Abstract
We study the effect of differences in product quality on new technology diffusion. We propose a model in which heterogeneity in perceived product quality affects consumer adoption. If consumers experientially infer the quality of a technology, an increase in initial exposure to a low-quality product may inhibit subsequent diffusion. Incentives intended to speed up adoption may in fact have the opposite effect, if they propagate low-quality signals. We examine the predictions of the model using sales data for 11 hybrid-vehicle models between 2000 and 2006. Consistent with press reports that the first-generation Insight was perceived to be of lower quality than the first-generation Prius, we find that, conditional on overall hybrid vehicle adoption in the first 2 years, locations with a relatively high Prius market share experienced faster subsequent adoption than states with a relatively high Insight market share. We estimate the elasticity of new hybrid sales with respect to the Prius penetration rate is 0.30–0.58, while the elasticity with respect to the Insight penetration rate is \(-\)0.14 to \(-\)0.44.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Quotations from Sandahl et al. (2006), p. 6.5, 6.7, and 7.1.
For instance, state-level hybrid car incentives often have applied only to some hybrid models (see Gallagher and Muehlegger 2011, Table 2). A longer literature (e.g., Cohen 1991; Fri 2003; Helm 2010) studies the mixed history and political economy of government incentives for technological innovation and adoption.
A substantially redesigned Insight was reintroduced to the American market beginning in the 2010 model year, after our data set concludes.
A causal installed-base effect may occur for reasons other than consumer/social learning. For instance, “green envy” may create a similar diffusion pattern (Sexton and Sexton 2012), as may network effects or economies of scale in maintenance or repairs. While we are not able to identify or rule out any particular mechanism for generating the causal installed-base effect, we argue that the heterogeneity in diffusion that we see by models is most consistent with our model of consumer learning.
Young (2009) models diffusion and learning with heterogeneity among consumers, not technologies.
This paucity is also partly explained by the lack of significant data on this new technology. The Consumer Expenditure Survey, for example, contains data on vehicle ownership, but it only first asked respondents the fuel type of the vehicle (gasoline, diesel, or hybrid) in 2005. The 2006 data set only contains 119 observations of hybrid vehicles, out of more than 56,000 automobile observations.
Other papers that study diffusion in the automobile market without explicitly modeling firm behavior include Kloess and Muller (2011), Kopecy and Suen (2010), Lescaroux and Rech (2008), and Greenman (1996). Exclusion of firm behavior may bias coefficients up. However, our main results lie in the difference between the coefficient on the Prius and those on the Insight. It is unlikely that the bias from omitting firm behavior will affect this difference.
Consumers are risk-neutral and making decisions based only on the mean assessment of quality. An extension to this model could also consider risk-averse consumers, and the fact that a higher market share of hybrids affects both the mean and the uncertainty of consumers’ assessments.
For instance, the market share of hybrids in our simulations is higher than their market share in the US vehicle fleet, in the long run. The observed market share in the fleet as of our data set is so low that the qualitative effects are difficult to see in these simulations.
In this first case, the true qualities of the three models are \(\eta _A =10, \eta _B=9\), and \(\eta _C =8\). The priors for each of these three models are accurate. Initially, the non-hybrid model A has 100 % market share \((\hbox {S}_{\mathrm{A},1}= 1)\). The inference parameter \(\gamma = 0.9\). In each period, 2.5 % of consumers purchase a new vehicle. We simulate for 200 periods, where one period represents a quarter.
In this case, all of the parameters are the same as in the prior footnote, except that the priors for the hybrid vehicles are lower: \( \hat{\eta } _B \left( {{\Omega }_0 } \right) =7\) and \(\hat{\eta }_C \left( {{\Omega }_0} \right) =6\).
The models in the data are from Ford (Escape), Honda (Accord, Civic, Insight), Lexus (GS450h, RX400h), Mercury (Mariner), Saturn (VUE), and Toyota (Camry, Highlander, Prius).
Note that we need not worry about hybrid sales from before the start of our data set, since none of the models were introduced to the US market before 2000 Q1. The only exception to this is the Honda Insight, which was introduced in December 1999, so we are only missing that one month’s worth of sales.
As we discuss below, we classify states as Insight-intensive (Prius-intensive) if the cumulative hybrid vehicle market share of the Honda Insight at the end of the 2001 was above (below) the nationwide median. Since the criteria for grouping for Insight-intensive states is independent of a state’s hybrid penetration quartile, the number of states comprising each of the line plots varies. For example, for the group of states in the first penetration quartile, only one has lower an Insight market share lower than the national median in Q4 2001. In contrast, eight of the states in the fourth quartile have Insight market shares lower than then national median in Q4 2001. The second and third quartiles are relatively balanced—similar numbers of states in each quartile have Insight market shares above and below the national median in Q4 2001.
