The Effectiveness of Financial Incentives for Health Behaviour ChangeSystematic Review and Meta-Analysis

Think about the last time you tried to do something healthy that didn't feel great in the moment, like going for a run after work or skipping dessert when everyone else dug in. The benefits are real, but they’re delayed and a little fuzzy. The costs? They're right now. That gap is the opening that financial incentives try to close. If you give people something immediate and certain for doing the healthy thing, or chip away at the immediate costs, you tilt the everyday calculus in favor of better choices. Giles and colleagues explored this idea. They focused on what they call health-promoting financial incentives, which are cash or cash-like rewards—such as vouchers you can spend—or, in some cases, penalties like reduced benefits, paid out only if you actually engage in the desired behavior. There have long been concerns that external rewards might crowd out intrinsic motivation. Most of that concern comes from lab studies. Giles and colleagues asked a simpler question in real-world settings: do incentives change behavior now, and does any of that change last when the money stops? They took a thorough approach to credibility. The review was preregistered on PROSPERO, followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidance, and combed nine major databases plus trial registries and citation trails up to 2012. They maintained a tight scope: non-clinical adults in high-income countries; real behaviors, measured objectively or with validated self-report; and incentives that were guaranteed if the behavior occurred. Lotteries could exist as add-ons, but their uncertain payouts didn't count toward "incentive value." All values were converted to 2011 U.S. dollars, so money meant the same thing across studies. What did that result in? Sixteen studies across seventeen papers. Ten targeted smoking cessation, five focused on attendance for vaccination or screening, and one targeted physical activity. Most provided cash or vouchers; a couple used deposit contracts, where you put in your own money and receive it back only if you reach the goal. The certain incentive values ranged from just over five dollars to about seven hundred and eighty-six dollars. Intervention durations varied significantly—from two weeks to two years for smoking cessation—and follow-up periods stretched from a few weeks to two years. Almost all the evidence came from the United States, which affects how widely we can generalize these findings. Let's start where the data are richest: quitting smoking. In the short term—up to about six months—offering money roughly doubled to tripled quit rates compared to usual care. The pooled relative risk was 2.48, with a confidence interval from 1.77 to 3.46. That's a big, clean signal. Importantly, in this short timeframe, the size of the incentive didn’t explain away the effect. Meta-regression showed quit rates didn’t consistently rise or fall with follow-up time within that six-month window, or with the total dollars on offer. The picture here is straightforward: money on the table now helps people stop smoking now. If you push the horizon beyond six months, the story changes, but it doesn't collapse. At longer follow-up, the pooled effect shrank to a relative risk of 1.50, with the lower bound barely clearing one. In plain terms, some of the initial gains faded once payments stopped, but not all of them. There was also much more variability between studies. One clue inside that scatter is that formats mattered. Cash-only incentives produced a larger long-run effect than formats that bundled cash with more complex elements, and only the cash-only subgroup achieved conventional significance. In this longer-term slice of the smoking literature, there was actually a hint that bigger payments mattered more—the meta-regression linked higher incentive values to larger effects, with a small but positive coefficient. So, for sustained abstinence, both size and simplicity may be important. Publication-bias checks for smoking didn’t show the typical asymmetry that can occur when only flashy positive results get published. If smoking is the big endurance test for incentives, vaccination and screening are the sprints. Just show up once for a shot or a test. Here, incentives worked, and worked reliably across various conditions—from influenza and hepatitis B vaccination to cervical screening and tuberculosis skin test reading. Pooled across nine comparisons, the relative risk was 1.92, with a confidence interval from 1.46 to 2.53. That represents a strong lift for one-off preventive actions. There was a twist concerning how programs were structured: cash layered with other motivational components outperformed cash or vouchers alone. When cash was paired with reminders or supportive messaging, the relative risk increased to about 2.75. Cash or vouchers by themselves landed closer to 1.77. This doesn’t mean pure cash doesn’t work—it does—but combining it with other nudges seemed to amplify the outcome in this domain. The variability here was high, which is researcher speak for "the details differed significantly," but the overall direction was the same. On physical activity, the evidence was scant. One study—by Finkelstein and colleagues—used pedometers to track