Hotel Performance Metrics: Net Operating Income Forecasts
Looking Under the Top Line
By Bram Gallagher Economist, CBRE Hotels' Americas Research | October 01, 2017
The most widely reported hotel performance metrics include occupancy, average daily rate (ADR), and revenue per available room (RevPAR), with RevPAR frequently used as the measure of market financial strength. While RevPAR itself is a good proxy for profitability when the market is in a typical state, it potentially obscures the performance of hotels when the market veers away from the norm. Pass-through performance (i.e., the percent change in bottom-line income given the percent change in top-line revenue), or measures that are net of a market-wide average of expenses, are increasingly of interest as record-breaking occupancies make RevPAR movements an incomplete measure of hotel performance.
To analyze the change in market conditions, CBRE Hotels' Americas Research is developing a technique to produce forecasts of net operating income (NOI). This measure takes the sum of a hotel's revenue from all sources and subtracts from it all departmental expenses, property taxes, management fees, insurance expenses and capital reserves. Although NOI closely tracks RevPAR, it also allows expenses and non-room revenue to vary based on occupancy. This potentially gives a fuller impression of a hotel's performance during the recent period of persistently elevated occupancy.
To find out what the relationships between occupancy, expenses, and revenues are, we turn to the Trends® in the Hotel Industry data: approximately 65,000 hotel statements from 1994 to 2016. These statements are unweighted and so are, by themselves, not representative of all hotels in aggregate; however, the statements are from a broad geographic distribution and all levels of performance. We examine these properties' expenses and revenues to understand the effect of varying occupancy levels across markets. Real average daily rate (RADR) is included as a control variable with the assumption that ADR is correlated with unobserved factors like the quality of service or the cost-of-living immediately proximate to the hotel. Subsampling could be used to obtain market-specific effects or even to estimate chain-scale specific effects. Here we use all useable records from the entire sample, or 50,507 complete records, to obtain national results.
An important consideration when trying to determine the cause-and-effect relationship of occupancy on expenses and revenues is that while higher occupancy produces higher expenses as more rooms need to be cleaned and linens washed, higher expenses also may increase occupancy when they are used for marketing or enhancing the guest experience. We attempt to correct for this situation, called endogeneity, by using the 3-stage least squares technique on a system of equations. This widely-used technique in economic analyses produces consistent results in the presence of endogeneity, and in this case will provide a better estimate of the effect of occupancy on NOI holding other variables constant.
The model produced highly statistically significant results, and three-quarters of the variation in rooms expenses were explained by the variables in our model. Other expenses achieved a fit ranging from 60%-96% of the variation explained, although no other expense category is as generally variable as rooms expense. Considering the parsimony of our model's specification, this is good evidence that our model is doing its job of relating occupancy to expenses.
Trends® has an impressive, large sample of individual hotels. To get broadly representative NOI estimates, national-level STR data for supply, demand, and room revenue are added to the analysis and fitted to the relationships estimated from Trends®. This method then is applied to forecast RevPAR data from Hotel Horizons® to obtain forecast NOI. The assumption underlying this formulation is that while hotels that submitted statements to the Trends® sample are not representative of all the hotels in a market, the effect of occupancy and ADR on revenues is consistent within a sample. The result is a single time series of nationally representative annual data of observed RevPAR and estimated NOI from 1991 to 2016 and forecast RevPAR and NOI through 2020.