We later also allow for a state-specific linear time trend.
These data are obtained from Polk.
Instead of the difference between per capital Honda and Toyota registrations as the instrument, we also tried using the ratio, and using both Honda and Toyota registrations. Results are robust to any of these specifications.
We only use pre-hybrid registration data for one year, 1999. Therefore, interacting these state-level registrations with time indicators is necessary to avoid perfect multicollinearity in the first stage.
Mean fuel economy (combined) for Toyota and Honda 1999-year models was 24.6 and 24.7 mpg respectively. In contrast, average fuel economy across all makes and models was 21.6 mpg and average fuel economy for models of Chrysler, Ford and GM were 21.5, 19.1 and 16.9, respectively.
First-stage IV results are presented in Appendix Table 8. The coefficients on the Honda/Toyota registration differential instruments (interacted with time indicators) are negative in both the Prius and the Insight first stage regressions, indicating that states with relatively more Honda than Toyota registrations in 1999 had lower adoption of both types of hybrids. However, the magnitudes of the negative coefficients are greater on the Prius regressions than on the Insight regressions, suggesting that our instruments are correctly predicting relative Prius and Insight market shares. The coefficients on the assembly plant location indicators are generally insignificant in the Insight regression, but in the Prius regression, the Toyota plant indicators have positive coefficients and the Honda plant indicators have negative coefficients, as we would expect. F-tests on the joint significance of the instruments in the first stage are significant at the 1 % level for both Prius and Insight penetration rates, and they are well above the conventional bound of 10 indicating weak instruments. The instruments pass a Sargan test of overidentifying restriction (\(p\)-value \(=\) .7088) when state and time fixed effects are included.
In early model years, the Prius’s design looked very similar to its non-hybrid equivalent; only in later model years (2003 and later) did the Prius have a distinctive design. The Insight, by contrast, was distinctive from the start. Sexton and Sexton (2012) argue that the Prius’s unique design offered its owners a status bonus relative to the Civic hybrid; their data are from more recent model years in which the Prius was redesigned. In the last 2 years of our dataset, Toyota’s dominance of the hybrid market accelerated (most of the sales in the “other” category in Fig. 6 were Toyotas), and thus one concern may be that if the results in Table 3 are coming primarily from these last 2 years, then the IV assumptions are not met. To explore this we replicate the regressions from Table 3 excluding the last 2 years of data. These regressions are reported in Appendix Table 9, and they demonstrate that the regressions results are robust to dropping these years.
See http://www.theautochannel.com/news/press/date/20000720/press021003.html; downloaded August 2014.
These regression results do not include state-fixed effects. When including those in the regressions, the U-shape for the OLS coefficients is even more pronounced, while the Coefficients in the IV regressions are monotone decreasing.
Department of Energy, Alternative Fuels Data Center, accessed on April 8, 2014. http://www.afdc.energy.gov/uploads/data/data_source/10581/10581_cc_vehicle_history.xlsx.
In the regressions at the state-year level, the restriction assumption is that the pre-hybrid Honda and Toyota per capita registrations do not directly affect total hybrid sales. Here, in the regressions at the state-model-year level, the assumption would be that these pre-hybrid registrations do not directly affect the sales of any particular hybrid model. This assumption does not seem credible, since Toyota registrations should directly affect, say, Camry sales.
Because these regressions are at the state-quarter level, we cannot include a full set of state-by-quarter fixed effects. Including state-by-year fixed effects is possible, but due to limited degrees of freedom the resulting coefficient estimates are insignificant. The regressions at the state-quarter-model level (in Table 6 and column 1 of Table 7) include state-by-model fixed effects and model-by-quarter fixed effects.
Choi (1997) provides a theoretical model that includes both informational spillovers and network externalities.
As an example, the Energy Improvement and Extension Act, passed into law October 2008, provides tax credits for plug-in hybrid purchases up to $7,500. The model could also apply to other new types of automobiles besides hybrids, for instance crossovers or minivans.
