Introduction

This course covers topics related to unemployment. Unemployment represents a waste of productive resources and therefore is something that should be limited. In addition to its wastefulness, unemployment generates other, large costs to society. People who are unemployed suffer from lower mental and physical health than employed workers. Even employed people in areas with high unemployment report lower well-being. Accordingly, good economic policy should stabilize unemployment at a desirable level and avoid periods of elevated unemployment.

This course presents several models of unemployment. These models explain why unemployment exist, and why unemployment fluctuate over the business cycle. The course also strives to justify all the assumptions introduced in the models.

These models are extremely helpful to design effective policies to tackle unemployment. First, the models are useful to determine what is the socially efficient level of unemployment. This socially efficient level of unemployment is a key ingredient to good labor-market policies. Second, the models explain how various labor market policies should respond to fluctuations in unemployment.

Lecture videos
  1. Why do we care about unemployment?
  2. Absence of unemployment in the competitive model
Main readings
  • Darity and Goldsmith (1996) – This survey reviews the psychosocial consequences of unemployment. It finds that exposure to unemployment severely damages psychological health. One reason is that joblessness creates a sense of helplessness: that one’s life is no longer under their control. Furthermore, numerous benefits of work are lost during unemployment, including a structured daily routine, regular interactions and shared experiences with others, a sense of purpose and a source of personal status and identity.
  • Frey and Stutzer (2002) – This survey provides evidence that personal unemployment and aggregate unemployment significantly reduce happiness. The evidence refers to the pure effect of unemployment, which controls for the income loss from unemployment.
Additional readings
  • Di Tella, MacCulloch, and Oswald (2003) – This paper finds in well-being surveys conducted in Europe and the United States that unemployment make people unhappy. Being unemployed makes you unhappy, and having unemployment in your country makes you unhappy. In fact, after controlling for income and other personal characteristics, becoming unemployed appears as painful as divorcing.
  • Winkelmann and Winkelmann (1998) – This paper uses panel data from Germany to show that unemployment causes unhappiness (and not the other way around). It finds that unemployment significantly reduces life satisfaction, and that the non-pecuniary cost of unemployment is much larger than the pecuniary cost.
  • Borgschulte and Martorell (2018) – This paper uses administrative data from the US military to compute revealed-preference estimates of the cost of unemployment. It finds that unemployment is indeed very costly.
  • Hussam, Kelley, Lane, and Zahra (2022) – This paper reports the results from a field experiment in Bangladesh designed to measure the psychosocial cost of unemployment. The experiment shows that paid employment raises psychosocial well-being substantially more than the same amount of cash alone. In fact, two-thirds of employed workers would be willing to forgo cash payments and continue working for free. The experiment illustrates just how large the psychosocial cost of unemployment is.

Labor market facts and matching function

This section reviews basic facts about the labor market and unemployment in the United States. It then introduces the matching function—the main tool that we will use to model the labor market and unemployment. The matching function summarizes the complex process through which workers searching for jobs and firms searching for employees meet each other.

Lecture videos
  1. Organization of the labor market
  2. Unemployment rate
  3. Vacancy rate
  4. Beveridge curve
  5. Job-finding rate
  6. Vacancy-filling rate
  7. Job-separation rate
  8. Matching function
  9. Labor market tightness
Lecture notes
Main readings
Additional readings
  • Barnichon (2010) – This paper measures the US job-vacancy rate since 1951 by combining the Conference Board’s help-wanted advertising index (available since 1951) with the BLS measure of job vacancies (available since 2001). Critically, the Conference Board’s index is adjusted to account for the shift from print job advertising to online job advertising in the 1990s.
  • Petrosky-Nadeau and Zhang (2021) – This paper combines various historical datasets to extend the BLS measure of US unemployment rate (which starts in 1948) and the Barnichon measure of US vacancy rate (which starts in 1951). The resulting measures of the US unemployment and vacancy rates go back to World War 1 (1919).
  • Shimer (2012) – This paper shows that in the United States, unemployment fluctuations are caused by fluctuations in the job-finding rate and not fluctuations in the job-separation rate.
  • Michaillat and Saez (2024) – This paper combines data on unemployment and job vacancies to build a new recession rule. The rule detects all US recessions since 1929, without errors, and it detects them faster than well-known existing rules. The rule shows that the US economy may have entered a recession as early as March 2024.
Practice material

Matching model of the labor market

This section introduces the matching model of the labor market. This is the model that we will use to study unemployment and labor market policies in this course. Early work on this model was conducted by Peter Diamond, Dale Mortensen, and Christopher Pissarides in the 1970s and 1980s, so the canonical version of the model is often called the Diamond-Mortensen-Pissarides model, or DMP model. For their work, Diamond, Mortensen, and Pissarides received the Economics Nobel Prize in 2010. In the matching model, unlike in the neoclassical model, all trades are mediated by a matching function. Nevertheless, we can construct labor supply and labor demand curves, and use them to solve the model.

