1# RECORD c98a43acf9eb55645cbcf1d65650ddce2b8ab2c62# TITLE: The social value of a ridesharing platform: a hedonic pricing approach3# YEAR: 2022.0 VENUE: Empirical Economics DOI: 10.1007/s00181-022-02310-94# AUTHORS: Yongwook Paik; C. Makridis5# PAGES: 26 CANDIDATE TABLE PAGES (1-based): [11, 16, 17, 18, 19]6# ABSTRACT: Governments around the world have been facing the challenges of regulating the ever-increasing prevalence of ridesharing platforms, leading to a proliferation of academic and policy debates about the social value of such platforms. However, the overall social value is still unclear because when analyzing the net benefit provided by ridesharing platforms, prior studies have focused on either supply side (i.e., drivers) or demand side (i.e., passengers) users without considering nonusers. Conversely, some studies that have considered the spillover effect of a ridesharing platform on nonusers have not simultaneously considered the net benefit for users. In this study, we propose that a hedonic pricing model can be used as a reasonably straightforward approach to collectively quantify the social value of a ridesharing platform for both users and nonusers. Using this approach, we exploit plausibly exogenous variation in the staggered entry of Uber into different metropolitan areas between 2010 and 2016 and find that the entry of Uber leads to, on average, a 2.8% increase in median housing prices per square foot within a metropolitan area. We also find that there is substantial spatial heterogeneity, with benefits concentrated in areas with greater dependence on public transportation and higher levels of traffic delay. By quantitatively computing the social value of a ridesharing platform with a hedonic pricing model, our study helps government regulators generate policies that are more likely to promote public welfare and become less susceptible to political whims.789===== PAGE 1 (score 7, numeric-paren 0, table-words 0, var-words 7) =====1011Empirical Economics (2023) 64:2125–215012https://doi.org/10.1007/s00181-022-02310-913The social value of a ridesharing platform: a hedonic pricing14approach15Yongwook Paik116· Christos A. Makridis217Received: 15 June 2021 / Accepted: 13 September 2022 / Published online: 25 September 202218© The Author(s) 202219Abstract20Governments around the world have been facing the challenges of regulating the ever-21increasing prevalence of ridesharing platforms, leading to a proliferation of academic22and policy debates about the social value of such platforms. However, the overall social23value is still unclear because when analyzing the net benefit provided by ridesharing24platforms, prior studies have focused on either supply side (i.e., drivers) or demand25side (i.e., passengers) users without considering nonusers. Conversely, some studies26that have considered the spillover effect of a ridesharing platform on nonusers have27not simultaneously considered the net benefit for users. In this study, we propose that a28hedonic pricing model can be used as a reasonably straightforward approach to collec-29tively quantify the social value of a ridesharing platform for both users and nonusers.30Using this approach, we exploit plausibly exogenous variation in the staggered entry31of Uber into different metropolitan areas between 2010 and 2016 and find that the entry32of Uber leads to, on average, a 2.8% increase in median housing prices per square foot33within a metropolitan area. We also find that there is substantial spatial heterogeneity,34with benefits concentrated in areas with greater dependence on public transportation35and higher levels of traffic delay. By quantitatively computing the social value of a36ridesharing platform with a hedonic pricing model, our study helps government reg-37ulators generate policies that are more likely to promote public welfare and become38less susceptible to political whims.39Keywords Platform · Hedonic pricing · Social value · Ridesharing · Uber40Yongwook Paik and Christos A. Makridis have equal contribution and are listed in random order.41B Yongwook Paik42ywpaik@kaist.ac.kr43Christos A. Makridis44cmakridi@stanford.edu45146KAIST College of Business, Korea Advanced Institute of Science and Technology, 85 Hoegiro,47Dongdaemoon-gu, Seoul 02455, Korea48249Stanford University, 210 Panama Street, Stanford, CA 94305, USA50123515253===== PAGE 11 (score 22, numeric-paren 0, table-words 6, var-words 8) =====5455The social value of a ridesharing platform: a hedonic pricing approach56213557Table 1 Characterizing the quasi-random nature of Uber entries into metropolitan areas. Sources: Census58Bureau, Zillow, Quarterly Census of Employment and Wages, 2010–2016. The table reports the coefficients59associated with logit regressions of Uber’s entry (at a metro × month frequency) on housing price per square60foot, employment growth, and wage growth conditional on year and month fixed effects and a range of61metro × year demographics, including logged population, the share of males, the age distribution (0–17,6218–34, 35–64), the race distribution (the