Houston festival attendance and economic-impact evidence, 2009–2017

Two festivals, nine years, one American city with a large and well-covered local press. That should be a good case for measuring what live music does to a metropolitan economy. It is not, and the reason is instructive: the published record consists of a handful of figures produced by different parties, using different measures, for different purposes, none of them audited and only one of them accompanied by a stated method. This study sets every one of them out with the measure it represents and the limitation that travels with it, and then declines to draw a trend.

The complete published record

Five measures that are not one measure

The table contains five distinct measure types and they answer different questions. Reported total attendance counts bodies as a reporter estimated them. Study attendance is a model input. Paid attendance counts admissions sold. Ticket price is what was advertised, not what was realised after fees, comps and tier mix. Economic output is a modelled multiple of assumed spending. Placing these on a single axis would produce a chart in which a press estimate, a model input and a recollection appear as three observations of the same quantity. They are not, and the archive stores the measure type on every figure precisely so that this cannot happen by accident.

This is not a Houston problem. It is the standard state of festival evidence in the United States, and it is why claims about the economic contribution of live events are so difficult to check. The figures that circulate most widely are usually the largest available number, stripped of the measure that made it meaningful.

The one study with a documented method

The 2012 University of Houston analysis is the only figure set in the record with a stated approach. As reported, it produced approximately $14 million in total local output, 104 full-time-equivalent jobs, $5.4 million in labour income and $8.2 million in value added, on an attendance input of 81,000 and with more than a quarter of attendance attributed to out-of-town visitors ¹.

Its limits are the ordinary limits of commissioned input-output work. It was paid for by the party whose event it measured. It models rather than observes: output, income and value added are derived from a spending assumption run through regional multipliers. It depends on the attendance input and on the visitor share, both of which are estimates. And its treatment of displacement — spending that would have occurred in the city anyway, redirected rather than created — is not visible in the reporting. None of that makes the study worthless. It makes it a single, dated, sponsor-commissioned estimate for one edition, which is exactly how it should be quoted.

The two figures that must not be upgraded

Two entries in the table are frequently quoted as though they were measurements, and neither is. The first is the Day for Night 2016 figure of nearly $20 million for the city. It appears in the record only as a study referred to in a later Houstonia retrospective ². The underlying study has not been retrieved; its sponsor, method and geographic boundary are unknown. It cannot responsibly be compared with the University of Houston figures, and it is recorded here as a secondary reference rather than as evidence of impact.

The second is the 2017 production cost of $6 million to $7 million per edition. This is the founder's account in a 2019 interview, given alongside the statement that none of the three editions was profitable ². It is a producer's forecast-style statement about cost, not a measured economic impact, and it says nothing about what the event returned to the city. Any statement of local impact derived by multiplying a production cost is a producer's projection and should be labelled as one wherever it appears.

What better evidence would look like

• Attendance stated as a measure: gate scans, tickets sold, or unique attendees, with the definition printed next to the number.
• Impact studies published in full, with the spending survey, multipliers, visitor share and displacement treatment disclosed, not summarised in a press release.
• Prices recorded as realised yield alongside advertised face value, so fee load and tier mix are visible.
• A stated geographic boundary for every impact claim, because a figure for the city and a figure for the metropolitan area are different figures.
• Independent sponsorship, or failing that, an explicit statement of who paid for the study.

Until those conditions are met, the useful thing an archive can do is refuse to average. Each row above is retrievable, dated and attributed, and each is labelled with what it actually measures. That is a smaller claim than a trend line and a far more durable one.