Lecture videos
  1. Properties of a good model according to Kuhn
  2. Notations
  3. Labor market flows
  4. Assumption of balanced flows
  5. Computing labor supply
  6. Properties of labor supply
  7. Organization of the firm
  8. Computing the recruiter-producer ratio
  9. Properties of the recruiter-producer ratio
  10. Profits of the firm
  11. Problem of the firm
  12. Computing labor demand
  13. Properties of labor demand
  14. Review of labor supply and demand
  15. Irrelevance of unemployment dynamics
  16. Finding the wage in a competitive model
  17. Finding labor market tightness in a matching model
  18. Graphical representation of the solution
  19. Complete description of the solution
Lecture notes
Main readings
  • Kuhn (1957, chapters 1 and 5) – This book studies the Copernican Revolution in astronomy and in the process isolates the three properties of a good model: economy, accuracy, and fruitfulness.
  • Pissarides (2001, chapter 1) – This chapter introduces the canonical version of the matching model of the labor market (the DMP model).
Additional readings
  • Rogerson, Shimer, and Wright (2005) – This survey reviews a range of search and matching models.
  • Pissarides (2011) – In this Nobel lecture, Christopher Pissarides presents the history of the DMP model. He also discusses several applications of the model.
  • Mortensen (2011) – In this Nobel lecture, Dale Mortensen presents a brief history of the DMP model and a basic version of the model. He then discusses how the model can help understand the Great Recession of 2008–2009.
  • Michaillat (2024) – This paper develops a basic matching model of the labor market and uses it to study the impact of immigration on local workers. The model and its presentation are similar to those in the course. The paper highlights how different assumptions alter the effects of immigration in the model. The paper shows that despite its simplicity, the model developed in the course is helpful to understand contemporary economic and political issues.
Practice material

Wage functions

This section discusses the labor market institutions that determine wages, such as unions, minimum wages, and corporate policies. It then discusses various wage functions that can be used in the matching model, such as fixed wages, rigid wages, and bargained wages.

Lecture videos
  1. Unemployment over the business cycle
  2. Unemployment in the matching model
  3. Sources of business cycles
  4. Wages over the business cycle
  5. Wages in the matching model
  6. Unions
  7. Minimum wage
  8. Efficiency wages
  9. Fixed wages
  10. Rigid wages
  11. Bargained wages
  12. Surplus of the firm
  13. Surplus of the worker
  14. Surplus sharing
  15. Properties of bargained wages
Lecture notes
Main readings
  • Bewley (2005) – This survey explains how wages are set, why wages are rigid, and in particular why wages do not fall in recessions. It finds that firms avoid pay cuts because cuts damage morale and therefore reduce productivity, increase turnover, and complicate recruiting.
  • Haefke, Sonntag, and Van Rens (2013) – This paper estimates that the real wages of new hires are somewhat rigid. The elasticity of real wages with respect to productivity is between 0.7 and 0.8, so less than 1.
Additional readings
  • Akerlof (1984) – This survey reviews various theories of efficiency wages. These theories try to explain how firms set wages in practice. They consider the effect of wages on productivity, attachment to the firm, retention, hiring, and so on.
  • Jacoby (1984) – This chapter analyzes the development of internal labor markets in American manufacturing firms. It explains how internal labor markets replaced spot labor markets, and how such replacement provided workers with an employment system that was more bureaucratic, more rule-bound, and more secure. In particular, wages paid to workers were much more rigid in internal labor markets.
  • Raff and Summers (1987) – This paper examines Henry Ford’s introduction of the five-dollar day in 1914 in his automobile factory. It argues that Ford’s decision to increase wages dramatically is the consequence of labor problems of the kind featured in efficiency-wage theories. Moreover, it finds that the five-dollar day resulted in substantial queues for Ford jobs, significant increases in productivity, and maybe surprisingly, significant increases in profits.
  • Mas (2006) – This paper finds evidence in favor of efficiency-wage theories: when police officers receive a wage below what they consider a fair wage, they are disappointed, and their performance on the job drops. This finding confirms that fairness might be an important source of wage rigidity.
Practice material

Unemployment fluctuations

This section discusses unemployment fluctuations in the matching model. We first show that the model with rigid wages generates realistic fluctuations in unemployment. We then show that the model with bargained wages is unable to generate such fluctuations. The reason is that bargained wages are too flexible.