share of African Americans), the share of married individuals,63and the education distribution (no high school, high school, some college, college). The table shows that64contemporaneous and one-year lagged housing prices, employment, and wage growth do not predict Uber’s65entry into a metro area. Standard errors are clustered at the metro level and are shown using brackets66Dependent variable =67Uber entry into area68(1)69(2)70Housing price growtht71−2.8672−6.1173[2.15]74[4.74]75Employment growtht762.50773.3678[4.08]79[3.72]80Wage growtht810.21820.3383[1.89]84[1.55]85Housing price growtht-1861.9587[3.08]88Employment growtht-189−0.5790[1.97]91Wage growtht-192−0.1493[1.05]94R-squared950.76960.7697Sample size989347999232100Metro controls101Yes102Yes103Year + month FE104Yes105Yes1064 Results1074.1 Net effect of Uber on local property values108Table 2 presents our main results associated with Eq. (1) when our outcome variable is109logged median housing prices per square foot at the zip code level. Not surprisingly, the110cross-sectional estimator suggests that there is a strong positive correlation between the111areas Uber enters and housing prices when we do not control for confounding factors.112For example, Column (1) shows that the property values tend to be 11.0% higher in113places where Uber entered. However, once we add control variables for demographic114conditions to model (2), the conditional correlation declines to an implied 5.4% rise115in property values following the entry of Uber. However, these correlations may still116be potentially biased by time-invariant factors across locations or by seasonal factors117because fixed effects are not added.118123119120121===== PAGE 12 (score 36, numeric-paren 0, table-words 7, var-words 21) =====1221232136124Y. Paik, C. A. Makridis125Table 2 Estimating the effects of uber entry on housing values. Sources: Census Bureau, Zillow, 2010-2016.126The table reports the coefficients associated with regressions of logged median housing prices per square127foot on an indicator for the month × year that Uber entered a metropolitan area, conditional on metropolitan128controls, which include logged population, logged(traffic delay indicator), logged(air quality indicator), the129share of males, the age distribution (0–17, 18–34, 35–64), the race distribution (white, black), the share of130married individuals, and the education distribution (no high school, high school, some college, college).131Standard errors are clustered at the metro level and are shown using brackets132Dependent variable =133ln(median housing price/sq ft)134(1)135(2)136(3)1371[Uber entry]1380.110∗∗∗1390.054∗1400.028∗∗141[0.020]142[0.028]143[0.010]144R-squared1450.021460.301470.98148Sample size149398,209150225,176151224,824152Metro controls153No154Yes155Yes156Zip code FE157No158No159Yes160Year + quarter FE161No162No163Yes164*** p value < 0.01, ** p value < 0.05, * p value < 0.10165Consistent with this view, the inclusion of zip code and time fixed effects reduces166the marginal effect of Uber’s entry to 0.028, as shown in model (3). Nonetheless, the167estimated effect of Uber’s entry on the local housing market remains positive and statis-168tically significant, implying that there is a 2.8% net gain in median property values, on169average, after Uber’s entry. To the extent that we interpret Column (3) as our preferred170causal estimate of Uber’s entry, approximately 25.5% (= 0.028/0.110) of the overall171association between entry and property values is driven by the causal impact that Uber172has on local amenities—that is, after we control for selection effects.15 The R-squared173in model (3) is quite large (e.g., 0.98) relative to models (1) and (2), reflecting the fact174that much of the variation in median zip code-level housing prices is explained by175time-invariant fixed factors within a zip code (e.g., neighborhood conditions, location176of house, size of house, number of bedrooms and bathrooms, parking space) and hous-177ing market seasonality trends. It also implies that our hedonic pricing model does not178suffer materially from potentially omitted variables, as the variation in housing prices179is already sufficiently accounted for in our preferred specification (R-squared = 0.98)180(Oster 2019). Given that the mean of the median housing price per square foot across181all zip codes in our sample is $98.62/sq ft, Column (3) implies that the marginal effect182of Uber’s entry would increase the median housing price per square foot to $101.38/sq183ft. This increase would imply, for example, that the value of a 2,000-square-foot home184would undergo a gradual increase of approximately $5,520 after Uber becomes avail-185able in the local market. Thus, our result is not only statistically significant but also186economically significant and realistic.18715 We find it useful to decompose selection and causal effects in our setting to inform policy debates, as is188common in many other settings (e.g., in labor economics; see Lazear (2000)). Understanding the magnitude189of the causal effect, above and