The First Result: Hiring Rose Without a Measured Wage Pass-Through
This chapter reads the PBM reform’s first labor-market results: when employment changed, how large the change was, and whether the lower cost appeared in the contractual wages of new hires. The aim is not to repeat that Brazil’s reform “created jobs,” but to understand what the event-study pattern supports and what it leaves open. That distinction matters because the relationship between employment, wages, and local labor-market concentration is the evidence the paper uses to move beyond a simple tax-incidence story.
Employment responds on a schedule
The employment estimate is dynamic rather than a single jump attached to a policy date. Before sectors entered the PBM, the event-study coefficients show no differential employment trend relative to the comparison path. That means sectors that would later receive the reform were not visibly pulling away from the never-treated sectors in the measured outcome before implementation. The absence of a pre-trend does not prove that treatment timing was random, but it makes one simple alternative less plausible: that the post-reform gap was merely the continuation of a divergence already underway.
The pattern changes after implementation. Employment rises in the first quarter after a sector enters the new regime, strengthens over the following quarters, and then settles at roughly 5 percent above the comparison path. The gradual shape is economically informative. A payroll-tax cut does not create a worker, a vacancy, or a new firm at the instant legislation changes. Employers have to adjust hiring, production, and perhaps entry decisions. The fact that the measured response begins after implementation and grows before stabilizing fits a process in which firms use the lower labor cost over time.
The estimate is a comparison of sector-level employment paths. It does not say that every firm in every treated sector increased its workforce by 5 percent. A sector average can combine firms that expand, firms that do not change, firms that enter, and firms that contract. It also measures formal employment in the administrative data. The result is therefore best stated narrowly: relative to the comparison path, formal employment in the treated sectors increased by about 5 percent after the PBM implementation in the studied period.
The paper translates this magnitude into a semi-elasticity. The average effective reduction in the payroll-tax burden was about half of the original 20 percent payroll tax, or approximately ten percentage points of the effective burden. A 5 percent employment increase associated with that change corresponds to about a 0.5 percent employment increase for each percentage-point reduction in the effective burden. A semi-elasticity is useful here because the tax change is measured in percentage points while employment is measured proportionally. It does not mean that a one-percentage-point reduction always produces exactly 0.5 percent more employment in another country or under another tax design.
Treatment intensity provides a second way to test whether the labor-cost channel is plausible. The researchers divide sectors according to the expected reduction in their tax burden. Sectors above the median reduction eventually show about 7.3 percent higher employment. Sectors below the median show about 2.2 percent, with a slower and less precise response. Both groups appear to respond, but the larger expected cost reduction is associated with the sharper employment increase.
This is a dose-response pattern, not a perfect experiment inside the experiment. The high- and low-intensity groups differ in their economic composition, and the expected tax changes come from sector calculations rather than an identical firm-level charge observed for every employer. Still, the pattern is relevant. If the increase had nothing to do with the labor-cost change, there would be less reason to expect sectors receiving larger reductions to expand more. The result strengthens the cost channel while leaving room for differences in sector trends or other mechanisms.
The intensity comparison also clarifies why the policy cannot be summarized by its headline statutory rates. A firm does not make a hiring decision from the percentage written in the law; it responds to the change in the cost of employing an additional worker relative to the revenue that worker helps generate and to the wages available elsewhere. The paper’s sector calculations are an approximation of that change. They are useful for testing whether larger effective relief is associated with a larger response, but they are not a complete description of each firm’s expected return from hiring.
A different pattern would have weakened the labor-cost interpretation. If employment had begun rising well before implementation, if high-intensity sectors had responded no more than low-intensity sectors, or if the effect appeared only in a group with a separate pre-existing shock, the event study would offer a less persuasive account. None of those observations would by itself prove the tax cut had no effect, but each would make the reported chain harder to defend. The observed timing and intensity pattern work in the opposite direction while leaving the paper’s stated identification limits intact.
The magnitude also needs a denominator. A five percent change in a large sector represents many more jobs than the same percentage change in a small sector. The paper later estimates the total number of formal jobs associated with the policy, but the event study itself is not a fiscal accounting exercise. Its primary job is to identify the relative change in sector employment around implementation. Treating the percentage as an unconditional number of jobs, or treating the semi-elasticity as a universal production rule, would go beyond what the design measures.
The dynamic pattern and the intensity gradient do not settle why the adjustment appears mainly in employment. They establish that the reform is associated with a meaningful change in the quantity of formal labor used by the treated sectors. The next margin is whether firms also bid more for workers when the cost of employing them falls.
What the wage result does—and does not—say
The paper finds small and statistically insignificant effects on wages. The relevant RAIS measure is the contractual hourly wage of new hires, and the estimated effect remains close to zero across the main specifications and across the higher- and lower-intensity sectors. In plain terms, employers hired more people without a detectable increase in the contractual hourly pay recorded for newly hired workers.
