Economics Group Special Commentary Anika R. Khan, Senior Economist anika.khan@wellsfargo.com (704) 410-3271 Azhar Iqbal, Econometrician azhar.iqbal@wellsfargo.com (704) 410-3270 Is the Hotel Cycle Near the End? Real GDP Growth May Not be the Best Indicator to Tell The hotel sector has outperformed in recent years and has seen strong gains across all key performance indicators. Real revenue per available room (RevPAR), which is the product of occupancy and average daily rate (ADR), rose 6.2 percent in 2015, marking the sixth straight year of positive gains. 1 Moreover, the occupancy rate hit its highest level on record last year, finishing the year at 65.5 percent, which helped to push real ADR up to its strongest pre-recession pace. Strength in the hotel sector can be traced back to solid U.S. domestic demand, an increase in international travel in large gateway markets, and a low level of new supply. Although real RevPAR is still improving, the pace began slowing noticeably in Q4 and continued to moderate in Q1. Moreover, the rate of real RevPAR growth for lower-end hotels fell into negative territory in Q1. 2 With the recent slowdown in real RevPAR growth and the lodging cycle in its sixth year of expansion, we explore whether broad measures of U.S. economic activity, including real GDP, real final sales, the broad trade-weighted dollar index (TWD), employment and the Leading Economic Index (LEI) are statistically useful in predicting real RevPAR growth. We also estimate the effect to real RevPAR growth given a shock to these macroeconomic variables. Using this framework, we find LEI and real final sales to have the strongest predictive power. Our analysis suggests it would take a 3 percent decline in LEI to pull real RevPAR into negative territory; and, real final sales would require a 4 percent drop. With these broad measures of economic activity still posting positive gains, we expect real RevPAR growth to continue to grow in the coming quarters driven by real ADR growth. That said, real ADR growth is also slowing. Figure 1 Figure 2 1 Real RevPAR vs. Real ADR Seasonally Adjusted, Year-over-Year Percent Change 1 9% Occupancy Rate Seasonally Adjusted, Year-over-Year Percent Change 9% The hotel sector has outperformed in recent years and has seen strong gains across all key performance indicators. 6% 6% 3% 3% -3% -3% -1-1 -6% -6% -9% -9% Real RevPAR: Q1 @ 1.4% Real ADR: Q1 @ 2.2% -2-2 88 90 92 94 96 98 00 02 04 06 08 10 12 14 16-12% -12% Hotel Occupancy: Q1 @ -0.8% 88 90 92 94 96 98 00 02 04 06 08 10 12 14 16 Source: Smith Travel Research and Wells Fargo Securities, LLC 1 We seasonally adjust real RevPAR growth by using the X-12 ARIMA Seasonal-Adjustment Program in Eviews 7. 2 We use Smith Travel Research s Chain Scale classification and define mid- and economy-scale as lower-end. This report is available on wellsfargo.com/economics and on Bloomberg WFRE.
Growth in hotel demand is closely tied to the U.S. economic business cycle, with occupancy, real ADR and supply following a cyclical pattern. Late-Cycle Indicators Emerge in the Lodging Sector 3 Growth in hotel demand is closely tied to the U.S. economic business cycle, with occupancy, real ADR and supply following a cyclical pattern. Since 1990, we have had two lodging cycles that have exhibited fairly predictable behavior: First, occupancy increases as economic activity and labor market conditions advance, increasing to levels above the long-run average as demand exceeds supply. Second, occupancy reaches a cycle peak, and real ADR growth continues to expand, contributing the lion s share to real RevPAR growth as hotel operators adjust pricing. Responding to the early surge in demand, hotel construction also ramps up. Third, the high levels of ADR eventually cause the occupancy rate to moderate, while new construction continues to grow. Finally, the increase in the stock of hotel rooms pull occupancy and real ADR growth lower, and real RevPAR growth slows. The current hotel cycle is not any different. Over the past six or so years, the hotel sector has outperformed most asset classes, with hotel demand growth skyrocketing, occupancy reaching an all-time high and real RevPAR growth advancing between 3.3 percent and 8.0 percent per annum since 2010. 