beyond the selection effect, provides insight into other factors that might190explain Uber’s entry and the adoption of ridesharing services.191123192193194===== PAGE 16 (score 18, numeric-paren 0, table-words 3, var-words 8) =====1951962140197Y. Paik, C. A. Makridis198Table 4 Heterogeneous treatment effects of Uber entry on housing values. Sources: Census Bureau, Zillow, 2010–2016. The table reports the coefficients associated with199regressions of the logged median housing price per square foot on an indicator for the month × year that Uber entered a metropolitan area and interaction terms testing200various heterogeneous effects of Uber’s entry. Metro controls include logged population, logged income, the share of males, the age distribution (0–17, 18–34, 35–64), the201race distribution (white, black), the share of married individuals, and the education distribution (no high school, high school, some college, college). Heterogeneity effects202are tested by comparing areas above versus below the median value of a key characteristic that represents various local conditions. The median values of these key variables,203all obtained from zip code-level census data, refer to the following: at least 26.4% of individuals with a college degree or more, at least 11.3% of individuals below the204poverty line, the share of workers over the age of 16 who use public transportation (including taxis) as a means of transportation to work equal to at least 0.8%, average time205to commute to work of at least 25 min, and vehicles per capita (16 + and older) equal to at least 0.52. Standard errors are clustered at the metro level and are shown using206brackets207Dep. var. =208ln(median housing price/sq ft)209(1)210(2)211(3)212(4)213(5)214(6)215(7)216(8)2171[Uber entry]2180.017∗2190.021∗∗∗2200.0042210.033∗∗∗2220.040∗∗∗223−0.0012240.023∗∗225−0.04226[0.009]227[0.007]228[0.007]229[0.008]230[0.009]231[0.009]232[0.009]233[0.17]2341[college > median]2350.312∗∗∗2360.274∗∗∗237[0.013]238[0.017]239× 1[Uber entry]2400.022∗∗2410.013242[0.008]243[0.010]2441[poverty > median]245−0.207∗∗∗246−0.111∗∗∗247[0.017]248[0.017]249× 1[Uber entry]2500.0152510.000252[0.006]253[0.006]2541[pct public transport > median]255−0.122∗∗∗256−0.046∗∗∗257[0.017]258[0.015]259× 1[Uber entry]2600.047∗∗∗2610.018∗∗262[0.009]263[0.008]2641[commute > median]265−0.021266−0.054∗∗∗267123268269270===== PAGE 17 (score 13, numeric-paren 0, table-words 3, var-words 3) =====271272The social value of a ridesharing platform: a hedonic pricing approach2732141274Table 4 (continued)275Dep. var. =276ln(median housing price/sq ft)277(1)278(2)279(3)280(4)281(5)282(6)283(7)284(8)285[0.022]286[0.021]287× 1[Uber entry]288−0.008289−0.013290[0.008]291[0.12]2921[vehicles per capita > median]2930.142∗∗∗2940.033∗295[0.019]296[0.019]297× 1[Uber entry]298−0.022∗∗299−0.007300[0.008]301[0.009]3021[traffic delay > median]303−0.042∗∗304−0.037∗∗305[0.019]306[0.018]307× 1[Uber entry]3080.056∗∗∗3090.052∗∗∗310[0.019]311[0.017]3121[air quality > median]3130.023∗∗∗3140.021∗∗315[0.007]316[0.008]317× 1[Uber entry]3180.0173190.000320[0.020]321[0.015]322R-squared3230.683240.633250.593260.583270.603280.583290.583300.70331Sample Size332397,601333397,601334397,601335397,464336397,464337224,856338397,601339224,784340Metro Controls341Yes342Yes343Yes344Yes345Yes346Yes347Yes348Yes349Metro FE350Yes351Yes352Yes353Yes354Yes355Yes356Yes357Yes358Year + Quarter FE359Yes360Yes361Yes362Yes363Yes364Yes365Yes366Yes367*** p value < 0.01, ** p value < 0.05, * p value < 0.10368123369370371===== PAGE 18 (score 10, numeric-paren 4, table-words 0, var-words 6) =====3723732142374Y. Paik, C. A. Makridis375Consequently, Uber is more appreciated and highly valued in markets with more376educated people.377Column (2) shows that areas where Uber enters have higher property values (β =3780.021***) regardless of the level of poverty. Unsurprisingly, areas with higher levels379of poverty are also negatively correlated (β = −0.207***) with property values in380that area. In addition, areas with higher levels of poverty potentially benefit more from381Uber’s entry and may gain a point estimate of 1.5% (β = 0.015) more in property values382than areas with lower levels of poverty. While this result may potentially suggest the383benefits offered by the gig economy platform, caution is needed, as the coefficient is384not statistically significant.385Column (3) shows that zip codes where people rely more on using public trans-386portation (compared to driving their own car) to commute to work are less desirable387residential areas in general (β = −0.122***) but may benefit more from Uber’s entry388than places where people do not need public transportation as much because they use389their own cars. In fact, the net gain that Uber offers to local markets is actually concen-390trated in zip codes where dependence on public transportation is relatively high, as the391own effect (1[Uber entry]), while still positive, becomes statistically insignificant in392Column (3). The interaction effect in Column (3) implies that there is an