Introduction Data Results Conclusions HOV Stickers and the Consumer Adoption of Hybrids: Evidence from California Kenneth Gillingham and Calanit Kamala Yale University and UC Berkeley November 7, 2012 USAEE Conference Introduction Data Results High Occupancy Vehicle (HOV) Lanes HOV lanes were originally intended to promote carpooling: Conclusions Introduction Data Results Conclusions HOV Lanes on Many Congested Highways HOV lanes must be appealing to many commuters: Introduction Data Results Clean Air Access Stickers for Hybrids Some states co-opt HOV lanes by allowing hybrids on them: Conclusions Introduction Data Results Conclusions Hybrid Policies in the Literature Hybrids have been a hot topic recently: Non-incentive factors influencing adoption: Kahn (2007), Heutel & Muehlegger (2011) Incidence of hybrid tax credits: Sallee (2011), Beresteanu & Li (2011) Effect of federal or state incentives for hybrids: Diamond (2009), Chandra, Gulati & Kandilkar (2010), Gallagher & Muehlegger (2011), Beresteanu & Li (2011) Market willingness-to-pay for a sticker: Shewmake & Jarvis (2011) Effect of hybrid sticker program on travel speeds: Bento et al. (2011), Jang & Cassidy (2011) Introduction Data Results Research Questions This paper focuses on California’s CAVS program: 1 What was the program’s effect on hybrid sales? How many of the stickers were given to free-riders? 2 What was the program’s effect on hybrid prices? What is the incidence of the policy? Conclusions Introduction Data Results California CAVS program dates Conclusions Introduction Data Results Data Sources All new vehicle registrations in CA, 2001-2009 (R.L. Polk). VIN, make, model, series, subseries, date of purchase, list price, transaction price, zip code of purchaser. All 85,000 CAVS stickers, 2005-2007 (CA DMV) VIN, zip code of purchaser, date of sticker. Monthly retail gasoline prices in each county in CA (OPIS). Zip-code demographics (2000 Census). County-level monthly unemployment (BLS) Monthly Consumer Confidence Index (Conference Board) County-level monthly median house prices (CA Realtor’s Association) Conclusions Introduction Data Results Conclusions Summary statistics Observation is a zip code-year-month Variable number vehicles sold hybrids sold eligible hybrids sold stickered vehicles sold CAVS time period transaction price (2010$) baseline list price (2010$) county unemployment rate avg house price (000s 2010$) gasoline price (2010$) percent green party percent yes on Prop 84 percent yes on Prop 87 county avg commute time zip code density zip code median income zip code percent 65+ zip code percent 18- Mean 73.5 1.75 1.13 0.38 0.28 27346.8 27402.03 6.77 447.07 2.68 0 0.51 0.44 25.91 3.66 61138.64 12.33 23.98 s.d. 248.43 5 3.08 1.44 0.45 6972.98 7242.03 2.76 207.66 0.64 0 0.13 0.14 4.62 5.64 29074.57 6.93 7.08 Min 1 0 0 0 0 10000 10030 2.8 94.44 1.25 0 0.1 0 13.4 0 0 0 0 Max 24488 542 541 133 1 100000 109899 31.3 1195.37 5 0.04 1 0.85 43.1 52.18 375000 100 41.3 N 210,531 210,531 210,531 210,531 210,531 199,843 204,746 209,733 209,733 209,606 162,330 162,244 162,244 207,045 161,743 161,138 161,138 161,138 Introduction Data Results Conclusions 0 new vehicles sold 2,000 4,000 6,000 Vehicle Sales 2000m1 2002m1 2004m1 2006m1 month of registration eligible hybrids 2008m1 sticker recipients 2010m1 Introduction Data Results Conclusions 0 percent of vehicles sold .02 .04 .06 Percent Hybrid Sales 2000m1 2002m1 2004m1 2006m1 month of registration eligible hybrids 2008m1 sticker recipients 2010m1 Introduction Data Results Conclusions 0 percent of vehicles sold are eligible hybrids .02 .04 .06 Sales by Proximity to HOV Lanes 2000m1 2002m1 2004m1 2006m1 month of registration HOV counties 2008m1 non−HOV counties 2010m1 Introduction Data Results Conclusions 20,000 avg vehicle price (2010$) 25,000 30,000 35,000 Hybrid Prices 2000m1 2002m1 2004m1 2006m1 month of registration all new vehicles 2008m1 new eligible hybrids 2010m1 Introduction Data Results Conclusions 22,000 eligible hybrid price (2010$) 24,000 26,000 28,000 30,000 Hybrid Prices by Proximity to HOV Lanes 2000m1 2002m1 2004m1 2006m1 month of registration HOV counties 2008m1 non−HOV counties 2010m1 Introduction Data Results Conclusions Primary Empirical Specifications qzt = β0 + β1 CAVSzt ∗ HOVcountyzt + βXzt + µm + εzt where for zip code z and month t: qzt is the number/percent of eligible or stickered hybrids sold CAVSzt is a dummy for 1/2005 - 2/2007 HOVcountyzt is a dummy for the county containing an HOV lane Xzt is a vector of controls (e.g., gas prices, unemployment, demographics, quadratic time trend) µm are month-of-the-year m fixed effects Also run a specification with hybrid prices as the dependent variable Introduction Data Results Conclusions Identification Identification of β1 is based on both time series and cross-sectional variation