job offer in any given week with probability 0.25. Offers pay between $400 and $1,200 a week, every wage in that range equally likely. Her asking wage is $1,000 a week. How many weeks should she expect to search?

Your answer (weeks)

Check answer

3.

Why can unemployment insurance that is not experience-rated increase temporary layoffs?

1Laid-off workers collect benefits paid for by premiums that do not rise with the firm's own layoffs, so other employers and workers bear part of the cost of each layoff2Firms receive the benefits paid to the workers they lay off3Workers on temporary layoff are not counted as unemployed, so layoffs cost firms nothing4Experience rating forces firms to keep paying laid-off workers their full wage

Check answer

Practise more questions on this topic

Sources

1. Krueger, A. B. and Mueller, A. I. (2016). A contribution to the empirics of reservation wages. American Economic Journal: Economic Policy, 8(1), 142 to 179. NBER working paper 19870. ↩

2. Government of Canada, EI regular benefits: Eligibility, page updated October 15, 2026. ↩ ↩²

3. Statistics Canada, Table 14-10-0354-01, Regional unemployment rates used by the Employment Insurance program, three-month moving average, seasonally adjusted. ↩

4. Government of Canada, EI regular benefits: How much you could receive, page updated October 19, 2026. ↩ ↩²

5. Government of Canada, Temporary Employment Insurance measures to respond to major changes in economic conditions, page updated October 6, 2026. ↩

6. Canada Revenue Agency, EI premium rates and maximums. ↩

7. Chetty, R. (2008). Moral hazard versus liquidity and optimal unemployment insurance. Journal of Political Economy, 116(2), 173 to 234. NBER working paper 13967. ↩

8. Christofides, L. N. and McKenna, C. J. (1996). Unemployment insurance and job duration in Canada. Journal of Labor Economics, 14(2), 286 to 312. ↩

9. Lemieux, T. and MacLeod, W. B. (2000). Supply side hysteresis: The case of the Canadian unemployment insurance system. Journal of Public Economics, 78(1 to 2), 139 to 170. NBER working paper 6732. ↩

10. Feldstein, M. (1976). Temporary layoffs in the theory of unemployment. Journal of Political Economy, 84(5), 937 to 957. ↩

11. Pavosevich, R. (2020). The cost of layoffs in unemployment insurance taxes. Monthly Labor Review, US Bureau of Labor Statistics, April 2020. ↩

12. Topel, R. H. (1983). On layoffs and unemployment insurance. American Economic Review, 73(4), 541 to 559. Card, D. and Levine, P. B. (1994). Unemployment insurance taxes and the cyclical and seasonal properties of unemployment. Journal of Public Economics, 53(1), 1 to 29. NBER working paper 4030. ↩

13. Meyer, B. D. (1990). Unemployment insurance and unemployment spells. Econometrica, 58(4), 757 to 782. NBER working paper 2546. ↩

14. Katz, L. F. and Meyer, B. D. (1990). The impact of the potential duration of unemployment benefits on the duration of unemployment. Journal of Public Economics, 41(1), 45 to 72. NBER working paper 2741. ↩

15. Card, D., Chetty, R. and Weber, A. (2007). The spike at benefit exhaustion: Leaving the unemployment system or starting a new job? American Economic Review, 97(2), 113 to 118. NBER working paper 12893. ↩

16. Woodbury, S. A. and Spiegelman, R. G. (1987). Bonuses to workers and employers to reduce unemployment: Randomized trials in Illinois. American Economic Review, 77(4), 513 to 530. ↩