2020 promises to be the most challenging year on record for the U.S. lodging industry. While the U.S. economy appears to be at a standstill at the moment, it is likely to limp forward in fits and starts for the remainder of 2020 and faces a long, hard road to recovery. Lodging industry fundamentals are likely to remain weak through at least 2020, or until employment growth gathers steam in a broad based recovery. Unfortunately, there are still many unknowns in store for the U.S. lodging industry and it is clear to those of us who track its movements that it will be a slow and bumpy ride through 2020 and into 2021.
One of the biggest question marks hanging over the industry’s future relates to the nature of structural changes brought on by the health crisis and employment meltdown. To what extent has there been a structural shift in the historical relationship between underlying economic drivers of lodging demand and the demand for hotel accommodation throughout the country? Only time will tell if the industry faces a permanent paradigm shift. For now, we are left with the challenge of forecasting the likely performance of the lodging industry with our long held assumptions in place!
Given the wide range of economic forecasts, it is not unusual to observe a wide range of RevPAR forecasts provided by lodging analysts. Making detailed projections about the timing and nature of the recovery of the US lodging industry might seem like a fool’s errand at this stage of the COVID-19 pandemic. Despite the unprecedented levels of uncertainty and the serious limits of predictability, a number of 2020 US RevPAR forecasts have been provided by industry analysis. They include CBRE Research (-36.9%), STR/Tourism Economics (-50.6%), Morgan Stanley (-32%) and Baird Equity Research (-60%).
These point-estimate forecasts are the most probable numbers, not the only possible numbers. Because they are single numbers, point estimates are almost always above or below the parameters they are supposed to estimate. Without additional information, point estimates are less than perfect for making hotel investment and operating decisions! Point forecasts provide no indication of the uncertainty in the number, and uncertainty is an important consideration in decision making. For example, a RevPAR forecast of $100+/- $10 may lead to a much different investment decision than a forecast of $100+/- $50.
An interval estimate or confidence interval on the other hand, is a range of values that contains the true parameter value with a known probability. The probability that the interval contains the true value, is the confidence level.
A confidence interval is therefore a standard way of describing the precision of a measurement of some parameter, such as a RevPAR forecast. There are two common ways of stating confidence intervals both of which provide exactly the same information, for example: 1) Total U.S. RevPAR is forecast to grow by 5% plus or minus 2% with 90% confidence. 2) Total U.S. RevPAR is forecast to grow by between 3% and 7%, with 90% confidence.
There are three pieces of information in a confidence interval: the lower bound, the upper bound, and the confidence level. In the above example, the lower bound is 3%, the upper bound is 7%, and the confidence level is 90%. Since properly applied confidence intervals incorporate tested assumptions, these are reliable tools to make hotel investment and operating decisions.
The confidence interval is calculated using statistical techniques such as bootstrapping that insure that the confidence level is objective. (See details on the “bootstrapping” technique below). Confidence intervals computed using a smaller confidence level will be smaller than those computed with a larger confidence level; it is not possible to directly compare two confidence intervals made at differing confidence levels.
Bootstrapping is a statistical method that can be used to estimate statistical properties from a finite data set (in this case, annual RevPAR growth rates over the period 1988-2019).The technique involves sampling a large number (we use 5,000) of data sets from our historical data sets of these variables. Given a data set with m data points (32 years), a bootstrap data set is a data set of m points chosen at random with replacement from the original data set as illustrated in Exhibits 1 & 2 below. The statistical properties for which we have used bootstrapping include RevPAR growth rates, confidence intervals and associated coefficients and R2s. The advantage of bootstrapping over other analytical methods is its great simplicity – it is straightforward to apply the bootstrap to derive estimates of confidence intervals.
What is the best confidence level? Most confidence intervals use a confidence level of 90 percent or 95 percent, but these levels are not right for every situation. To pick the most appropriate confidence level, we need to think about the investment decision to be made, and the potential effects of making a bad decision. When using a confidence interval to make a decision with two choices, there is always a decision rule. If the decision value is inside the interval, one choice will be made, but if the decision value is outside the interval, the opposite choice will be made. When thoughtfully applied, confidence intervals and confidence bounds are powerful tools for making hotel investment decisions.
As an example of the bootstrapping technique, we have provided RevPAR forecasts and associated confidence intervals for the Anaheim/Santa Ana MSA hotel market using STR/Tourism Economics RevPAR forecast for the entire U.S. lodging industry of -50.6% for 2020, and 63.2% for 2021 and the relationship between Total U.S. RevPAR growth and the RevPAR growth for the Anaheim/Santa MSA hotel market.
Exhibit 1: Simple Least Squares Regression Uncertainty Surrounding Anaheim/Santa Ana MSA RevPAR % Growth by Using the Bootstrap Technique
Exhibit 2: Simple Least Squares Regression Uncertainty Surrounding Anaheim/Santa Ana MSA RevPAR % Growth by Using the Bootstrap Technique
The width or tightness of the RevPAR growth confidence intervals differ markedly across different markets reflecting the underlying volatility of fundamentals such as supply and demand and their relationship with economic drivers such as total U.S. employment.
Exhibit 3: Simulated Distribution of 2020 Anaheim/Santa Ana MSA 2020 RevPAR Growth
Based on the US RevPAR forecast of STR/Tourism Economics for 2020 and 2021 and the historical relationship between RevPAR growth for the Total U.S and the Anaheim/Santa Ana MSA hotel market, we forecast a RevPAR decline for the Anaheim/Santa Ana MSA hotel market of 55.3% in 2020 and a growth of 70.9% in 2021. There is a 90% probability that the market’s RevPAR growth will lie between –49.7% and -64.1% and a 60% probability that it will fall between –53.3% and -56.5% in 2020 as illustrated in Exhibit 3 and the table below.
Exhibit 4: Simulated Distribution of Forecasted RevPAR & RevPAR Growth % for 2020 & 2021 Anaheim/Santa Ana MSA Hotel Market
The average beta for the market is 1.1. This means that on average a 1% increase in Total U.S. RevPAR translates into a 1.1% increase in RevPAR in the Anaheim/Santa Ana MSA hotel markets. Conversely a I% decline in US RevPAR translates into a 1.1% decline in RevPAR for the Anaheim/Santa Ana MSA hotel market. There is a 90% probability that beta falls between 1.0 and 1.2 as illustrated in Exhibit 5.
Exhibit 5: Simulated Distribution of Beta for Anaheim/Santa Ana MSA 2020 RevPAR Growth
The R2 or coefficient of determination for Anaheim/Santa Ana’s RevPAR growth rate is 0.94 and is negatively skewed as illustrated in Exhibit 6. There is a 90% probability that between 79% and 99.5% of the variation in the market’s RevPAR growth rate is explained by the variation in the Total U.S. RevPAR growth rate. Now that you know more about confidence intervals, don’t hesitate to use them. Remember, a point estimate by itself is just another useless number. Have confidence in your analysis and your investment decisions!
Exhibit 6: Simulated Distribution of R-Squared for Anaheim/Santa Ana MSA 2020 RevPAR Growth
Based on our bootstrapping techniques we are able to provide clients with forecasts for chain scales, hotel locations, regions and about 60 hotel markets throughout the country.