New Working Paper Examines Sensitivity of Synthetic Control Methods in Evaluating California Fast Food Minimum Wage Effects

The ongoing academic debate surrounding the economic consequences of minimum wage legislation has received a significant methodological update with the release of National Bureau of Economic Research (NBER) Working Paper 35931. Published in October 2026 and authored under DOI 10.3386/w35931, the study casts a critical eye on the analytical tools economists use to evaluate public policy. Specifically, the research evaluates synthetic control methods—a popular statistical technique used to build credible counterfactuals by matching pretreatment outcome paths—and demonstrates how sensitive these models can be to minor variations in specification choices.

By testing California’s landmark $20 fast-food minimum wage law through 1,440 plausibly defensible analytical specifications, the researchers uncovered a wide dispersion of estimated employment effects. Although an overwhelming majority of these specifications—over 96 percent—yielded negative estimates for employment changes, the absolute range of the outcomes varied dramatically. This wide spread underscores a vital vulnerability in modern empirical research: the susceptibility of complex methodologies to the less-visible degrees of freedom exercised by researchers during the modeling process.

Background Context and Policy Origins

To understand the weight of Working Paper 35931, one must examine the contentious policy landscape that inspired it. California’s legislative journey toward a sector-specific minimum wage began in earnest with Assembly Bill 1228 (AB 1228). Signed into law to address labor standards, wage adequacy, and collective bargaining structures within the fast-food industry, AB 1228 authorized the creation of a Fast Food Council empowered to set industry-wide labor standards, culminating in a mandatory baseline wage of $20 per hour for fast-food workers across the state.

Proponents of the legislation argued that a substantial wage floor was long overdue for frontline workers who faced high living costs, demanding physical labor, and historical wage stagnation. They contended that higher pay would reduce employee turnover, boost morale, and inject crucial purchasing power into local consumer markets without significantly hampering business viability. Conversely, opponents, including prominent restaurant associations, franchise owners, and business advocacy groups, warned that an abrupt and steep hike in labor costs would trigger immediate inflationary pressures, accelerate automation initiatives, and force widespread job cuts or reduced operating hours.

As economists rushed to measure the empirical reality of AB 1228 following its implementation, they deployed a variety of econometric frameworks. The labor market effects of California’s policy have been fiercely debated, with varying conclusions documented in both the initial and revised versions of NBER Working Paper 34033, as well as in NBER Working Paper 35171. These conflicting findings highlighted the urgent need to look beyond the headline numbers and interrogate the statistical machinery generating them.

Methodological Deep Dive: Synthetic Control Versus Difference-in-Differences

At the heart of the new NBER study is a methodological critique of synthetic control methods. In modern applied econometrics, when a policy is implemented in a specific jurisdiction—such as California—researchers often struggle to find a single, identical control state that mirrors the treated region’s prior economic trajectory. Synthetic control methods attempt to solve this problem by creating an artificial "control state" constructed as a weighted average of un-treated states that, when combined, closely match California’s pretreatment economic path.

However, the construction of these synthetic controls requires numerous researcher decisions: which donor states to include, which economic predictors to prioritize, what time frame to use for the baseline fit, and how to weight various lagged outcomes. These choices represent the "less-visible degrees of freedom" that can subtly steer the final empirical verdict.

To quantify this sensitivity, the authors of Working Paper 35931 constructed 1,440 distinct, defensible specifications for estimating the employment impact of the $20 fast-food minimum wage. The results demonstrated that while the directional finding remained remarkably consistent—with more than 96 percent of the models pointing toward a negative effect on employment—the magnitude of that effect swung wildly depending on the precise econometric pathway chosen.

Interestingly, the study provides a comparative baseline by evaluating simpler and far more transparent econometric techniques, such as difference-in-differences methods. The estimates derived from these traditional difference-in-differences models consistently landed near the centers of the comparable distributions generated by the complex synthetic control specifications. This finding suggests that while sophisticated models promise greater precision by closely fitting pretreatment paths, they may also introduce unnecessary fragility if their underlying specifications are not rigorously stress-tested across multiple plausible scenarios.

Chronology of the Debate

The release of Working Paper 35931 in October 2026 marks the latest chapter in an evolving multi-year academic and public policy chronology:

• September 2023: California passes AB 1228, paving the way for the establishment of the Fast Food Council and the scheduled implementation of the $20 minimum wage.
• April 2024: The policy takes effect, immediately triggering intense public debate, media scrutiny, and corporate restructuring across major fast-food chains in the state.
• Late 2024 to 2025: Early economic working papers, including initial drafts of NBER Working Paper 34033 and Working Paper 35171, begin circulating. These papers report divergent employment impacts, ranging from negligible job losses to more pronounced reductions in low-wage workforce counts.
• October 2026: NBER releases Working Paper 35931, shifting the analytical focus from the raw employment figures alone to a rigorous meta-evaluation of the econometric tools used to calculate them.

Implications for Economic Research and Public Policy

The insights contained within Working Paper 35931 extend far beyond the debate over California’s fast-food industry. They strike at the core of how empirical economics informs public policy design in the United States and globally.

In policy circles, research findings are frequently weaponized by political actors on both sides of the aisle. Proponents of labor market interventions often highlight studies showing minimal disemployment effects, while critics point to research indicating severe job destruction. By demonstrating that synthetic control models can yield a vast array of quantitative estimates depending on minor specification choices, the new working paper serves as a cautionary tale for policymakers and science communicators alike. It highlights the danger of relying on a single, isolated empirical specification to justify sweeping economic interventions or condemnations.

For the academic community, the paper provides a strong methodological mandate: transparency and robustness checks must become standard operating procedure. When researchers deploy advanced matching techniques, they must transparently report how sensitive their conclusions are to alternative formulations of the control group and baseline periods. The fact that traditional, more transparent difference-in-differences estimators produced results centered within the distribution of the synthetic control estimates also breathes new life into classical econometric tools, reminding researchers that complexity does not automatically equate to superiority.

As states and municipalities continue to experiment with sector-specific wage floors, occupational licensing reforms, and localized labor market regulations, the methods used to study these interventions will remain under intense scrutiny. Working Paper 35931 ensures that future economic evaluations will be approached with a heightened awareness of researcher degrees of freedom, ultimately pushing the discipline toward more robust, reproducible, and transparent policy analysis.