E[CAVSzt ∗ HOVcountyzt ∗ εzt ] = 0 relies on no county-specific shocks during the time period of the program that lead to more hybrid purchases. Introduction Data Results Conclusions Effect of Program on Eligible Hybrid Sales Dependent variable: eligible hybrids sold in zip code (mean = 1.1 per month) CAVS*HOVcounty county unempl rate avg house price gasoline price (1) base (2) demog (3) mon-of-yr dummies (4) time trend (5) county FE 0.238*** (0.050) -0.015* (0.006) 0.003*** (0.000) 0.578*** (0.023) -1.560*** (0.109) 0.658*** (0.065) -0.027*** (0.007) 0.001** (0.000) 0.980*** (0.040) 2.921*** (0.745) -0.015* (0.007) 0.036*** (0.011) -3.297*** (0.383) 0.521*** (0.057) -0.056*** (0.010) 0.001 (0.000) 0.586*** (0.038) 2.842*** (0.739) -0.016* (0.007) 0.037*** (0.011) -2.441*** (0.389) 0.446*** (0.056) -0.069*** (0.011) 0.000 (0.000) 0.531*** (0.035) 2.870*** (0.739) -0.016* (0.007) 0.038*** (0.010) -97.695*** (9.822) 0.108* (0.048) -0.145*** (0.011) 0.004*** (0.000) 0.364*** (0.022) -0.434 (0.940) -0.049** (0.016) 0.017 (0.013) -34.801*** (5.479) N N N N Y N N N Y Y Y N Y Y Y N Y Y Y Y 0.072 209,606 0.167 155,372 0.171 155,372 0.175 155,372 0.193 155,372 % yes on Prop 84 county avg commute zip code density constant zip demographics month-of-year FE quadrat. time trend county FE R-squared Observations *** indicates significant at 1% level, ** significant at 5% level standard errors clustered on zip code in parentheses Introduction Data Results Conclusions Effect of Proximity to HOV Lanes on Stickers Dependent variable: sticker hybrids sold in zip code (mean = 0.38 per month) HOVcounty county unempl rate avg house price gasoline price (1) base (2) demog (3) mon-of-yr dummies (4) time trend (5) county FE 0.893*** (0.067) 0.033*** (0.010) 0.002*** (0.000) 0.239*** (0.025) -1.507*** (0.172) 1.061*** (0.110) 0.045*** (0.012) -0.000 (0.000) 0.521*** (0.041) 0.288 (0.613) -0.008 (0.007) 0.037*** (0.010) -2.767*** (0.411) 1.061*** (0.111) 0.045*** (0.012) -0.000 (0.000) 0.518*** (0.038) 0.286 (0.614) -0.008 (0.007) 0.037*** (0.010) -2.762*** (0.411) 1.059*** (0.111) 0.040*** (0.012) -0.000 (0.000) 0.349*** (0.036) 0.233 (0.613) -0.009 (0.007) 0.037*** (0.010) -1104.955*** (76.422) 1.306** (0.447) 0.020*** (0.005) 0.001* (0.000) 0.395*** (0.034) -1.348 (0.882) -0.020 (0.016) 0.014 (0.012) -1003.460*** (71.028) N N N N Y N N N Y Y Y N Y Y Y N Y Y Y Y 0.092 59,835 0.242 41,806 0.242 41,806 0.247 41,806 0.274 41,806 % yes on Prop 84 county avg commute zip code density constant zip demographics month-of-year FE quadrat. time trend county FE R-squared Observations *** indicates significant at 1% level, ** significant at 5% level standard errors clustered on zip code in parentheses Introduction Data Results Conclusions Effect of Program on Eligible Hybrid Prices Dependent variable: transaction price (mean = 27,347 2010$) CAVS*HOVcoun*eligible county unempl rate avg house price gasoline price (1) base (2) demog (3) mon-of-yr dummies (4) time trend (5) county FE 214.017*** (34.171) -22.521 (16.768) 5.348*** (0.381) -489.454*** (35.244) 26331.269*** (225.431) 137.483*** (23.157) -29.082 (18.343) -0.181 (0.601) -134.153* (55.423) 3677.352* (1528.822) -54.902*** (12.383) 32.503 (17.174) 22306.374*** (892.706) 89.976*** (23.602) -45.408 (27.360) -1.937* (0.796) -624.163*** (72.220) 3534.778* (1507.048) -53.570*** (12.256) 39.229* (17.347) 22447.063*** (919.890) 67.094** (24.933) -40.947 (28.888) -2.343** (0.850) -702.222*** (67.168) 3552.002* (1500.791) -52.367*** (12.245) 41.539* (17.418) -1.86e+05*** (26800.418) 38.321 (22.881) -203.324*** (13.409) 1.873*** (0.285) -1040.751*** (33.372) 4269.535* (2012.853) -32.577 (20.322) -22.617 (29.487) -7.26e+04*** (10652.060) N N N N Y N N N Y Y Y N Y Y Y N Y Y Y Y 0.032 199,542 0.224 151,989 0.236 151,989 0.239 151,989 0.280 151,989 % yes on Prop 84 county avg commute zip code density constant zip demographics month-of-year FE quadrat. time trend county FE R-squared Observations *** indicates significant at 1% level, ** significant at 5% level standard errors clustered on zip code in parentheses Introduction Data Results How Many Free-riders? We can get a first-order approximation on free-ridership: All stickers from hybrids purchased before 9/2004 must be free-riding: 30% Stickers from hybrids purchased in non-HOV counties may be free-riding: 7% Using estimate of induced sales in HOV counties: Even if there was a 40% increase in sales due to the policy, free-riding: 38% These imply free-riding on the order of 75%! Conclusions Introduction Data Results Conclusions and Future Work The CAVS program did have an effect on sales of hybrids By increasing demand, CAVS also increased prices of hybrids However, a very high level of free-ridership Future work: Estimate the incidence of the policy Examine the interaction with federal tax incentives Conclusions