Lecture videos
  1. Matching model with rigid wages
  2. Business-cycle shocks
  3. Labor supply shocks with rigid wages
  4. Labor demand shocks with rigid wages
  5. Elasticities
  6. Fluctuations in labor market tightness with rigid wages
  7. Elasticity of labor supply
  8. Elasticity of the recruiter-producer ratio
  9. Elasticity of labor market tightness
  10. Wage rigidity required to generate realistic fluctuations
  11. Matching model with bargained wages
  12. Labor supply shocks with bargained wages
  13. Labor demand shocks with bargained wages
  14. Fluctuations in labor market tightness with bargained wages
  15. Value from unemployment
  16. Bargained wages cannot generate realistic fluctuations
Lecture notes
Main readings
  • Shimer (2005) – This paper shows that the canonical matching model of the labor market (the DMP model) cannot generate realistic fluctuations in unemployment and vacancies. This anomaly of the DMP model is known as the Shimer puzzle. The reason behind the Shimer puzzle is that the wage bargaining protocol assumed in the DMP model—Nash bargaining—produces wages that are too flexible.
  • Hall (2005) – This paper shows that a matching model with fixed real wages generates large fluctuations in unemployment and vacancies—larger in fact that the fluctuations observed in the United States. Hence fixed real wages solve the Shimer puzzle.
Additional readings
  • Hagedorn and Manovskii (2008) – This paper shows that if workers are almost indifferent between working and not working, then the DMP model can generate realistic fluctuations in unemployment and vacancies. A downside of this assumption, however, is that it is completely inconsistent with the large psychosocial cost of unemployment documented at the beginning of the course. Another downside is that it implies that the socially efficient rate of unemployment is above 20%.
  • Hall and Milgrom (2008) – This paper proposes a form of wage bargaining that produces somewhat-rigid wages. With such wage rigidity, the matching model generates realistic fluctuations in unemployment and vacancies.
  • Blanchard and Gali (2010) – This paper blends the matching model of the labor market with a New Keynesian model of the product market. It then shows that under wage bargaining, unemployment and vacancies do not fluctuate at all in the model. This is a strong form of the Shimer puzzle. By contrast, when the real wage is a subproportional function of labor productivity, the model generates realistic fluctuations in unemployment and vacancies.
  • Eliaz and Spiegler (2013) – This paper insert reference-dependent preferences into a matching model. Because of these preferences, wage cuts relative to a reference point make workers feel that they have been treated unfairly, which dampens their intrinsic motivation and reduces their output. As a result, firms avoid cutting wages. Because wages are rigid downward, unemployment and vacancies respond more to shocks than under Nash bargaining.
Practice material

Frictional and rationing unemployment

This section turns to the sources of unemployment over the business cycle. In the DMP model, unemployment becomes vanishingly small when unemployed workers search sufficiently hard for jobs. Unemployment also becomes vanishingly small when recruiting costs are sufficiently low. In other words, the DMP model does not feature job rationing. This lack of job rationing is difficult to reconcile with the long queues of unemployed workers at job bureaus and factory gates observed during the Great Depression.

The section then develops a matching model with job rationing. In the model, in recession, jobs are lacking, so some unemployment remains even if workers are desperate to find a job. The model features both frictional unemployment—caused by difficulties in matching workers and firms—and rationing unemployment—caused by a lack of job. The model describes well good and bad times. In bad times, labor demand is low so rationing unemployment is high. Hence total unemployment is high. But, maybe surprisingly, frictional unemployment is low. In that case, workers queue for jobs and it is easy for firms to fill vacancies. Conversely, in good times, labor demand is high so rationing unemployment is low and total unemployment is low. Frictional unemployment is higher than in bad times. In that case, it is easy for workers to find jobs but firms struggle to fill job vacancies.