approximately3934.7%increaseinpropertyvaluesafterUber’sentryinareaswhereusageofpublictrans-394portation is high but not elsewhere. The estimate is in line with theory because it is395precisely these areas (i.e., where residents are less dependent on driving their own car)396that stand to gain the most by having a more robust and active ridesharing economy,397as more individuals value having access to a low-cost alternative transportation mode.398It is worth noting that the construction of our variable, 1[pct public transport], relies399on 2010 census data that combine taxi users and users of other public transportation400modes. Thus, it may be useful to consider the heterogeneous effect of Uber’s entry401based on residents’ average commute time to tease out some of the nuances and gain402more insight because as commute time increases, it probably becomes less likely that403people rely solely on taxis on a daily basis.404Column (4) shows that zip codes where people experience longer commute times405may be less desirable residential areas in general (β = −0.021, although not statisti-406cally significant) and that, consistent with previous results, the net gains Uber offers to407a local market are positive and statistically significant (β = 0.033***). In addition, the408interaction term (β = −0.008) in Column (4), while negative, shows that there is no409statistically significant difference between locations where people experience longer410versus shorter commute times. This nonsignificance most likely occurs because peo-411ple living in locations with longer commute times rely on their own cars or low-cost412transit rather than taxis. Thus, Uber (which functions similarly to local taxis) may not413provide much benefit to consumers with longer commute times, especially if they use414Uber sporadically rather than for commuting purposes on a daily basis. Therefore,415the results in Columns (3) and (4), taken together, imply that Uber is similar to a taxi416service that may be used for shorter trips (e.g., the “last mile” after getting off a train417station) or for occasional rides, rather than longer commutes, but potentially offers418greater convenience and cost-effectiveness.419123420421422===== PAGE 19 (score 11, numeric-paren 3, table-words 0, var-words 8) =====423424The social value of a ridesharing platform: a hedonic pricing approach4252143426In Column (5), the interaction term shows that the entry of Uber may actually be427undesirable in areas where the level of vehicle ownership is already high, as the coef-428ficient is negative and statistically significant (β = −0.022**). If vehicle ownership429is already high, then there may be less demand for ridesharing services or for other430public transportation options (e.g., the correlation between the share of the population431taking public transportation and the measure of vehicles per capita within a zip code432is −0.67). Nonetheless, these areas must still endure the negative externalities of433ridesharing services, such as increased congestion (e.g., frequent stops by rideshar-434ing drivers to load and unload passengers) or increased air pollution (Keating 2019).435Hence, areas with high vehicle ownership seem to appreciate the entry of Uber the436least because people rely on driving their own cars.437Column (6) shows that, unsurprisingly, zip codes that suffer more traffic delays are438less desirable residential areas in general (β = 0.042***) but may benefit more from439Uber’s entry than other places where traffic delays are not as bad because consumers440can benefit from the convenience of using a low-cost ridesharing service such as Uber441rather than stressfully driving to navigate local traffic congestion. The interaction442effect in Column (6) implies that there is an approximately 5.6% gain in property443value after Uber’s entry in areas where traffic delay is severe compared to areas where444traffic delay is not so severe. Similar to Column (3), the net gain Uber offers to local445markets is concentrated in zip codes where traffic delay is relatively high, as the own446effect (1[Uber entry]) becomes statistically insignificant in Column (6). The result in447column (6), as well as the result in column (3), shows that local traffic conditions are448an important factor that unleashes the potential benefits Uber offers.449Column (7) shows that areas with better air quality may be more desirable (β =4500.023***) to reside in general. However, the interaction term between air quality and451Uber’s entry is not significant (β = 0.017), which implies that the benefit of Uber’s452entry does not differ across areas depending on their air quality. Column (8) includes all453interaction terms simultaneously and finds that the inclusion of all interaction terms454does not materially alter our results and interpretations. Importantly, while we find455that the benefit of Uber’s entry had substantial spatial heterogeneity depending on456various local market conditions, we find that the benefits were concentrated in areas457with greater dependence on public transportation or with higher levels