4 In fact, the top of the range for real RevPAR growth during this cycle, at 8.0 percent, is also a record high. However, the overall pace of real RevPAR growth is moderating with Q1 posting the lowest year-over-year reading since early 2010, slowing from 4.4 percent in Q4 quarter to 1.4 percent in Q1, which is much slower than expected. Previous lodging cycles show that even when occupancy peaks, real ADR growth can be the dominant driver of activity boosting overall real RevPAR growth for several years. During the past lodging cycle, which spans more than seven years (trough to trough), the occupancy rate peaked in early 2006, with real RevPAR growth having continued to advance almost two years after the zenith on the back of ADR (Figure 3). Last year, ADR accounted for almost three-fourths of real RevPAR growth and is expected to be the sole driver in the coming years. One study found a weak and lagged statistical relationship between ADR and occupancy as hotel operators appear reluctant to adjust rental rates quickly in response to short-run changes in occupancy caused by the economic business cycle. 5 Moreover, the three-month annualized rate of change has been running above the 12-month change in recent months, corroborating the expectation that real RevPAR growth has a little more room to improve. That said, the pace of real ADR growth is slowing, with most pointing to the strong dollar as the key driver behind the moderation. Empirically, we find a statistically significant relationship between growth in the broad tradeweighted dollar and real ADR growth, but the variance in ADR can barely be explained by the dollar (Figure 4). Figure 3 Figure 4 68% 66% 64% Occupancy Rate Cycles Seasonlly Adjusted, Percent Peak: 1994Q4 @ 65.2% Peak: 2006Q1 @ 64.1% Peak?: 2015Q4 @ 65.9% 68% 66% 64% 6% 4% 2% Real ADR vs. Trade-Weighted Dollar Year-over-Year Percent Change 62% 62% 6 Trough: 1991Q1 @ 60.9% 6-2% 58% 58% -4% 56% Trough: 2001Q4 @ 57. 56% -6% 54% Trough: 2009Q2 @ 54.1% 52% 5 87 89 91 93 95 97 99 01 03 05 07 09 11 13 15 54% 52% 5-8% -1 y = 0.0704x + 0.3472 R² = 0.0279-12% -2-1 1 1 2 2 Source: Smith Travel Research, Federal Reserve Board, and Wells Fargo Securities, LLC 3 Corgel, J.B., Predictive Powers of Hotel Cycles, Real Estate Issues: Winter 2003-2004. 4 We deflate RevPAR by using the quarterly CPI Index, which is indexed to 2014. 5 Wheaton, William C. and Rossoff, Lawrence, The Cyclic Behavior of the U.S. Lodging Industry, Real Estate Economics: Volume 26, Issue 1, March 1998. 2
Another indicator worth watching is real RevPAR growth for lower-end hotels (mid- and economy-scale), which posted a negative reading in Q1. On the back of weak real ADR growth for lower-end hotels and declining occupancy (Figure 5), real RevPAR growth fell 1.8 percent in Q1 marking the slowest pace in more than four years. During Q1, the real room rate fell sharply, increasing only 0.6 percent following a 2.2 percent increase in Q4. On the other hand, high-end real RevPAR growth rose 2.6 percent in Q1, after increasing only 0.7 percent in Q4 (Figure 6). These results are not surprising given the path of the U.S. economy during the current expansion, where much of the economic activity has been concentrated at the top of the income distribution. However, the sustained improvement in labor market conditions and wage and salary growth for low- and middle-income households suggests there is still room for improvement in the coming quarters. Given these late-cycle signs in the hotel sector, we take an empirical approach to determine the effect of a shock to broad measures of real economic activity to real RevPAR growth. Figure 5 Figure 6 8 7 Occupancy Rate by Chain Scale Seasonally Adjusted, Percent High-End Scale: Q1 @ 73.8% Lower-End Scale: Q1 @ 57.3% 8 7 1 1 RevPAR Growth by Chain Scale Seasonally Adjusted, Year-over-Year Percent Change 1 1 7 7 6 6-1 -1 6 6 5 5 04 05 06 07 08 09 10 11 12 13 14 15 16 Source: Smith Travel Research and Wells Fargo Securities, LLC 5 5-2 High-End Scale: Q1 @ 2.6% -2 Lower-End Scale: Q1 @ -1.8% -3-3 05 06 07 08 09 10 11 12 13 14 15 16 Empirical Results: Does Macroeconomics Matter for RevPAR? Figure 7, shows the statistical relationship between the year-over-year change in real GDP and real RevPAR growth. Although there is a correlation, we would like to determine the effect to real RevPAR growth given a shock to economic activity measured by real GDP, and other macroeconomic variables including real final sales, employment growth, TWD and LEI. 6 To make this assessment, we first test whether these variables are statistically useful in predicting real RevPAR growth. Next, we explore which variables have the highest predictive power as we can include only so many variables in a traditional econometric model. Last, we estimate the effect of a change in an economic variable on real RevPAR growth. We utilize the Granger causality test to determine whether these potential predictors are statistically useful in explaining real RevPAR growth. We believe the Granger causality test is a better method as it seeks to reveal any statistical causality ( cause and effect ) between variables. 