Income targeting and surge pricing
-
Upload
gpano -
Category
Data & Analytics
-
view
293 -
download
1
Transcript of Income targeting and surge pricing
Income Targe,ng & Surge Pricing
Fish VP Analy,cs, Zenefits
(thanks to Uber data science) 11/18/15
Tradi,onal Economics
Behavioral Economics (Kahneman & Tversky, 1979)
Camerer et al. (1997) – Income Targe,ng
• “Daily targe,ng makes exactly the opposite predic,on of the intertemporal subs,tu,on hypothesis: When wages are high, [the worker] will reach their target more quickly and quit early; on low-‐wage days they will [work] longer hours to reach the target.”
• Taxi cab driving is a natural context for studying this... – “Schedules” are flexible – “Wages” fluctuate daily – Wages are correlated within day, but weakly across days – Heterogeneity of drivers – Strong wage proxies – Good data (for 1997)
Rejects posi,ve response to wages
Becer data, same context, & important component of mission
“push a bucon and get a ride in minutes”
Lye “Prime Time” and Uber “Surge”
• Wage Flexibility
– Lye “Prime Time” • Prime Time adds a percentage to your
ride subtotal. “When ride requests greatly outnumber available drivers, our system will automa,cally turn on Prime Time.”
– Uber “Surge Pricing” • “At ,mes of high demand, the number
of drivers we can connect you with becomes limited. As a result, prices increase to encourage more drivers to become available.”
Hall, Kendrick, Nosko (2015) • Inves,gates 2 events
– Ariana Grande concert at MSG – Technical glitch on NYE
• Strong response to surge pricing (counter to Diakopoulos, 2015)
– Drivers go to the surge area – Riders less inclined to request a ride
• Focus on economic efficiency of Surge – Increased total surplus (riders + drivers)
• Surge Pricing transfers some surplus from rider to driver • Higher value riders matched (lower value riders drop out) • Increased driver supply allows more matches
– Matching completed under 15 minutes is high with aid of Surge Pricing
Customer response Driver response
Natural Experiment (Technical Glitch)
Results
Driver (non) response to (non) surge
Evidence of posi,ve intertemporal subs,tu,on
Selec,on? Hall & Krueger (2015)
• Sor,ng by most opportunis,c
– Valuing flexibility • “A variety of ques,ons made it
clear that Uber's driver-‐partners value the flexibility that the Uber plaoorm permits, and many are drawn to Uber in large part because of this flexibility.”
– Outside op,ons • Most of Uber’s driving partners
con,nued full-‐ or part-‐,me jobs. “Uber’s driver-‐partners also oeen cited the desire to smooth fluctua,ons in their income as a reason for partnering with Uber.”
What I liked
• Novel, extensive data
• Simple, Clear, Robust – The result is in the visualiza,on, not the model specifica,on
• Making the most of a “natural experiment”
What I didn’t like
• The measure of (driver) responsiveness and efficiency • Diakopoulos finds heterogeneity of impact in Washington, D.C. neighborhoods
• Sharing limited results publicly – Focused on economic efficiency of Surge Pricing – Not es,ma,ng a coefficient of elas,city – Not exploring the data for more results
• What happens to the app openers who did not request
• Responsiveness to an unknown shock is the more relevant/interes,ng es,ma,on – New Years Eve and Ariana Grande are predictable events
Next steps & learnings • What mo,vates Lye/Uber drivers?
– Higher wages – Intertemporal subs,tu,on
• Farber (2005) vs. Camerer et al. (1997) – Farber argues cumula,ve hours dominate (increasing disu,lity)
• Do drivers during a surprise surge drive longer? • How does Surge Pricing impact long-‐term driver response?
• Do we observe heterogeneity in response? – Camerer et al. found posi,ve intertemporal effects in high experienced drivers, and nega,ve effects in low experienced
– Can we get increased economic efficiency through • Experience? • Informa,on / Training