When the Securities and Exchange Commission proposes a new rule, it follows a “notice-and-comment” process. It first discloses the proposed changes, then it invites constituents to write comment letters addressing those changes, and finally it settles on what to adopt. That open invitation is at the heart of the “notice-and-comment” process the Administrative Procedure Act has required since 1946. But which comment letters’ views are actually incorporated into the rules?
In a new paper, we try to answer that question on a scale that has not been possible before. Using large language models to read and code more than 65,000 comment letters covering 417 rules proposed between 1995 and 2024, we map who said what and how those stated positions line up with the SEC’s text at each stage of rulemaking. The picture we find is familiar in tone but newly precise in detail: Resourceful actors’ views tend to appear in proposals and final rules, while retail investors’ views are most likely reflected in the SEC’s decision to drop a proposed rule.
The novelty here is largely methodological. Past studies of SEC rulemaking have mostly looked at handfuls of rules or narrow groups of commenters, because reading and classifying tens of thousands of letters by hand is impractical. We instead use Google’s Gemini 2.0 Flash to pull every request for comment (“RFC”) out of each proposing release, then match each answer in each letter to the specific question it addresses. The result is a dataset built at the answer-to-question level, with roughly 395,000 individual positions. We also classify each writer into one of four groups: firms, sophisticated investors (think hedge funds and institutional players), experts (academics, auditors, lawyers), and retail investors. We then ask whether each writer’s stated position agrees with the rule text at two stages: the original proposal and the eventual outcome—adopted or dropped.
At the proposal stage, firms have the clearest advantage. Their stated preferences agree with the SEC’s draft text 64 percent of the time, compared with 39 percent for retail investors. After controlling for the specific question being asked, we find that firms are about 7.9 percentage points more likely than the average letter writer to be aligned with a proposal. Retail investors are about 7.1 percentage points less likely. Experts and sophisticated investors fall in between.
But that picture changes by the time a rule is adopted. Firms’ advantage shrinks dramatically—to roughly 1.2 percentage points for RFCs that ultimately get adopted, and it turns sharply negative (about minus 8.8 percentage points) for RFCs the SEC drops. Retail investors flip in the opposite direction. They are not more aligned with the rules that get adopted; in fact, they are slightly less so. But they are much more aligned with the RFCs the SEC ends up dropping, by about 7 percentage points. We read this as consistent with a procedural reality: Modifying a proposal can be legally risky and procedurally costly, so when commenters push back hard, abandoning a proposed provision is often easier than rewriting it. Retail investors, in other words, seem to influence what does not happen more than what does.
We next ask whether speaking with one voice helps a group’s positions show up in the final rule. We define a group as “in agreement” when at least 80 percent of its writers take the same position on an RFC, then test whether the outcome tracks that internal consensus. The answer is generally yes—unified groups tend to see better alignment—but the result varies by the identity of the writer. When firms, sophisticated investors and experts agree among themselves, that consensus is associated with RFCs being adopted. Retail investor consensus, by contrast, is again associated with dropped RFCs. Sophisticated investors are the only group whose unified voice tracks the outcome whether the RFC is adopted or dropped.
When groups disagree with each other, the losers are predictable. We look at scenarios where at least 80 percent of one group takes one position while at least 80 percent of another takes the opposite position. Retail investors and experts are the two groups most likely to come out on the losing side. When firms and retail investors clash, retail investors are about 4.4 percentage points less likely to see their views aligned with the outcome. When experts and retail investors clash, both lose ground, and firms’ alignment actually goes up.
To explain these patterns, we look at proxies for resources. Two stand out. First, participation in associations—industry groups like SIFMA, professional organizations, retail-investor nonprofits—is associated with greater alignment at both the proposal and adoption stages. The views of firms that submit comments through an association are about 5 percentage points more likely to be aligned with proposed rules and 2.5 percentage points more likely to agree with adopted rules. Second, meetings with SEC officials are associated with sizable jumps in proposal-stage alignment. Firms that meet with the agency are roughly 7.2 percentage points more likely to see their views aligned with proposals. The retail investors who manage to meet with the SEC see an even larger jump, but this happens rarely—only about 1 percent of retail-investor responses come with a meeting, versus 15 percent of firm responses.
We also examine which letters are cited in the adopted rules. Adopting releases often quote specific comment letters to explain regulatory choices. Firms are more likely to be cited; retail investors are less so—consistent with the alignment results. But experts appear to be cited far more often than their actual influence on the final rule would predict. The SEC, in other words, seems to use expert letters to frame the rationale for adopting rules even when those letters’ specific positions did not carry the day.
Finally, we ask which groups’ support is associated with a proposal being adopted, rather than dropped. Taken one at a time, each group’s support individually predicts adoption. But when we place all four groups in the same regression, retail-investor support falls out as a statistically significant predictor while firm, sophisticated-investor and expert support remain. We read this as further evidence that retail-investor participation behaves more as a veto than a constructive force: it is associated with proposals dying, not with new rules being born.
There are limits to the conclusions we can draw from our research. We measure alignment between stated positions and rule text, not the causal effect of comment letters on outcomes. Comment letters are one factor regulators account for in designing regulations. Other factors, such as meetings, unofficial communications (a concern that SEC Chair Paul Atkins has publicly flagged) and pre-proposal interactions, matter too. We also do not claim that our findings prove regulatory capture; an agency that consistently sides with one group could be doing so for other, good reasons given the SEC’s often conflicting missions of protecting investors, encouraging capital formation and defending transparent markets.
Still, the empirical picture is striking, especially for those who read SEC rulemaking in the context of administrative law. Notice-and-comment is the formal channel for democratic participation in agency rulemaking. Our evidence suggests that participation alone does not translate into representation. The actors with resources, organization, and access–firms, large investors, professional associations–are more likely to see their preferences reflected in the text the SEC publishes. Retail investors do have influence, but mostly on the rules that are never adopted. For lawyers, scholars, and policymakers thinking about how to make rulemaking work better, our takeaway is that what looks like an open forum can produce lopsided outcomes—even without any single actor doing anything wrong.
Chuck Downing is a PhD student at MIT’s Sloan School of Management, Gabriel Pündrich is an assistant professor at the University of Florida’s Warrington College of Business, and Gabriel Voelcker is an assistant professor at Dartmouth College’s Tuck School of Business. This post is based on their working paper, “Constituent Interests in SEC Rules: Evidence from Voluntary Comment Letters,” available here.