Technically, the DMP model does not feature job rationing because its labor demand is perfectly elastic with respect to wages and labor market tightness. Once we introduce a labor demand that is downward sloping with respect to wages and tightness, job rationing appears and not all unemployment is frictional. The easiest way to generate a downward-sloping labor demand is by assuming that the production function has diminishing returns to labor.

Lecture videos
  1. Introduction to job rationing
  2. Evolution of matching models: standard model (DMP model)
  3. Evolution of matching models: rigid-wage model
  4. Evolution of matching models: job-rationing model
  5. All unemployment is frictional in standard and rigid-wage models
  6. Standard model with zero recruiting cost
  7. Rigid-wage model with zero recruiting cost
  8. The case of infinite job-search effort
  9. Standard model with infinite job-search effort
  10. Rigid-wage model with infinite job-search effort
  11. Generating job rationing
  12. Unemployment with zero recruiting cost
  13. Defining rationing unemployment
  14. When is rationing unemployment positive?
  15. Unemployment with infinite job-search effort
  16. Measuring frictional and rationing unemployment
  17. Frictional and rationing unemployment over the business cycle
  18. Frictional and rationing unemployment in a calibrated model
Lecture notes
Main readings
  • Michaillat (2012) – This paper first establishes that the DMP model does not have job rationing and then develops a matching model with job rationing. The job-rationing model departs from the DMP model by assuming that the production has diminishing returns to labor instead of constant returns to labor, and that the real wage is given by a rigid wage norm instead of Nash bargaining. In that model unemployment can be decomposed into rationing and frictional components. In recessions, the rationing component is large while the frictional component is small.
  • Crepon, Duflo, Gurgand, Rathelot, and Zamora (2013) – This paper reports the results from a randomized experiment designed to evaluate the direct and indirect effects of job-placement assistance on the labor market outcomes of young educated jobseekers in France. It finds that jobseekers who are given job-placement assistance find jobs more rapidly, while jobseekers in the same labor market who are not assisted take longer to find a job. This finding suggests that jobs are somewhat rationed, and that assisted jobseekers move ahead of non-assisted jobseekers in job queues. The spillovers of job-placement assistance are particularly strong in bad times, when jobs are scarcer.
Additional readings
  • Akerlof, Rose, and Yellen (1988) – This paper develops an early matching model of the labor market with job rationing. The model reproduces a number of stylized facts of the US labor market regarding quits and vacancy chains. While the model in not in the DMP tradition, it does feature a Beveridge curve.
  • Michaillat and Saez (2015) – This paper adds aggregate demand to the model in Michaillat (2012). This is done by adding a product market to the labor market with a similar matching structure. Aggregate demand shocks generate fluctuations in unemployment and vacancies along the Beveridge curve. In that model, unemployment can be decomposed into three components: Keynesian unemployment (due to insufficient aggregate demand), classical unemployment (due to high real wages), and frictional unemployment (due to matching frictions).
  • Michaillat and Saez (2022) – This paper builds a dynamic version of the model in Michaillat and Saez (2015), which is static. In this model the central bank can influence aggregate demand and unemployment through interest rates.
  • Michaillat and Saez (2024) – This paper uses the dynamic model in Michaillat and Saez (2022) and generates a Phillips curve by introducing price competition through directed search. To ensure that unemployment fluctuates, the model assumes price rigidity through quadratic price-adjustment costs. The Phillips curve produced by the model guarantees divine coincidence: inflation is on target when unemployment is efficient.
Practice material

Efficient unemployment and unemployment gap

Unlike in neoclassical models, in matching models there is no guarantee of efficiency. The prevailing unemployment rate is generally inefficient: either too high or too low.

This section defines and computes the socially efficient rate of unemployment in the matching model. It then extends the analysis to all models with a Beveridge curve—not only the matching model. It develops measures of the efficient labor market tightness and efficient unemployment rate. It also develops a measure of the unemployment gap to assess how far the unemployment rate is from its socially efficient level. The measure depends on three sufficient statistics: recruiting cost, social value of nonwork, and elasticity of the Beveridge curve.

The section finally shows that in the United States, the unemployment gap is generally positive and is sharply countercyclical. This means that the US labor market is generally inefficiently slack and especially inefficiently slack in slumps. For instance, the unemployment gap reached 6 percentage points during the Volcker Recession and the Great Recession.