References
Axsen J, Mountain DC, Jaccard M (2009) Combining stated and revealed choice research to simulate the neighbor effect: the case of hybrid-electric vehicles. Resour Energy Econ 31(3):221–238
Berestanu A, Li S (2011) Gasoline prices, government support, and the demand for hybrid vehicles in the United States. Int Econ Rev 52(1):161–182
Bollinger B, Gillingham K (2012) Peer effects in the diffusion of solar photovoltaic panels. Market Sci 31(6):900–912
Chandra A, Gulati S, Kandlikar M (2010) Green drivers or free riders? An analysis of tax rebates for hybrid vehicles. J Environ Econ Manag 60(2):78–93
Chevalier J, Mayzlin D (2006) The effect of word of mouth on sales: online book reviews. J Mark Res 43(3):345–354
Choi J (1997) Herd behavior, the ’Penguin Effect’, and the suppression of informational diffusion: an analysis of informational externalities and payoff interdependency. RAND J Econ 28(3):407–425
Cohen L (1991) The technology pork barrel. Brookings Institutional Press, Washington, DC
Colombo M, Mosconi R (March 1995) Complementarity and cumulative learning effects in the early diffusion of multiple technologies. J Ind Econ 43(1):13–48
Consumer Reports. “Fuel Efficient Cars 12/00”. December 2000. www.consumerreports.org
De Haan P, Mueller M, Peters A (2006) Does the hybrid Toyota Prius lead to rebound effects? Analysis of size and number of cars previously owned by Swiss prius buyers”. Ecol Econ 58(3):592–605
Dover Y, Goldenberg J, Shapira D (2012) Network traces on penetration: uncovering degree distribution from adoption data. Mark Sci 31(4):689–712
Fri R (2003) The role of knowledge: technological innovation in the energy system. Energy J 24(4):51–74
Gallagher K, Muehlegger E (2011) Giving green to get green? Incentives and consumer adoption of hybrid vehicle technology. J Environ Econ Manag 61(1):1–15
Greenman JV (1996) The car park: diffusion models revisited. Energy Econ 18(1–2):107–128
Griliches Z (1957) Hybrid corn: an exploration in the economics of technological change. Econometrica 25(4):501–522
Helm D (2010) Government failure, rent-seeking and capture: the design of climate change policy. Oxf Rev Econ Policy 26(2):182–196
Jensen R (1983) Innovation adoption and diffusion when there are competing innovations. J Econ Theory 29(1):161–171
Kahn M (2007) Do greens drive hummers of hybrids? Environmental ideology as a determinant of consumer choice. J Environ Econ Manag 54(2):129–145
Kahn M, Vaughn R (2009) Green market geography: the spatial clustering of hybrid vehicles and LEED registered buildings. BEJEAP Contrib 9(2):1–24
Kloess M, Muller A (2011) Simulating the impact of policy, energy prices and technological progress on the passenger car fleet in Austria—a model based analysis 2010–2050. Energy Policy 39(9):5045–5062
Kopecy K, Suen R (2010) A quantitative analysis of suburbanization and the diffusion of the automobile. Int Econ Rev 51(4):1003–1037
Lamberson PJ (2008) The diffusion of hybrid electric vehicles. In: Future research directions in sustainable mobility and accessibility, report by sustainable mobility accessibility research and transformation at the University of Michigan Center for Advancing Research and Solutions for Society
Lescaroux F, Rech O (2008) The impact of automobile diffusion on the income elasticity of motor fuel demand. Energy J 29(1):41–60
Mau P, Eyzaguirre J, Jaccard M, Collins-Dodd C, Tiedmann K (2008) The ‘Neighbor Effect’: simulating dynamics in consumer preferences for new vehicle technologies. Ecol Econ 68(1–2):504–516
McFarland J, Herzog H (2006) Incorporating carbon capture and storage technologies in integrated assessment models. Energy Econ 28(5–6):632–652
Medlock K, Soligo R (2002) Car ownership and economic development with forecasts to the year 2015. J Transp Econ Policy 36(2):163–188
Narayanan S, Nair HS (2013) Estimating causal installed-base effects: a bias-correction approach. J Mark Res 50(1):70–94
Phil P (2007) Once frumpy, green cars start showing some flash. New York Times, New York
Rose N, Joskow P (1990) The diffusion of new technologies: evidence from the electric utility industry. RAND J Econ 21(3):354–373
Sallee J (2011) The surprising incidence of tax credits for the Toyota Prius. Am Econ J Econ Policy 3(2):189–219
Sandahl LJ, Gilbride TL, Ledbetter MR, Steward HE, Calwell C (2006) Compact fluorescent lighting in America: lessons learned on the way to market. Pacific Northwest National Laboratory Report for the Department of Energy
Sexton S, Sexton A (2012) Conspicuous conservation: The Prius halo and willingness to pay for environmental bona fides. Working Paper
Stoneman P, Toivanen O (1997) The diffusion of multiple technologies: an empirical study. Econ Innov New Technol 5(1):1–17
Young HP (2009) Innovation diffusion in heterogeneous populations: contagion, social influence, and social learning. Am Econ Rev 99(4):1899–1924
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Heutel, G., Muehlegger, E. Consumer Learning and Hybrid Vehicle Adoption. Environ Resource Econ 62, 125–161 (2015). https://doi.org/10.1007/s10640-014-9819-3
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10640-014-9819-3