7 Results based on the Granger causality analysis are reported in Table 1, and we find all predictors, with the exception of TWD, are statistically useful. Based on this empirical evidence, we find that macroeconomic variables matter for real RevPAR growth. In fact, the LEI is the single-best predictor of real RevPAR growth using the Chi-Square test value, with the smallest error, on average. We also find that the second-best predictor is real -2-2 We utilize the Granger causality test to determine whether these potential predictors are statistically useful in explaining real RevPAR growth. 6 Note, we are using growth rates (year-over-year percent change) for all variables. Typically, growth rates of a macroeconomics series tend to be stationary (as it is a differenced form) compared to level form of a macroeconomics variable. Stationarity of the dataset is necessary to obtain reliable statistical results and non-stationary dataset, usually, produces spurious statistical results. 7 For more details about the Granger causality and Chi-Square see, Silvia, John, Iqbal, A et al (2014). Economic and Business Forecasting: Analyzing and Interpreting Econometrics Results. Wiley 2014. 3
final sales followed by real GDP and employment. The dollar index is the only statistically insignificant variable based on the results from the Granger Causality Test (Table 1). 8 Figure 7 Table 1 9 1-1 Real RevPAR vs. Real GDP Year-over-Year Percent Change -2 y = 2.0517x - 4.7114 R² = 0.4549-2 -6% -4% -2% 2% 4% 6% Testing the Causal Relationship: The Granger Causality Test Dependent Variable= Real RevPAR Regressor Outcome LEI 41.69* Real Final Sales 18.00* Real GDP 15.31* Employment Growth 11.07* Trade Weighted Dollar (TWD) 0.93 Our statistical analysis shows that a one percentage point drop in the LEI growth rate produces the largest effect on real RevPAR growth. Source: Smith Travel Research, U.S. Department of Commerce, U.S. Department of Labor, Federal Reserve Board, the Conference Board and Wells Fargo Securities, LLC Once the Granger causality test results confirmed the statistical usefulness of these predictors, we then build a vector autoregression (VAR) model to estimate the likely effect of a change in the predictor (LEI for example) on real RevPAR growth. The VAR model allows us to shock the predictor and then estimate the effect of a one percentage point decline in the predictor on real RevPAR growth. Based on the results, our final VAR model consists of real RevPAR (our target variable), real GDP, real final sales and LEI. 10 The LEI Deserves Attention from the Hotel Industry Our statistical analysis shows that a one percentage point drop in the LEI growth rate produces the largest effect on real RevPAR growth, suggesting LEI is a leading and statistically meaningful indicator for the hotel sector (Figure 8). We find that real RevPAR dropped by 0.47 percentage point in Q1, with the largest drop (in response to a one percentage point drop in the LEI), estimated to occur in Q4, at -1.57 percent. Indeed, it would take a 3 percent decline in LEI to pull real RevPAR growth into negative territory. Given that the LEI model produces the largest Chi-Square test statistics based on the Granger causality test and largest effect on real RevPAR growth using the VAR model, our analysis suggests that LEI should be considered in forecasting hotel activity. Real final sales also has strong predicative power in explaining real RevPAR growth. Figure 9 shows that a one percentage point drop in the real final sales growth rate shows a 0.38 percentage point reduction in real RevPAR growth in Q1, with the largest decline noted in the fifth quarter (-1.13). Here, we find it would require a 4 percent drop in real final sales to see a decline in real RevPAR growth. Surprisingly, the effect of a drop in real GDP on real RevPAR does not yield an intuitive result after the first quarter. In response to a one percentage drop in real GDP, real RevPAR growth shows a decline of 0.14 percentage point in Q1, but increases in every quarter thereafter, which is contrary to the expected result. In this case, we put more weight on LEI and real final sales to estimate real RevPAR growth. 