Lecture videos
  1. Why do we need to know the efficient unemployment rate?
  2. Natural rate of unemployment and NAIRU
  3. Defining the efficient unemployment rate
  4. Problem of the social planner
  5. Solution to the planner’s problem
  6. Computing efficient labor market tightness
  7. Graphical representation of matching efficiency
  8. Why is the labor market generally inefficient?
  9. Description of Beveridgean models
  10. Social welfare in Beveridgean models
  11. Efficiency in Beveridgean models
  12. Graphical representation of Beveridgean efficiency
  13. Comparative statics for the efficient unemployment rate
  14. Graphical representation of the unemployment gap
  15. Sufficient-statistic formula for efficient labor market tightness
  16. Implementing the sufficient-statistic formula
  17. Unemployment gap in the United States
Lecture notes
Main readings
  • Michaillat and Saez (2021) – This paper derives sufficient-statistic formulas for efficient labor market tightness and efficient unemployment rate. It applies the formulas to the United States and finds that the US labor market is generally inefficiently slack, and especially inefficiently slack in slumps.
  • Michaillat and Saez (2024) – This paper shows that under simple but realistic assumptions, the sufficient-statistic formula for the efficient unemployment rate from Michaillat and Saez (2021) reduces to $u^\ast = \sqrt{uv}$. This unemployment rate marks not only social efficiency but also full employment. In the United States, 1930–2024, the full-employment rate of unemployment (FERU) $u^\ast$ averages 4.1%.
Additional readings
  • Hosios (1990) – This paper shows that in a DMP model, unemployment is efficient when workers’ bargaining power equals the elasticity of the matching function with respect to unemployment.
  • Chetty (2009) – This survey describes the sufficient-statistic method for welfare and policy analysis.
  • Robinson (1946) – This essay discusses various definitions of full employment and explains why full employment does not mean zero unemployment.
  • Gokten, Heimberger, and Lichtenberger (2024) – This paper uses the formula $u^\ast = \sqrt{uv}$ to compute the FERUs in selected European countries (Germany, Sweden, Austria, Finland, UK) between 1970 and 2022. It compares the FERUs and unemployment gaps in Europe to those in the United States.
Practice material

Labor-demand policies

Over the business cycle, fluctuations in labor demand generate fluctuations in unemployment that are inefficient. Labor market policies that can influence labor demand should therefore counterbalance these fluctuations and bring the unemployment rate closer to its efficient level. This section discusses how minimum wage, payroll tax, and public employment—all policies that directly influence labor demand—should respond to unemployment fluctuations over the business cycle.

Lecture videos
  1. Active and passive labor market policies
  2. Modeling a minimum wage
  3. Labor supply and demand with a minimum wage
  4. Designing an optimal policy
  5. Optimal minimum wage
  6. Evidence on the minimum wage
  7. Other possible effects of the minimum wage
  8. Modeling a payroll tax
  9. Labor supply and demand with a payroll tax
  10. Optimal payroll tax
  11. Public employment in the United States
  12. Matching process with public employment
  13. Labor supply and demand with public employment
  14. Graphical representation of public employment
  15. Computing the public-employment multiplier
  16. Slopes of the labor supply and demand curves
  17. Expression for the multiplier
  18. Multiplier is positive but less than one
  19. Multiplier is countercyclical
  20. Multiplier in slumps
  21. Multiplier in booms
  22. Optimal public employment
Lecture notes
Main readings
  • Michaillat (2014) – This paper establishes that in a matching model with job rationing, the public-employment multiplier is positive but less than one, and the multiplier is larger when the unemployment rate is higher. These predictions are consistent with a growing body of evidence from the United States and abroad.
  • Auerbach and Gorodnichenko (2012) – This paper finds that in the United States, government multipliers are larger when the unemployment rate is higher than when the unemployment rate is lower.
Additional readings
  • Crepon and van den Berg (2016) – This survey assesses the effectiveness of various active labor market policies: policies designed to improve labor market outcomes of jobseekers. Policies covered include job-search assistance (to stimulate labor supply) and job subsidies (to stimulate labor demand). The survey highlights spillovers from such policies. Job-search assistance is not found to be very effective, which is unsurprising in a world with job rationing. Job subsidies are found to be moderately effective.
  • Neumark and Shirley (2022) – This survey reviews evidence on the effects of the minimum wage on employment in the United States. Overall, the evidence points toward negative effects of minimum wages on employment of less-skilled workers.
  • Neumann, Fishback, and Kantor (2010) – This paper describes public employment in the United States during the First New Deal and Second New Deal. It then documents the effect of public employment on the private labor market, unearthing evidence of crowding out of private employment by public employment.
  • Auerbach and Gorodnichenko (2013) – This paper finds that in a large number of Organization for Economic Cooperation and Development (OECD) countries, government multipliers are larger in recessions than in expansions.
Practice material