That is not a contradiction in the data. Employment and wages are different margins of adjustment. A firm can respond to a lower cost by hiring more at roughly the prevailing wage, especially if it can expand output or attract workers from outside the formal labor market. It can also retain part of the saving. Whether the wage rises depends on how strongly employers compete for workers, how easily workers move among employers, and how much labor the firm wants at the existing wage.
The wage result must be stated with its measurement boundary. The paper observes contractual hourly wages for new hires, not every worker’s total compensation. The estimate does not directly test incumbent wages, bonuses, benefits, hours, job stability, or the value workers place on formal registration. Nor does a statistically insignificant coefficient prove that the true effect is exactly zero. It means the study does not precisely distinguish the estimated change from zero for this wage measure and comparison. A small wage effect remains possible, as does a wage response in an unmeasured margin.
This distinction also separates two questions that are often collapsed. The first is incidence: who captures the tax saving, workers or firms? The second is adjustment: through which quantity or price does the market respond? A null new-hire wage estimate suggests that the measured price did not absorb much of the change, but the employment increase shows that the quantity margin did respond. The paper’s profit estimate adds another possible destination for the saving. Without observing all compensation and all informal work, the evidence cannot assign every dollar of relief, but it can reject the idea that the only relevant outcome is a one-for-one wage adjustment.
Even with those qualifications, the null result is central. If the tax saving were passed through mainly as higher pay, the clearest evidence should appear in the wages of workers hired after the cost change. Instead, the paper observes a sizeable employment response alongside little measured wage movement. The result makes it harder to explain the reform through a standard story in which wages immediately absorb the employer-side tax change and employment remains unchanged.
One possible explanation would be wage rigidity: legal rules, internal pay structures, or norms might prevent firms from lowering or raising wages freely. The paper’s contribution is to show that an employment response does not require the entire explanation to rest on such frictions. Its oligopsony framework allows several employers to compete for labor while retaining some power over wages. A firm may pay less than a worker’s marginal contribution, but it cannot reduce pay without making recruitment and retention more difficult. Lowering labor cost can then make expansion attractive even when the wage paid to a new hire hardly changes.
Local labor-market concentration gives that mechanism a testable geographic dimension. In a less concentrated market, a treated firm faces more employers that could hire the same workers. It has less scope to benefit by quietly keeping the saving while offering a less attractive wage than its competitors. Expanding employment is a more plausible use of the relief. In a more concentrated market, workers may face fewer comparable employers, so firms have more room to retain the gain as profit. The average wage estimate can remain small in both settings, while the employment response differs sharply.
The model does not imply that wages must never rise. Its general prediction is that a labor subsidy can increase both employment and wages, with employment effects larger than wage effects for most plausible values of the relevant substitution elasticities. The Brazilian estimates fit the employment-dominant part of that prediction. They provide weaker information about a concentration gradient in wages because the wage coefficients are small and imprecise. This is an important distinction: the data are more decisive about where the jobs response appears than about the exact pattern of wage incidence.
The selective nature of the PBM is again important. The reform did not lower costs for every employer competing in a local market. A treated standard-regime firm could become more competitive relative to firms left under another regime or outside the eligible sectors. A tax change common to all employers would not create the same relative-cost difference. In that alternative setting, firms might compete the saving away through wages, or the change might have a different effect on total employment. The paper’s result therefore belongs to a policy with selective exposure, not to the abstract phrase “payroll-tax cut” alone.
This is also why the average employment estimate should not be read as a direct measure of worker welfare. More formal jobs can have value beyond the recorded hourly wage, but this paper does not estimate those gains. A job increase can coexist with a large transfer to firms, and the paper’s evidence does not by itself show how gains are distributed among workers, incumbent employees, entrants, and employers. Local labor-market concentration is not just a technical moderator of the coefficient; it is a clue about who has the bargaining position to capture the relief.
The employment and wage estimates also do not yet identify the source of the added employment. The sector may have grown because existing firms hired more workers, because new firms appeared, because informal work became formal, or because firms changed tax regimes. Nor do they fully settle whether untreated employers experienced local spillovers. The event study answers the first question—whether formal sector employment changed after the PBM, relative to the comparison path—before the channel analysis answers the second.
That ordering protects the interpretation. The tax cut is associated with more formal employment, the response begins after implementation, larger expected reductions are linked to larger increases, and the available new-hire wage measure does not move significantly. These are meaningful findings, but they are not yet a count of net new jobs or a complete statement about worker welfare. To understand what “more employment” means, the next step is to follow the response through firm growth, observed entry, tax-regime switching, and formalization.