steps over up to four weeks and paid cash for increases in activity. During the intervention, participants in the incentive group logged, on average, sixteen more minutes of activity per day compared to controls. If you force that into a risk ratio, you get a striking number over five, but the confidence interval is wide, which signals imprecision. With one short trial, you can’t draw broad conclusions about increasing walking for money, other than "it might help, but we need more than one snapshot." Putting everything together still shows a benefit. Across all behaviors at the longest follow-up available, the pooled relative risk was 1.62, with a confidence interval from 1.38 to 1.91. That global number comes with substantial variability; these are different behaviors in different people under different designs. But it tells you the core point: in non-clinical adults in wealthy countries, financial incentives outperformed usual care more often than not, and by a meaningful margin. Now, a layer deeper. When Giles and colleagues examined the entire dataset to see what happens as you follow people over time, the effect tends to decrease. The logarithm of the relative risk diminished as follow-up lengthened, with a small negative coefficient. In the same overall analysis, larger incentive values were also tied to slightly smaller effects, again with a small negative coefficient. This pattern contradicts the intuitive idea that more money always buys more change. It creates an interesting tension with the long-term smoking data, where larger payments were linked to larger effects. The takeaway is nuance: the relationship between size, timing, and behavior isn’t one-size-fits-all. How and when you pay may be just as crucial as how much you pay. A word about methods, because trust in the numbers relies on trust in the methodology. Random-effects meta-analyses were employed throughout, which is appropriate when you expect studies to differ beyond random variance. For smoking, analyses were divided into follow-ups of up to six months and beyond six months, and outcomes were taken at the longest available point per study to avoid double counting. When a single trial included multiple incentive arms, the shared control group was managed carefully to ensure it didn’t inflate precision. The risk of bias was assessed using Cochrane tools. Allocation sequence generation and concealment were weak spots in a few studies, but most were judged to have low or unclear risk overall, and one high-risk outlier wasn’t included in the meta-analyses because the data weren’t usable. Funnel plots for smoking didn’t indicate clear publication bias. This isn’t a perfect evidence base, but it's a conscientious reading of what we have. What should you take away from all this? First, incentives reliably boost short-term quitting, and a portion of that advantage persists months after payments end. For one-off preventive actions like vaccination or screening, paying people will likely increase attendance, and you may see even better results if you combine the payment with simple motivational supports. Second, the format and the time horizon matter. Cash-only approaches appeared more effective for maintaining smoking abstinence over time, while mixed approaches outperformed cash alone for encouraging participation in screenings. Third, more isn’t always better. Across the full set of behaviors, larger incentives were associated with slightly smaller effects, even as the long-term smoking data suggests the opposite. That’s a call to design with care, not a recommendation to cut costs. There are limits here. Nearly all the studies were based in the United States. Cultural norms, baseline access to care, and the social meaning of a cash reward can differ significantly across countries, so we don’t know how these programs would operate elsewhere. The variability was high in several pooled analyses, reflecting how differently these interventions were constructed. The evidence regarding complex, sustained behaviors beyond smoking—such as physical activity or long-term diet change—is limited. For policymakers or employers listening to this during your commute, the bottom line is pragmatic. If you want more people to do a specific healthy thing in the near term, paying them works. If you care about that change sticking, pay attention to design. Make rewards immediate and certain. Keep formats simple when the goal is sustained abstinence, and consider pairing payments with supportive cues when the goal is a one-off appointment. And don’t assume you can buy a linear dose response with ever-bigger incentives; the relationship is more complex than that. Looking ahead, the toughest and most valuable tests are those that haven’t been run yet. We need head-to-head trials that vary timing, certainty, and format to identify which levers matter for specific behaviors. We need studies conducted outside the United States to understand how context shapes responses. And we need longer follow-ups that tell us not only if people respond while the incentives are active, but how much value remains when they cease. For now, the ledger reads like this: money talks, especially in the short term. If you listen closely to how it speaks—when, how much, and in what form—you can make it communicate something useful about health.