Economists have long debated why labour markets do not work as smoothly as textbooks suggest. People remain in the same jobs for months even when better-paid opportunities are available, while companies complain that they cannot find employees willing to take the risk of moving. A new study based on a direct survey of 1,000 employed Americans and a structural model of the labour market measures the cost of this caution in concrete terms for the first time. It also adds an unexpected twist to the debate over why unemployment insurance is needed in the first place.
The logic begins with a simple question that no one had previously asked directly: how much extra would a worker need to be paid to accept an additional one percentage point risk of being unemployed a year later? A representative survey of 1,008 employed Americans conducted in spring 2025 produced a remarkably consistent answer: a 1.63% pay rise for every additional percentage point of dismissal risk. The figure was almost identical to the prediction of the authors’ theoretical labour-market model, even though the model had not been calibrated to reproduce this particular result. Its counterpart was 1.53%, an unusually close match between theory and observed behaviour by the standards of macroeconomics.
The variation around that average was not random noise but a systematic signal. Workers in already secure jobs demand almost twice as much compensation for additional risk as those who are already in precarious employment: a 2.11% pay rise against 0.96% for the same percentage point increase in risk. The pattern is similar for income. The higher a person’s current pay, the more compensation they require to put it at risk. Most revealingly, workers who admit they would have to cut spending sharply after losing their job demand almost twice as much compensation for risk as those with a financial cushion, 1.72% against 0.98%. This is direct evidence that the less able people are to smooth a fall in income themselves, the more tightly they hold on to their current job, even when it is less productive.
The scale of this distortion is clear from the way workers are actually distributed across jobs. Only 26% of all job offers are for secure positions with a risk of dismissal no higher than average. Yet 68% of employed workers are in such jobs. This is neither coincidence nor a random selection effect. It reflects a deliberate flight from risk. Workers actively reject riskier but more productive opportunities and gravitate towards secure positions even when those jobs pay less.
The model puts a number on what this collective caution costs the economy. If imperfections in insurance markets were eliminated entirely, hypothetically making every worker indifferent to risk, job-to-job transitions would rise by 12% and aggregate economic productivity by 0.19%. The absolute number may look modest, but it captures the pure loss from a single mechanism: excessive caution when changing jobs, something that had not previously been included in standard models of the labour market.
The most surprising part of the study then overturns the conventional view of unemployment benefits. These payments are usually discussed as insurance for people who have already lost their jobs, allowing them to avoid grabbing the first poor offer that appears. The authors show, however, that benefits serve just as much, if not more, as insurance for those who are still employed. Knowing there is a cushion if the move goes wrong makes a worker more willing to accept a competitor’s riskier but better-paid and more productive offer. When the model sharply cuts benefits from a typical US replacement rate of 40% of pay to 8%, effectively eliminating insurance while retaining only basic social assistance, productivity falls by 1.29%. Three-quarters of the effect comes from the behaviour of employed workers who become too frightened to risk changing jobs, not from the standard channel in which unemployed people become less selective when looking for their first job after dismissal.
There is also a counterintuitive side effect for employers. Cutting benefits should, in theory, make recruitment easier because unemployed workers become more willing to compromise. In practice, the opposite happens. The arrival rate of new job offers falls by almost 2% because employed workers respond to less generous benefits in precisely the opposite way: they accept competitors’ approaches less often, making vacancies aimed at poaching existing employees harder to fill. Part of the saving on benefits is consumed by a less dynamic labour market, particularly in the part that normally moves people into more productive jobs.
What happens during an economic downturn deserves separate attention. A simulated recession on the scale of the Great Recession raises the price of risk, defined as the difference between the value of the current job and the value of unemployment, by 52% at the moment of impact. This shows up less in the overall number of job-to-job moves than in their character. The share of transitions in which workers accept lower pay in return for greater security almost doubles, from 7% to 14%. The authors describe this as moving up the safety ladder. A temporary 50% increase in unemployment benefits almost completely neutralises the effect, although it produces a higher peak in unemployment because people spend longer searching for a suitable position rather than accepting the first job available.
Applied to Kazakhstan, the logic produces an even sharper picture than in the US study. Kazakhstan formally has a similar instrument: payments from the State Social Insurance Fund with an income replacement rate of up to 45%, slightly higher than the paper’s baseline US scenario of 40%. But differences in the details move Kazakhstan much closer to the model’s extreme paycheque-to-paycheque case, which produces the steepest risk-compensation curve and the greatest damage to mobility.
The first difference is the duration of payments, which ranges from one to six months depending on how long a person has participated in the system. Insurance therefore expires just as a financial cushion begins to be genuinely necessary. The second and far more important difference is coverage. Only participants in the mandatory social insurance system who have made formal contributions are eligible. Yet the self-employed account for almost one-quarter of all workers in Kazakhstan, and a significant share operate informally without any social guarantees. Add employees who also work without formal registration, and a substantial part of Kazakhstan’s labour force receives not 45% of previous pay after losing a job, but nothing. This is precisely the paycheque-to-paycheque scenario in which the model predicts not moderate but dramatically greater caution over job choices.
Kazakhstan’s labour market also has a structural feature that the paper does not examine directly but towards which its logic clearly points. A significant share of employment is concentrated in the state and quasi-state sectors, including large national companies and publicly funded organisations that are perceived as secure and stable. The paper’s mechanism predicts that when insurance against risk is weak and short-lived, workers will be drawn towards precisely these employers. The riskier but potentially more productive private sector is left without an inflow of skilled labour. This is exactly the distortion the authors identify as a source of lost productivity, amplified in Kazakhstan by the narrow coverage and short duration of unemployment insurance.
This leads to an unconventional policy conclusion that follows directly from the paper. Expanding Kazakhstan’s unemployment insurance system, particularly by extending it to self-employed and informally employed workers, should be discussed not only as a social-protection measure but also as a potential instrument for accelerating the movement of labour from the less productive quasi-state sector into private business. In other words, it could become a lever for the economic diversification that has featured in government policy for years. The paper, of course, models a relatively competitive labour market, while Kazakhstan adds frictions of its own: the monopsony power of large quasi-state employers, weak administration of the informal sector, and regional and skills mismatches absent from the original model. Its logic should therefore be treated as a direction for analysis, not as a literal transfer of the specific percentages.
Elvira Baturova, EconomyKZ Research Group, exclusively for EconomyKZ.org