8 One potential reason why TWD shows a statistically insignificant relationship to real RevPAR growth is that our econometric method estimates a historical relationship on average between dollar and RevPAR and historically the dollar s effect on RevPAR is not statistically meaningful. 9 We ranked Regressors using the Chi-Square Test Statistics, where a higher Chi-Square value may indicate that the corresponding variable is more useful than others. For example, the LEI model produces the highest Chi-Square values and hence best predictor among competitors. All variables, except the Dollar Index, are statistically significant at 1 percent. 10 We also tested employment growth along with LEI, GDP and Final Sales, but the effect of an employment decrease was not meaningful on real RevPAR growth. Therefore, we did not include employment growth in our final model 4
Figure 8 Figure 9 1.0 0.5 Effects of a LEI "Shock" On Real Hotel RevPAR Growth, Percentage Points Real Hotel RevPAR Two Standard Deviation Upper-Limit Two Standard Deviation Lower-Limit 1.0 0.5 1.0 0.5 Effects of a Real Final Sales "Shock" On Real Hotel RevPAR Growth, Percentage Points Real Hotel RevPAR Two Standard Deviation Upper-Limit Two Standard Deviation Lower-Limit 1.0 0.5 0.0 0.0 0.0 0.0-0.5-0.5-0.5-0.5-1.0-1.0-1.0-1.0-1.5-1.5-1.5-1.5-2.0 1 2 3 4 5 6 7 8 9 10 11 12-2.0-2.0 1 2 3 4 5 6 7 8 9 10 11 12-2.0 Source: U.S. Department of Commerce, The Conference Board and Wells Fargo Securities, LLC How Likely Are We to See Consecutive Declines in LEI? Now that we have established LEI as the strongest predictor of real RevPAR growth, we take a closer look at the trend in the indicator. The Conference Board s LEI is made up of 10 components, which are thought to be forward-looking. With the recent increase in the equity market, the index rose 2.2 percent year over year in March (Figure 10). However, the weak factory sector, slowing building permits and consumer expectations have weighed down the headline reading over the past six months (Figure 11). That said, with headwinds from the manufacturing sector abating, the overall index should strengthen further in the coming months. Given the empirical results that it would take a 3 percent decline in LEI to pull real RevPAR growth into negative territory, we explore the likelihood of the index falling one percent in three consecutive quarters. We find that as a leading indicator for the economic business cycle, LEI typically falls ahead of the downturn, and consecutive quarterly declines of one percent or more are common throughout the time series. For example, the National Bureau of Economic Research, the official arbiters of recession, marked Q4 of 2007 as the start of the recession. LEI posted its first negative quarterly reading five quarters before the start of the recession, registering straight declines of one percent or more in the three quarters before the start of the downturn. The lesson here is that once LEI begins to show back-to-back declines it does not only spell trouble for the cyclical hotel sector, but signals weakness for the overall economy. Figure 10 Figure 11 1 Leading Economic Index Year-over-Year Percent Change 1 Net Contribution to Leading Economic Index Past Six Months, Percentage Points Interest Rate Spread 1.21 We find that as a leading indicator for the economic business cycle, LEI typically falls ahead of the downturn. 1 1 Leading Economic Index 0.65 Leading Credit Index Stock Prices 0.22 0.16 Consumer Goods 0.08 Initial Claims 0.04-1 -1 Manuf. Hours Worked Consumer Expectations -0.04 0.00 Building Permits -0.05-2 LEI Year/Year Change: Mar @ 2.2% -2 04 05 06 07 08 09 10 11 12 13 14 15 16-2 -2 N.D. Cap. Goods Ex-Air -0.10 ISM New Orders -0.46 March 2016-1.0-0.5 0.0 0.5 1.0 1.5 2.0 Source: The Conference Board and Wells Fargo Securities, LLC 5
