What one star is worth

A single star is worth measurable money, and economists have been pricing it for two decades. Independent restaurants in Seattle gained 5 to 9 percent more revenue for every extra star in their average Yelp rating, according to a Harvard Business School analysis of city tax records. A half-star lift made San Francisco restaurants 49 percent more likely to sell out at peak hours. Hotels across eleven cities, from London to Los Angeles, could raise prices 11.2 percent for the same booking odds once their review score climbed a full point. Four separate research designs, four different platforms, one consistent direction: a rating moves demand before a customer ever walks in.

The Harvard estimate: five to nine percent a star

The most-cited number in the reputation-economics literature comes from Michael Luca, who matched every Yelp review of a Seattle restaurant to that restaurant’s actual revenue reported to the Washington State Department of Revenue between January 2003 and October 2009 — 3,582 restaurants in total, an average of 1,587 trading in any given quarter. Rather than simply correlating higher ratings with higher sales, Luca exploited a quirk in how Yelp displays scores: averages are rounded to the nearest half star, so a restaurant rated 3.24 shows as 3 stars while one rated 3.25 shows as 3.5 — a difference invisible in underlying quality but highly visible to a browsing customer. Comparing restaurants on either side of that rounding line isolates the causal effect of the displayed rating rather than the food behind it. The result: a one-star increase raised revenue 5 to 9 percent, but only for independent restaurants. For the 143 chain-affiliated restaurants in the sample — about 5 percent of the market in any quarter — the effect was statistically indistinguishable from zero, because a chain’s brand already does the reputational work a rating does for an independent.

The pattern repeats in reservations, hotel rooms and book sales

Luca’s rounding-threshold design has since been reused across formats, and the direction holds. In the San Francisco Bay Area, economists Michael Anderson and Jeremy Magruder applied the same trick to restaurant reservation data and found that an extra half-star rating made a restaurant 19 percentage points — 49 percent — more likely to be fully booked during peak hours, with the effect largest where diners had the least other information to go on.

Hotels show a comparable pattern at a different order of magnitude. A Cornell Hospitality Report tracked monthly STR performance data against ReviewPro’s Global Review Index and Travelocity ratings across 11 major markets — Berlin, Chicago, London, Los Angeles, Madrid, Miami, Milan, New York, Prague, Rome and San Francisco — from January 2010 to June 2012, more than 50,000 property-months in all. A 1 percent rise in a hotel’s reputation score was tied to up to a 0.89 percent increase in average daily rate, a 0.54 percent increase in occupancy and a 1.42 percent increase in revenue per available room. Scaled up, a hotel that lifted its review score a full point — from 3.3 to 4.3, say — could raise its price 11.2 percent while keeping the same odds of being booked.

Books show the asymmetry more than the average. Judith Chevalier and Dina Mayzlin compared sales-rank movements for the same titles at Amazon.com and Barnesandnoble.com — 2,387 books tracked across data-collection windows in May 2003, August 2003 and May 2004 — and found that one-star reviews dragged sales down harder than five-star reviews pulled them up. In one illustrative case from the paper, turning a single five-star review on a mid-ranked title into a one-star review was modeled to cost that book roughly 20 sales a week.

Four studies, four sectors, one direction

The headline numbers are not directly comparable — different rating scales, different outcome variables, different eras — but laid side by side they show the same sign every time.

Why the money moves asymmetrically

The mechanism repeats across all four studies. Consumers facing hundreds of reviews rarely read them all — Luca notes that many Yelp restaurants carry upward of 200 — so the rounded average becomes a shortcut for judging quality that would otherwise take real time to assess. That shortcut carries more weight precisely where other signals are weakest: independents without brand recognition, restaurants competing for a single peak-hour table, hotels in an unfamiliar city. It also carries more weight from certain reviewers — Luca finds that reviews from Yelp’s “elite” members move revenue almost twice as much as ordinary reviews — and it carries more weight downward than upward, matching the loss-averse reading pattern Chevalier and Mayzlin document at Amazon. None of the four studies claims its exact multiplier transfers cleanly to how Google Business Profile, Trustpilot or Tripadvisor display ratings today; each result is tied to the specific rounding and ranking mechanics of the platform it studied.

Methodology

Sources

Michael Luca, “Reviews, Reputation, and Revenue: The Case of Yelp.com,” Harvard Business School Working Paper 12-016 (2011, revised 2016) — 3,582 Seattle restaurants matched to Washington State Department of Revenue tax records, January 2003–October 2009. Michael Anderson and Jeremy Magruder, “Learning from the Crowd: Regression Discontinuity Estimates of the Effects of an Online Review Database,” The Economic Journal 122(563), 2012 — San Francisco Bay Area restaurant reservation data matched to Yelp.com, data as of 2011. Chris K. Anderson, “The Impact of Social Media on Lodging Performance,” Cornell Hospitality Report 12(15), November 2012 — STR performance data for hotels in 11 markets (Berlin, Chicago, London, Los Angeles, Madrid, Miami, Milan, New York, Prague, Rome, San Francisco) matched to ReviewPro and Travelocity ratings, over 50,000 property-months, January 2010–June 2012. Judith Chevalier and Dina Mayzlin, “The Effect of Word of Mouth on Sales: Online Book Reviews,” Journal of Marketing Research 43(3), August 2006 — 2,387 books tracked for sales-rank and review changes at Amazon.com and Barnesandnoble.com across data-collection windows in May 2003, August 2003 and May 2004.

Scope limits

All four studies rely on pre-2013 data from US platforms; the hotel study also covers six European cities and London, but none of the four has a Ukraine-specific or post-2015 replication we could verify for this brief, so the exact percentages should not be read as current benchmarks for today’s review platforms. Each result is tied to the specific display mechanics of the platform studied — Yelp’s half-star rounding, Amazon’s and Barnesandnoble.com’s sales-rank systems — and the original authors do not claim the effect sizes transfer automatically to other rating systems or sectors. The Chevalier and Mayzlin finding is directional (one-star reviews weigh more than five-star reviews) rather than a single point estimate, and it is reported here as such rather than forced into a percentage the paper does not state.

Updates

This brief will be revised as new data is published. Corrections are handled under our corrections policy.