Labor Market ABM Model
How do search frictions, firm hiring rules, and heterogeneous worker skills generate unemployment duration distributions, wage dispersion, and Beveridge curve dynamics that aggregate models leave out?
Understand the mathematics
Follow the same equations and assumptions one section at a time or in a continuous reading.
Hover or focus a symbol defined in this model to find its exact matches. Select it to keep the highlight; Escape clears the selection.
Odd+D Protocol: Model Specification
PURPOSE: The model explores how decentralized search-and-matching between heterogeneous workers and firms generates aggregate labor market outcomes (unemployment, vacancies, wage distribution, duration distribution) that cannot be derived from a representative-agent matching function. The target audience is labor economists and policymakers evaluating active labor market programs, unemployment insurance design, and minimum wage effects.
ENTITIES, STATE VARIABLES, AND SCALES: Workers ( to ) carry: skill type (discrete: low/mid/high or continuous), current wage (0 if unemployed), reservation wage, unemployment duration, search intensity, and cumulative income. Firms ( to ) carry: productivity, workforce (list of worker ids), vacancy count, wage offer schedule (may be skill-dependent), revenue, and retained earnings. The policy block carries: UI replacement rate, UI maximum duration, minimum wage, and hiring subsidy parameters. Temporal scale is quarterly; spatial dimension is absent in the baseline.
PROCESS OVERVIEW AND SCHEDULING: Each period: (1) aggregate demand shock arrives, (2) firms set production targets and compute desired workforce, (3) firms fire/lay off if over-staffed, (4) firms post vacancies, (5) unemployed workers search (sample vacancies each), (6) workers apply to acceptable vacancies, (7) firms rank applicants and hire, (8) wages are paid, consumption happens, revenue accrues, (9) UI payments are made, (10) state variables update. The fixed ordering ensures that firing precedes vacancy posting (firms know their labor gap before posting) and that search precedes hiring (workers and firms interact in a defined sequence).
The aggregate matching function is measured from simulation output, not imposed. The Cobb-Douglas form is an empirical approximation to the emergent relationship. The parameters and may shift endogenously with skill composition.
This single form is used in every tab. The reservation wage decays geometrically with duration; while UI eligibility lasts, the benefit level acts as a FLOOR under that decay, so a worker never accepts less than the benefit is worth. At exhaustion the floor disappears and the reservation wage resumes decaying toward the minimum wage. Note the UI term is a floor, not a markup: a freshly laid-off worker's reservation wage starts at ( is below it), never above it. Whether the floor actually binds is a calibration question, not a property of the form: the UI term wins the max only when . At the graph tab's seed values (, ) that requires above 0.83, outside the range that tab's slider offers, so the floor never binds there and the replacement rate does not move durations in the sketch.
Emergent vs. imposed matching functions
When the emergent matching function is estimated from ABM output, the elasticity parameters () are typically close to the DMP calibration () in normal times but shift significantly during deep recessions. Fagiolo, Dosi, and Gabriele (2004) document that the matching function becomes less efficient (lower ) when skill mismatch increases, a result that cannot be captured by a fixed matching function.
The baseline does not produce negative duration dependence
It is tempting to claim the ABM generates negative duration dependence emergently. Under the rules as specified it does not, and the sign is worth being explicit about. A longer spell LOWERS the reservation wage, which RAISES the acceptance probability. Firms rank applicants by skill fit and break ties toward the cheaper wage, which FAVORS long-duration workers. Vacancy sampling is i.i.d. across periods with no memory, so being passed over last period does not worsen this period's draw. The baseline therefore produces roughly flat or mildly POSITIVE duration dependence. Negative duration dependence requires a variant that is switched off by default: the SKILL-DECAY variant (skills depreciate with the spell, so long-duration workers stop meeting skill requirements) or the DURATION-SCREENING variant (firms rank applicants on duration as a signal, i.e. statistical discrimination). Both are listed in the variants section; either one flips the sign, and neither is in the baseline.
Transition checks
Sampled dynamics
The state is updated at observed dates. A line between dates is a display convention unless the model specifies behavior inside the interval.
Read adjacent points as adjacent sample calculations. Smoothness is a property of the estimator, not of the chart line.
Do not infer within-period behavior from a sampled estimator.
Variable timing
Latent state
The state updates at sample dates. The visible line is only as smooth as the estimated state sequence.
Observation equation
Observed series can move differently from latent states because measurement noise and loadings matter.