Building a Model to Measure Local Economic Performance
The Yelp Economic Average (YEA) is a composite measure of the economy, reflecting both business health and consumer demand among businesses in 30 sectors.
The eight root categories
The 30 business sectors, or categories — the “Yelp 30″ — are drawn from eight umbrella business categories on Yelp: restaurants, food, nightlife, local services, automotive, professional services, home services, and shopping.
Root categories’ share of the 30 components
The share of YEA components from each of these eight categories is based on each one’s share of the economy, as estimated from County Business Patterns reports.
Choosing the Yelp 30
Each of the Yelp 30 is chosen based on maximizing four criteria, relative to other candidates within its family of categories, as measured in the first quarter of 2016:
- Number of businesses on Yelp in the category;
- Consumer interest on Yelp for businesses in the category, as measured by activity such as page views, reviews, and photos;
- Number of the 50 metro areas — whose economic health we have been measuring a year and a half, originally as part of our Local Economic Outlook — in which the category is present;
- Uniform spread across the four Census Bureau-defined regions of the country.
Choosing baseline categories
We then chose baseline categories against which to compare the fortunes of the Yelp 30. This step helps remove changes due to seasonality and Yelp’s internal growth; what remains is a reflection of real economic patterns. We selected all other root categories not represented by the YEA components as baselines because they provided the most robust controls against seasonality and activity on Yelp.
Calculating the YEA scores
For each of the Yelp 30 in each quarter, its two scores — one for business population and one for consumer interest — are calculated as follows:
- Count the component’s total for the quarter;
- For consumer interest only: Count the baseline’s total for the quarter;
- For consumer interest only: Divide the component’s total by the baseline total to get the component’s score;
- Divide the component’s score for the quarter of interest by the component’s score in the equivalent quarter in 2016 — comparing, for instance, the fourth quarter of 2018 to the fourth quarter of 2016, to adjust for seasonality;
- Multiply by 100 to make 100 a typical score.
The two scores are then normalized to have the same variance, so that each contributes equally across components.
To reduce the effect of outliers, the overall score for both consumer engagement and business count is the median of each component’s score.
The YEA is the mean of the overall consumer engagement score and business-count score.
The YEA is separate from, and not meant to inform or predict, Yelp’s financial performance because our figures are adjusted to remove the effects of changes to usage of our product.
We calculated equivalent scores at the regional, state, and metro level to provide a local look at the state of the local economy. We also calculated scores by political party. We did so by reviewing the performance of states’ local economies (through business survival and consumer business interest), comparing states based on the political party that won the state in the 2016 presidential election.
We calculated annual scores by computing the mean of scores in each year’s four quarters.
Forecasting YEA scores
We used Facebook’s Prophet package to forecast the current quarter’s YEA scores based on prior quarter’s scores, at the national, business sector, and local levels.
Comparing YEA to GDP Growth
We compare the change in both indicators rather than the absolute magnitude because they’re measured and reported differently.