JFM referees know that market-quality variables are endogenous to almost everything — volume, volatility, information arrival, and prices co-move mechanically. So the bar is high for any causal liquidity/price-impact claim, and the journal especially rewards designs built on exogenous changes in market structure. The credible JFM toolkit is design-based, not control-saturated.
A recurring identification subtlety in market-structure work is that a rule change has both a mechanical effect (a wider tick arithmetically widens the minimum quotable spread) and a behavioral effect (liquidity suppliers and informed traders re-optimize). A credible JFM design isolates the behavioral channel, because the mechanical one is not a finding. Show the effect on stocks where the tick does not bind, decompose the spread change into the binding-tick component and the residual, or condition on pre-period spread relative to the new tick. Conflating the two is a frequent reviewer catch.
The SEC Tick-Size Pilot widened the quoting/trading increment for a randomized set of small-cap stocks. This is close to an ideal JFM design: random treatment assignment, a discrete date, and a treated/control split. The clean identification statement is one sentence — the effect of a wider tick on depth is identified by the random assignment of stocks to the pilot's test groups. The credible version shows flat pre-pilot trends in depth, estimates with assignment-level clustering, and separates the mechanical (tick-binding) effect from the behavioral (liquidity-supply) response. A weak version regresses depth on a post-dummy with controls and calls it causal.
Estimate and audit the identification claim, don't only argue it. Full map:
execution-with-mcp. JFM is market microstructure and asset pricing — liquidity, price discovery, and cross-sectional return tests where the factor-zoo multiple-testing haircut is salient.
detect_design → recommend → fit with as_handle=true → audit_result to list
the checks the design still owes.callaway_santanna / sun_abraham + bacon_decomposition +
honest_did_from_result (the pre-trend test is low-power, Roth 2022).effective_f_test + an anderson_rubin_ci (valid under weak instruments),
not a 2SLS t-stat alone.rdrobust (bias-corrected) + rddensity / mccrary_test for manipulation.oster_delta / sensemakr — how strong a confounder would have to be.Report the economic magnitude; route the full battery to the appendix; keep every
number reproducible. A run end-to-end (synthetic data, real returns) is in the
JF execution walkthrough. If StatsPAI/Stata are not connected, adapt the
vendored resources/code/ skeleton and flag any unverified number.
Knowing the field's natural experiments speeds design. Commonly exploited exogenous changes, each with its own caveats: decimalization (2001) — tick size from sixteenths to pennies; the SEC Tick-Size Pilot (2016-18) — randomized, the cleanest assignment; Reg NMS (2007) — order protection and access fees; MiFID (2007) / MiFID II (2018) — European venue competition and transparency; short-sale bans (2008, and country-specific) — abrupt constraint changes; maker-taker / fee pilots — rebate structure; circuit breakers / LULD — discrete trading halts; index reconstitutions — forced, scheduled order flow; venue launches/closures and dark-pool entry. For each, the identification hinges on (a) whether assignment is plausibly exogenous to the stock's liquidity trajectory and (b) whether a clean control group exists. Argue both explicitly; a shock is not self-justifying.
Identification is not finished until inference matches the data structure. Microstructure panels are correlated in two dimensions — the same stock is autocorrelated over time and all stocks co-move on a given day — so single-clustered or plain OLS standard errors overstate precision. Default to two-way clustering by stock and by day; use Newey-West when the time-series autocorrelation is the dominant concern; use a wild-cluster bootstrap when the number of treated venues or events is small (few-cluster bias). For event studies, account for cross-sectional correlation in abnormal liquidity across the event window. State the choice and its rationale where the design is described, not as an afterthought — a referee reads the clustering as part of the identification claim.
【Journal】Journal of Financial Markets (JFM)
【Skill】jfm-identification
【Branch】market-structure NE / intraday event / IV / structural
【Data-to-effect mapping】one sentence
【Identifying variation】<shock / window / instrument / moment>
【Endogeneity handled】how liquidity/flow endogeneity is broken
【Inference】clustering level + (if needed) weak-IV-robust set
【What it does NOT identify】<…>
【Source status】verified URL / 待核实 / not asserted
【Next skill】jfm-empirical-design