The standard finding is that women negotiate starting salary less often than men. The standard advice that follows is to negotiate more.
Both the finding and the advice are more complicated than that, and the complication is the useful part.
What the initiation research shows
Studies of negotiation initiation have found differences in propensity to negotiate, with the gap varying considerably by context.
The most informative finding in this literature is about ambiguity. When it is explicitly stated that salary is negotiable, the difference in initiation shrinks substantially or disappears. When negotiability is left unstated, the gap is largest.
This locates the mechanism. It is not a general reluctance to advocate; it is a difference in behaviour under ambiguity about whether advocacy is permitted.
The practical implication for an individual is direct: assume it is negotiable, because it almost always is, and the assumption is where the difference lies.
The implication for employers is equally direct: stating that an offer is negotiable, or removing negotiation entirely through banded pay, closes most of this particular gap at essentially no cost.
The backlash finding
The reason "just ask" is incomplete is that the same request has been found to produce different reactions depending on who makes it.
Experimental studies have found that evaluators rate women who negotiate as less likeable and report less willingness to work with them, at rates higher than for men making identical requests. The effect has been found more strongly when the evaluator is male.
Effect sizes vary and some replications have found weaker effects, so this should not be treated as an iron law. But the direction has appeared often enough that advising women to negotiate exactly as men do, without acknowledging a differential cost, is incomplete advice.
The correct response is not to avoid negotiating. It is to use framings that have been found to reduce the backlash effect.
Framings with evidence behind them
Relational account. Studies have found that framing a request with reference to a legitimate external reason and to the ongoing relationship reduces the social penalty. In practice: explaining briefly why the request is being made and connecting it to working well together, rather than presenting it as a pure demand.
Negotiating on behalf of something. Research has found that women negotiate more assertively and with better outcomes when representing someone else's interests than when representing their own, and that the backlash effect is smaller. The framing "this is what the role requires to be done properly" is doing some of this work.
Anchoring with external data. A request grounded in market data is harder to read as a personal demand. It also raises the anchor, and anchoring effects in negotiation are among the more robust findings in the field.
Asking about the range rather than stating a number first, when you have poor information. The first-offer advantage is real but depends on being informed. Anchoring low is worse than not anchoring.
What to negotiate other than salary
Base salary is the most consequential item because everything compounds from it, but it is frequently the least flexible, particularly in structured organisations.
Items that are often more flexible and materially valuable: start date, signing payment, review timing, title, professional development budget, remote or flexible arrangement, and — most underrated — the terms of the first performance review, including when it occurs and against what criteria.
Negotiating an early review with specified criteria converts an uncertain future into a defined one, and is frequently agreed because it costs nothing now.
The information problem
Most first negotiations fail on information rather than on nerve. A candidate who does not know the band is not in a position to negotiate regardless of technique.
Salary transparency legislation in a number of jurisdictions has made ranges visible in job advertisements, and studies of these regimes have generally found effects on posted pay and some narrowing of gaps. Where such information exists, it should be used.
Where it does not, the substitutes are professional body surveys, published pay scales in comparable public-sector roles, and asking people directly. The last is more effective than most people expect; the taboo on discussing pay operates overwhelmingly in employers' favour.
What the research does not support
It does not support the claim that closing the negotiation gap would close the pay gap. Negotiation differences account for a small share of the total, and the larger mechanisms are structural.
It does not support aggressive tactics. Studies of negotiation outcomes generally find that competitive tactics produce worse long-term relationships and, in employment contexts, worse outcomes than problem-solving approaches.
And it does not support the idea that failing to negotiate is a personal failing. In an environment where the same behaviour carries different social costs for different people, choosing not to incur the cost is a rational calculation, not a deficiency.
The single most useful sentence
If nothing else is retained from this: "Is there flexibility in the offer?"
It is a question rather than a demand, it costs almost nothing socially, it resolves the ambiguity that the research identifies as the actual mechanism, and it very frequently produces a yes.