Conclusion The hotel sector has outperformed most asset classes during the current expansion, with performance metrics hitting record levels. However, the overall pace of activity as measured by real RevPAR growth is moderating with Q1 posting the lowest year-over-year reading since early 2010. With occupancy likely reaching a peak last year and ADR growth moderating, it is understandable that investors are asking how much longer the current hotel cycle will last. Since 1990, the hotel sector has seen two cycles, which followed fairly predictable behavior. Once occupancy peaks, ADR growth does the heavy lifting for overall real RevPAR growth, with the sector continuing to advance. During the previous cycle (trough to trough: Q4 2001 Q2 2009), the occupancy rate peaked in early 2006, with real RevPAR growth having continued to advance almost two years after the zenith on the back of ADR. We expect a similar result during the current cycle; however, the pace of real ADR growth is slowing much faster than expected. Some have pointed to the strong dollar to explain the faster deceleration in ADR growth; however, empirical evidence suggests that the variance in ADR growth can barely be explained by the dollar. That said, forecasting ADR growth is not within the scope of this paper, but would be a worthwhile exercise to explore in a follow-up report. Another tell-tale sign the hotel sector is nearing the end of the cycle is the slowdown in lower-end real RevPAR growth. Given these late-cycle indicators, we explore whether broad measures of U.S. economic activity including real GDP, real final sales, employment, the TWD and LEI are statistically useful in predicting real RevPAR growth. We also estimate the effect of real RevPAR growth given a shock to these variables. Using this framework, we find LEI and real final sales to have the strongest predictive power. Our analysis suggests it would take a 3 percent decline in LEI to pull real RevPAR into negative territory, and real final sales would require a 4 percent drop. With these broad measures of economic activity still posting positive gains, we expect real RevPAR growth to continue to grow in the coming quarters driven by real ADR growth, but the pace of the moderation in real ADR growth bears watching. ** Special thanks to Wells Fargo Equity Research and Wells Fargo CMBS Research for providing added insight to this report. 6
Appendix I The VAR Approach Sims (1980) introduced the vector autoregression (VAR) modeling approach as an alternative to the large scale structural model, also known as macro-econometric model. 11 The basic idea behind a VAR approach is that instead of including hundreds of variables in a model, we can include a handful of variables (sometimes eight variables) to represent major sectors of an economy and then that model can be utilized for forecasting and policy analysis (see Sims [1980] for more details). A traditional VAR of n-variables will consist of n-equations, one equation for each variable. Each equation includes a constant and lag(s) of n-variables, including lag(s) of the left-hand-side variable. The lag order, how many lags of a variable, is denoted by P. Therefore, a VAR (P) of n- variables indicates up to p-lags of each variable are utilized in each equation. Here we share a simple example of a two-variable VAR model, which includes one lag of each variable, we will call it VAR(1), because P=1 in this case. Yt = α0 + α1 Yt-1 + α2 Xt-1 + ε1t Xt= β0 + β1 Yt-1 + β2 Xt-1 + ε2t Impulse Response Function In an n-variables VAR model, an impulse response function (IRF) shows the effect of a one percentage point increase in one variable in the current period on all variables in the model. The basic idea behind an IRF is that you increase one percentage point (or one unit) of one of the VAR variables in the current period, let us say GDP growth, assuming that the increase will disappear in the subsequent periods. Furthermore, we keep the VAR errors for other (n-1) variables equal to zero, that is actual values are equal to estimated values. That would allow us to generate the effect of an increase in one variable on all others variables. In addition, we can approximate the total effect of a change in the GDP on the other variable of interest (RevPAR for example), where the impact may be distributed over a prolonged period of time. 11 See Christopher Sims, Macroeconomics and Reality, Econometrica 48 (1980), p.1-48. 7
