jbes-literature-positioning

Literature Positioning (jbes-literature-positioning)

When to trigger

• A referee will ask "how is this different from method X already in the literature?"
• The contribution relative to the closest existing estimator/test/algorithm is fuzzy
• You are unsure whether your improvement is incremental or genuinely new
• You need to map your method onto the right strand (time series, panel, GMM, ML, Bayesian, etc.)

Why positioning is the methods-paper crux at JBES

JBES referees are method experts: they judge a paper first by what it adds to the existing toolkit. Because the journal explicitly welcomes adaptation of methods from machine learning and data science alongside classical econometrics, your closest competitors may live in two literatures at once — the statistics/ML method you adapt and the econometric problem you apply it to. Position against both. The contribution must be stated as a delta against named prior methods, not as a freestanding survey: which assumptions you relax, which rates you improve, which computational barrier you remove, or which empirical setting prior methods cannot handle.

Positioning protocol

1. Name the incumbents. List the 3–6 methods a referee would consider state of the art for your problem (estimator, test, or algorithm — with citations).
2. State your delta per incumbent. For each: weaker assumptions? better asymptotics/rates? valid under dependence/heavy tails/high dimension where they are not? faster/feasible at scale? Honest deltas only.
3. Locate the strand. Place the paper in its method family (e.g., HAC inference, GMM, factor models, quantile methods, debiased ML, Bayesian computation) so referees from that strand recognize the lineage.
4. Bridge to the application. Tie the methodological delta to the empirical payoff: the new method changes a substantive conclusion or makes a previously infeasible analysis feasible.
5. Concede gracefully. Where an incumbent dominates (simpler, or better in a regime you do not target), say so; over-claiming invites a hostile report.

Checklist

• 3–6 closest prior methods named and cited
• A concrete delta stated against each (assumptions / rates / robustness / computation)
• Both the statistics/ML side and the econometrics side covered if you adapt across fields
• The method family / strand is explicit
• The positioning connects to the empirical payoff, not just abstract properties
• Limitations vs. incumbents conceded honestly

Anti-patterns

• A chronological survey ("Smith (1990) did..., Jones (1995) did...") with no delta
• Comparing only to a strawman or to outdated methods, ignoring the current frontier
• Claiming novelty against econometrics while missing the identical idea in the statistics/ML literature (or vice versa)
• Vague superiority ("our method performs better") with no stated dimension of improvement
• Hiding the regime where an existing method still wins

Worked vignette: positioning a debiased-ML estimator across two literatures

A hypothetical JBES paper adapts double/debiased machine learning to estimate a heterogeneous treatment effect of credit-score thresholds on default, using bank loan-level data (figures illustrative). Because the idea lives in two literatures at once, the positioning must hit both. On the statistics/ML side the closest prior work is cross-fitted DML and causal forests; the delta is a dependence-robust cross-fitting scheme valid under the within-branch clustering of loan data, which iid DML ignores. On the econometrics side the incumbents are series/sieve semiparametric estimators; the delta is an illustrative 30% RMSE reduction at the same nominal coverage when nuisance dimension is high. The paper concedes that plain DML still wins under independence and low nuisance dimension — naming where an incumbent dominates pre-empts a hostile report. The delta then ties to the application: it changes which credit-score band shows the largest effect.

Referee-pushback patterns on positioning (venue-specific fixes)

Calibration anchor (hedged): JBES welcomes machine-learning and data-science adaptations, so your nearest competitor often sits outside econometrics — missing the identical idea in statistics/ML invites the sharpest rejection.

Output format

【Incumbents】[3–6 prior methods + citations]
【Delta per incumbent】method → what you improve (assumptions/rates/robustness/computation)
【Strand】method family this paper joins
【Cross-field check】statistics/ML side AND econometrics side covered? [Y/N]
【Empirical payoff】how the delta changes a substantive result
【Conceded】where incumbents still win: ...
【Next step】jbes-contribution-framing