of traffic delay458according to column (8).459Overall, our quantitative results from the hedonic pricing model point to the follow-460ing conclusions: after Uber’s entry, there is a nonnegligible average economic gain in461property value that is equivalent to an increase of 2.8% in the median housing value462per square foot. While our hedonic pricing model suggests that Uber has a net positive463social value on average, there is substantial heterogeneity across regions based on local464conditions. Regions with greater dependence on public transportation (as opposed to465driving their own vehicle) and with greater traffic delay benefit the most after Uber’s466entry. These findings are corroborated in Column (8) when all interaction terms are467included simultaneously, as this column shows results that are qualitatively similar468to those in Columns (1)–(7). In summary, our main findings are consistent with the469123470471472===== PAGE 20 (score 6, numeric-paren 0, table-words 0, var-words 6) =====4734742144475Y. Paik, C. A. Makridis476view that Uber, and ridesharing more generally, functions as a relatively low-cost,477time-efficient for-hire transportation mode that directly competes with local taxis.174784.4 Aggregate social value of a ridesharing platform and policy implications479We now compute the aggregate welfare gains from having access to ridesharing. Under480the assumption that the MWTP for access to ridesharing is homogeneous, we can take481our preferred gradient of 0.028 and scale it by the number of individuals who had482access to ridesharing as of the end of 2016.18 We take the total population in these483metropolitan areas with exposure to ridesharing platforms such as Uber and scale it484by the number of individuals, which equals approximately 209,256,147 people. As485of 2016, among the set of metropolitan areas with an Uber presence, the population-486weighted average of the median housing price per square foot was $130.9. Given that487we find that Uber entry is associated with a 2.8% causal increase in housing prices,488the marginal increase amounts to $3.67 per square foot. Once we scale this figure489across all individuals exposed to ridesharing in these markets, we obtain a total of490$768 million of net value created to society. Thus, our results show that ridesharing491platforms represent a net gain for society.492The policy implications of these results deserve some attention. While our results493show the net gain to society represented by Uber’s entry and the heterogeneity of social494value accrued across regions, our method cannot speak to the heterogeneity of social495value accrued to each agent (e.g., taxi drivers, Uber drivers, renters, homeowners, or496nonusers of Uber) within a market.19 We only know that in aggregate, the benefit that497residents (renters and homeowners alike) enjoy from having access to Uber outweighs498the cost that they suffer from the disruption generated by Uber. The net benefit is499thus translated into a higher willingness to pay for being a resident. In this regard,500some policymakers may be responding to the harm done by Uber to specific groups501by banning Uber against a background of society-wide gains. For example, this was502the case in Austin, Texas, London, UK and other international markets (Lee 2017).503In addition, due to spatial heterogeneity, if regions with greater dependence on public504transportation or greater traffic delays increase property taxes (for example, up to 2.8%,505according to our estimates20), the additional tax revenue can be used, for example, to50617 Paik et al. (2019) reports that Uber offers a significant economic benefit in the form of a 45% discount,507on average, relative to a typical taxi fare for the same 5-mile distance due to ridesharing.50818 If the MWTP for access to ridesharing is not homogeneous, then the welfare gains may be lower than509our estimate. The reason is that if homebuyers who care about Uber are likely driving up the prices for510homes, then individuals other than homebuyers may have a lower MWTP. Our welfare estimates will be511the upper bound in this case.51219 The welfare implication of Uber for each agent is the subject of many prior microlevel studies (e.g.,513Barrios et al., 2020; Cachon et al., 2017; Chen et al., 2019; Greenwood & Wattal, 2017; Hagiu & Wright,5142019; (Park, Pang, Kim, and Lee 2021) https://pubsonline.informs.org/doi/abs/10.1287/isre.2020.0978).515However, the purpose of this study is to synthesize and compute the aggregate social value created by Uber516that accrues to both users and nonusers of Uber so that we can complement prior studies.51720 Given that platform businesses such as Uber intentionally incur accounting losses for many years while518offering promotions and discounts to users, the increase in property taxes should be much lower than the5192.8% we estimated because it is likely that Uber will eventually begin to increase prices once it starts to520enjoy a strong network effect (Katz & Shapiro, 1985; Parker & Van Alstyne, 2005; Zhu & Iansiti, 2012).521123522