Wells Fargo Securities, LLC Economics Group Diane Schumaker-Krieg Global Head of Research, Economics & Strategy (704) 410-1801 (212) 214-5070 diane.schumaker@wellsfargo.com John E. Silvia, Ph.D. Chief Economist (704) 410-3275 john.silvia@wellsfargo.com Mark Vitner Senior Economist (704) 410-3277 mark.vitner@wellsfargo.com Jay H. Bryson, Ph.D. Global Economist (704) 410-3274 jay.bryson@wellsfargo.com Sam Bullard Senior Economist (704) 410-3280 sam.bullard@wellsfargo.com Nick Bennenbroek Currency Strategist (212) 214-5636 nicholas.bennenbroek@wellsfargo.com Anika R. Khan Senior Economist (704) 410-3271 anika.khan@wellsfargo.com Eugenio J. Alemán, Ph.D. Senior Economist (704) 410-3273 eugenio.j.aleman@wellsfargo.com Azhar Iqbal Econometrician (704) 410-3270 azhar.iqbal@wellsfargo.com Tim Quinlan Senior Economist (704) 410-3283 tim.quinlan@wellsfargo.com Eric Viloria, CFA Currency Strategist (212) 214-5637 eric.viloria@wellsfargo.com Sarah House Economist (704) 410-3282 sarah.house@wellsfargo.com Michael A. Brown Economist (704) 410-3278 michael.a.brown@wellsfargo.com Jamie Feik Economist (704) 410-3291 jamie.feik@wellsfargo.com Erik Nelson Economic Analyst (704) 410-3267 erik.f.nelson@wellsfargo.com Alex Moehring Economic Analyst (704) 410-3247 alex.v.moehring@wellsfargo.com Misa Batcheller Economic Analyst (704) 410-3060 misa.n.batcheller@wellsfargo.com Michael Pugliese Economic Analyst (704) 410-3156 michael.d.pugliese@wellsfargo.com Julianne Causey Economic Analyst (704) 410-3281 julianne.causey@wellsfargo.com Donna LaFleur Executive Assistant (704) 410-3279 donna.lafleur@wellsfargo.com Dawne Howes Administrative Assistant (704) 410-3272 dawne.howes@wellsfargo.com Wells Fargo Securities Economics Group publications are produced by Wells Fargo Securities, LLC, a U.S. broker-dealer registered with the U.S. Securities and Exchange Commission, the Financial Industry Regulatory Authority, and the Securities Investor Protection Corp. Wells Fargo Securities, LLC, distributes these publications directly and through subsidiaries including, but not limited to, Wells Fargo & Company, Wells Fargo Bank N.A., Wells Fargo Advisors, LLC, Wells Fargo Securities International Limited, Wells Fargo Securities Asia Limited and Wells Fargo Securities (Japan) Co. Limited. Wells Fargo Securities, LLC. is registered with the Commodities Futures Trading Commission as a futures commission merchant and is a member in good standing of the National Futures Association. Wells Fargo Bank, N.A. is registered with the Commodities Futures Trading Commission as a swap dealer and is a member in good standing of the National Futures Association. Wells Fargo Securities, LLC. and Wells Fargo Bank, N.A. are generally engaged in the trading of futures and derivative products, any of which may be discussed within this publication. Wells Fargo Securities, LLC does not compensate its research analysts based on specific investment banking transactions. Wells Fargo Securities, LLC s research analysts receive compensation that is based upon and impacted by the overall profitability and revenue of the firm which includes, but is not limited to investment banking revenue. The information and opinions herein are for general information use only. Wells Fargo Securities, LLC does not guarantee their accuracy or completeness, nor does Wells Fargo Securities, LLC assume any liability for any loss that may result from the reliance by any person upon any such information or opinions. Such information and opinions are subject to change without notice, are for general information only and are not intended as an offer or solicitation with respect to the purchase or sales of any security or as personalized investment advice. Wells Fargo Securities, LLC is a separate legal entity and distinct from affiliated banks and is a wholly owned subsidiary of Wells Fargo & Company 2016 Wells Fargo Securities, LLC. Important Information for Non-U.S. Recipients For recipients in the EEA, this report is distributed by Wells Fargo Securities International Limited ("WFSIL"). WFSIL is a U.K. incorporated investment firm authorized and regulated by the Financial Conduct Authority. The content of this report has been approved by WFSIL a regulated person under the Act. For purposes of the U.K. Financial Conduct Authority s rules, this report constitutes impartial investment research. WFSIL does not deal with retail clients as defined in the Markets in Financial Instruments Directive 2007. The FCA rules made under the Financial Services and Markets Act 2000 for the protection of retail clients will therefore not apply, nor will the Financial Services Compensation Scheme be available. This report is not intended for, and should not be relied upon by, retail clients. This document and any other materials accompanying this document (collectively, the "Materials") are provided for general informational purposes only. SECURITIES: NOT FDIC-INSURED/NOT BANK-GUARANTEED/MAY LOSE VALUE