Unemployment insurance

Fluctuations in unemployment raise another policy question: how should the generosity of unemployment insurance respond to unemployment fluctuations? This question was hotly debated during the Great Recession. Some argued that unemployment insurance should be reduced because it discouraged job search and would raise unemployment further. Other countered that unemployment insurance could be increased without raising unemployment much—as there were no jobs available for jobseekers.

This section weights the two sides of the argument. It discusses the different channels through which unemployment insurance affects the labor market. It then derives a formula for optimal unemployment insurance. The formula contrasts how generous unemployment insurance should be in good times and in bad times. The formula shows that the generosity of unemployment insurance should be adjusted over the business cycle; the adjustment depends on how unemployment insurance affects labor market tightness. The effect of unemployment insurance on tightness depends on the model: increasing unemployment insurance may raise tightness by alleviating the rat race for jobs or lower tightness by increasing wages through bargaining.

In the United States, unemployment insurance is more generous in bad times than in good times. The formula, combined with evidence from the US labor market, suggests that such countercyclical generosity is desirable. The reasons are that labor market tightness is inefficiently low in slumps and inefficiently high in booms (as we saw previously), and that increasing unemployment insurance raises labor market tightness.

Lecture videos
  1. Unemployment insurance in the United States
  2. Matching process with unemployment insurance
  3. Labor demand with unemployment insurance
  4. Social welfare with unemployment insurance
  5. Effort supply with unemployment insurance
  6. Labor supply with unemployment insurance
  7. Graphical representation of unemployment insurance
  8. Unemployment insurance in the job-rationing model
  9. Moral-hazard channel of unemployment insurance
  10. Microelasticity and macroelasticity of unemployment
  11. Rat-race channel of unemployment insurance
  12. Unemployment insurance in the rigid-wage model
  13. Unemployment insurance in the standard model
  14. Job-creation channel of unemployment insurance
  15. Optimal unemployment insurance
  16. Formula for optimal unemployment insurance
  17. Baily-Chetty term
  18. Effect of labor market tightness on welfare
  19. Effect of unemployment insurance on labor market tightness
  20. Cyclicality of optimal unemployment insurance
Lecture notes
Main readings
  • Landais, Michaillat, and Saez (2018a) – This paper studies optimal unemployment insurance in a matching model and obtains a sufficient-statistic formula for optimal unemployment insurance.
  • Landais, Michaillat, and Saez (2018b) – This paper applies the sufficient-statistic formula from Landais, Michaillat, and Saez (2018a) to the United States and finds that optimal unemployment insurance is countercyclical—more generous in bad times than in good times.
Additional readings
  • Chetty (2006) – This paper studies optimal unemployment insurance in a model with fixed tightness and derives the Baily-Chetty formula.
  • Lalive, Landais, and Zweimueller (2015) – This paper uses a natural experiment in Austria to measure the response of labor market tightness to unemployment insurance. The experiment consists of an increase in benefit duration from 52 to 209 weeks for eligible unemployed workers in a subset of regions. Ineligible unemployed workers in treated labor markets experienced significantly lower unemployment duration, which implies that labor market tightness rose in treated labor markets.
  • Marinescu (2017) – This paper examines how the tripling of the duration of unemployment benefits during the Great Recession impacted the US labor market. It finds that the number of job applications fell while the number of job openings remained that same. Therefore, labor market tightness (the ratio of job openings to job applications) increased when unemployment insurance became more generous.
  • Dieterle, Bartalotti, and Brummet (2020) – This paper shows substantial spillovers across US states when one state changes the duration of its unemployment insurance benefits and the other does not. In particular, workers from the low-benefit state flock to the high-benefit state, suggesting increased job-finding rates and thus increased labor market tightness in the high-benefit state.
Practice material