Theoretical Inflation for Unavailable Products by Rachel Soloveichik ESCoE Virtual Conference, September 16 th to 18 th Disclaimer: The views in this presentation reflect those of the author and not necessarily those of the Department of Commerce or the Bureau of Economic Analysis.
Preview of Theoretical Strategy • Hotel prices are higher in dense urban regions – Despite the higher prices, tourists still flock to cities with desirable amenities that aren’t available in rural regions • This observed behavior can be used to estimate theoretical inflation rates for unavailable products 2 Louisiana Vacations: New Orleans vs. Rural Region
Preview of Empirical Results • Time spent at retail and recreational locations is a proxy for product availability – Product availability started dropping in March, bottomed out in April and slowly recovered in May and June • Theoretical inflation was at least 1.2 percentage points above the CPI in Q1 and 4.9 percentage points in Q2 – Published economic statistics miss approximately one third of the theoretical drop in real consumption • Two data appendixes provide with detailed data on specific products and specific regions 3
Outline of Presentation • Review of price measurement literature • Development of new model to estimate theoretical prices for unavailable products • Empirical data on the exact unavailable products – Unavailable products include both products restricted under a government stay-in-place order and products that consumers voluntarily avoid • Regional estimates of product unavailability in the first and second quarter of 2020 – This unavailability is then used to calculate regional inflation rates 4
Theoretical Price Measurement Problem • Laspeyres price index formula: – Price Index T = w 10 (p 1T /p 10 )+ w 20 (p 2T /p 20 )+…+ w n0 (p nT /p n0 ) – This formula requires prices for every product • The CPI assumes that unavailable products without price data have similar inflation as comparable products with price data – This assumption appears to be accurate in normal economic times (Bradley 2003) – However, the assumption might not apply when stay-in-place policies suddenly make broad product categories unavailable 5
Relevant Price Index Literature • “New goods”: – (Hausman 1999), (Hausman 1997), (Petrin 2002), (Goolsbee and Petrin 2004), (Berndt et al. 1996), (Nordhaus 1996), (Diewert and Feenstra 2019), and (Diewert et al. 2019) • “Outlet substitution bias”: – (Reinsdorf 1993), (Hausman and Liebtag 2009), and (Greenlees and Mclelland 2008) • “Variety bias”: – (Feenstra 1994), (Broda and Weinstein 2010), and (Handbury and Weinstein 2014) • None of these literatures match COVID-19 6
New Price Measurement Model • Tourists choose a rural or urban vacation – Six products: Housing, goods 1&2, services 1&2, and amenity – The amenity is only available in urban regions • Assumption: rational tourists visit the region where a vacation budget buys the most utility – Weather and travel costs are similar in the two regions – Theoretical rural prices equal theoretical urban prices, so that – The price premium for the unavailable rural amenity is ip aR = [(1–w h p hR - (w g1 p g1R +w g2 p g2R )- (w s1 p s1R +w s2 p s2R )]/w a 8
Theoretical Prices for Unavailable Products • Assumption: the price premium for unavailable tourist amenities is smaller than the price premium for nonessential goods and services – Tourist amenities are generally considered more discretionary than nonessential products like clothing or elective surgery – Tourists are better able to plan around unavailable products • Comparing price indexes: Theoretical Prices ≥w h p hSIP +w g1 p g1SIP + w s1 p s1SIP +(w g2 + w s2 + w a )ip aR Quasi-BLS Prices =[w h p hSIP +(w g1 + w g2 ) p g1SIP + (w s1 + w s2 )p g1SIP ]/(1-w a ) 9
Data Used to Measure Theoretical Inflation • ip aR are calculated using these data sources: – BEA’s regional income account gives p hR , (w g1 p g1R +w g2 p g2R )/(w g1 +w g2 ), and (w s1 p s1R +w s2 p s2R )/(w s1 +w s2 ) – BEA’s travel and tourism account gives w h , w g1 +w g2 , w s1 +w s2 , and w a – The paper calculates ip aR = 1.59 in the average region • Theoretical prices are calculated using this data: – BLS reports low monthly inflation for p hSIP , p g1SIP , and p g1SIP – Appendix A reports product unavailability in a full stay-in-place policy: w h =0.20, w g1 =0.27 w g2 =0.05, w s1 =0.31, w s2 =0.17, w a =0.02 – Appendix B reports actual product unavailability for every region 10
Theoretical Inflation in Q1 and Q2 of 2020 11
Data on Actual Product Unavailability • Google’s Mobility Trends gives mobility changes – Most unavailable products are sold at retail and recreational locations, so this paper focuses on that category – Dataset is publicly available at the day/county level • American Time Use Survey (ATUS) gives normal mobility levels – The paper calculates mobility levels for each region and day type – Mobility for smaller regions is smoothed to reduce volatility • Weather Underground gives daily weather – Holding local policy fixed, mobility is higher in pleasant weather 12
ATUS Mobility for 2003-2018 Minutes Per Day at Retail and Recreational Locations 13 9/9/2020
Average Mobility for March 2020 Minutes Per Day at Retail and Recreational Locations 14 9/9/2020
Average Mobility for April 2020 Minutes Per Day at Retail and Recreational Locations 15 9/9/2020
Average Mobility for May 2020 Minutes Per Day at Retail and Recreational Locations 16 9/9/2020
Average Mobility for June 2020 Minutes Per Day at Retail and Recreational Locations 17 9/9/2020
Impact of Temperature on Mobility 18
Impact of Humidity on Mobility 19
March Adjustment for Weather and Day Minutes Per Day at Retail and Recreational Locations 20 9/9/2020
April Adjustment for Weather and Day Minutes Per Day at Retail and Recreational Locations 21 9/9/2020
May Adjustment for Weather and Day Minutes Per Day at Retail and Recreational Locations 22 9/9/2020
June Adjustment for Weather and Day Minutes Per Day at Retail and Recreational Locations 23 9/9/2020
Summary of Actual Mobility Changes • ATUS respondents normally spend 75 minutes per day at retail and recreational locations – About 75 percent of this time is spent on nonessential activities • Almost every region sees a mobility drop in 2020 – Adjusted time fell 13 minutes per day in March, 33 minutes per day in April, 23 minutes per day in May, and 15 minutes per day in June – Theoretical inflation was at least 1.2 percentage points above the CPI in Q1 and 4.9 percentage points above the CPI in Q2 • Wealthy urban regions saw larger mobility drops 24
Theoretical Inflation vs. Income in 2018 Bubble size is proportional to regional population 25 9/9/2020
Conclusion • Inflation is underestimated when many common goods and services are unavailable – The paper then develops a new model to estimate theoretical cost-of-living when products are unavailable • The paper collected detailed data on both potential and actual product unavailability – Theoretical inflation was at least 1.2 percentage points above the CPI in Q1 and 4.9 percentage points above the CPI in Q2 – The published economic statistics miss approximately one third of the theoretical drop in real consumption 26
Recommend
More recommend
