12:14:39 From Lita-NBER Staff : Welcome to the Summer Institute 2020 HC Workshop 12:41:35 From heidi williams to Lita-NBER Staff(Privately) : Lita, I think we can go ahead with the live-streaming. Thank you! 12:47:52 From heidi williams to Lita-NBER Staff(Privately) : Lita, is there a direct youtube link you could send me? 12:48:04 From heidi williams to Lita-NBER Staff(Privately) : Ah, got it — never mind! 12:51:29 From Leila Agha to Lita-NBER Staff(Privately) : Hi Lita! Am I supposed to log in a special way when it is time for my presentation, so that I can share screen? 12:54:09 From Leila Agha to Lita-NBER Staff(Privately) : Never mind—I see that I have this ability now 12:54:17 From Amy Finkelstein to Lita-NBER Staff(Privately) : lita the speaker victoria can't share her screen 12:55:02 From David Slusky : Looks like everyone can share screens. 12:59:24 From David Slusky : You could do a poll.... 13:02:26 From Adrienne Sabety : Heidi, I’m not sure YouTube is getting sound 13:02:29 From Adrienne Sabety : Or Lita 13:02:42 From Adrienne Sabety : I just got a note from someone watching youtube 13:02:54 From Brad Shapiro : I get sound on YouTube 13:03:05 From Peter Hull : same 13:03:22 From Adrienne Sabety : ok cool, thank you! 13:16:24 From Michael Esman Chernew : Just FYI, the Mass connector has a small business platform. Businesses can offer employees either vertical choice (1 carrier, different levels of plans) or horizontal (1 plan level but multiple carriers). I forget, but can find out, which is most popular. Depending on what you believe about employer agency, heterogeneity of work forces, etc.) in small businesses you may infer what employers think employees prefer 13:16:47 From Lita-NBER Staff to heidi williams(Privately) : The volume seems to be ok on my end as well as the NBER IT person desk computer and laptop computer. The problem is on the participants end. 13:17:05 From heidi williams to Lita-NBER Staff(Privately) : Thanks so much for checking! 13:17:08 From Adrienne Sabety : That's fascinating Mike -- will follow-up offline 13:18:10 From Adrienne Sabety : Thank you! 13:18:39 From Peter Hull : sorry if i missed it -- what is the source of the choice set variation and why is it uncorrelated with consumer health / utilization? 13:19:05 From Adrienne Sabety : We use the benefit design for Oregon teachers ("OEBB") 13:19:13 From Keith Marzilli Ericson : The theory you developed assumes that you can set the optimal price between plans-- would your key result (high WTP have higher value) change if prices were set in equilibrium? 13:19:15 From Shooshan Danagoulian : @Ben @Adrienne Thinking about populations with high MC of care and lower WTP, I think of Medicare population, especially around end-of-life care. When we think of WTP, are we thinking of constraints that consumers face? 13:19:17 From Adrienne Sabety : Every district, and even within the district, offers a different set of plans 13:19:43 From Amanda Kreider : FYI a couple of people watching on YouTube are saying that the audio is continuing to come in and out. 13:20:29 From Peter Hull : so is the assumption something like teachers don't choose where they work as a function of the generosity of plans ? 13:20:58 From heidi williams to Lita-NBER Staff(Privately) : Lita, not sure if you saw Amanda Krieder’s note but it sounds like the YouTube audio is still getting reports of being unreliable (going in and out). Is there any way to check that further? We had ~300 people yesterday with no complaints, so it does seem like there may be a problem and with a group that big on the live-stream it would be ideal to fix if possible. 13:21:15 From Adrienne Sabety : yes, exactly. We do a lot of work to support that point in the paper 13:21:16 From Keith Marzilli Ericson : Your result of the very sick not getting much risk protection value relies on them having similar marginal utility of income? Which would rule out a case where a chronically ill person paying OOP max every year has much lower wealth. 13:21:17 From David Slusky : Do you have any other data you can use to estimate how present biased people are (meaning they would pick the high deductible) and what kind of financial resources they have (meaning whether they can self insurance at the low end if they pick a high deductible)? 13:21:40 From Daria Pelech : What is it about population B that makes no interior solution possible? Is it because consumers’ costs are bimodal? 13:21:48 From Joseph P. Newhouse : It seems as if you assume there is no non-price rationing by a plan that is integrated with delivery 13:22:00 From klavetti : vote up @ Keith's point 13:22:18 From Adrienne Sabety : Defer to Ben on this, but think this question would be great to discuss at end 13:22:21 From Peter Hull : there's probably *both* horizontal and vertical differentiation in this market. Is that important for your empirical strategy? Do you need to "separate" them? 13:23:06 From Adrienne Sabety : Exactly, we account for that 13:23:21 From Adrienne Sabety : by "separating" them if you will 13:23:26 From heidi williams : (Logistics) let me suggest that when people respond to questions, please be sure to reference @name to flag who you are responding to (mostly a note to Adrienne for now, but others for later!) 13:23:39 From Adrienne Sabety : @heidi thank you 13:23:41 From Adrienne Sabety : ! 13:27:48 From Tim J. Layton : would the "no vertical choice is best" result hold if you could not shut down the outside option of opting out of the market altogether? Seems like the condition would be less likely to hold for the insurance-uninsurance vertical decision, given that uninsurance does not have an OOP max, unlike all of the options you all consider. 13:27:52 From Keith Marzilli Ericson : A clarifying point on the interpretation of Ericson & Sydnor 2017: it's not just a behavioral econ paper! Without risk adjustment (and with perfectly rational consumers), we argue choice is welfare reducing. With risk adjustment, we find tiny gains. 13:28:08 From Keith Marzilli Ericson : (Of course, we do spend plenty of time on mistakes too) 13:28:29 From Brad Shapiro : Upvoting @Tim's question 13:29:01 From Adrienne Sabety : @Tim Important point that we talk about in the paper. Worth talking about in the discussion 13:29:28 From Wes : I may have missed this, but does the model allow for choice of uninsurance? Selection into enrollment may impact premiums and choice across vertical choices. 13:36:49 From Lita-NBER Staff to heidi williams(Privately) : I am listening to HC workshop via my laptop my volume level works. Also, I am listening to the on my desktop as well and the volume level works. The NBER IT person David can raise and lower his volume on his computer. David informed me the problem is on their in. 13:37:51 From heidi williams to Lita-NBER Staff(Privately) : Thank you for checking! You might send a note to everyone, flagging that you are responding to Amanda Krieder’s question, just so they know you did look into it. Thank you! 13:41:48 From Fiona Scott Morton : would backward sorting be likely in situations where low income causes poor health/high spending and also lowers WTP? 13:42:26 From neale mahoney : I realize we don’t typically discus and discussant, but I want to push back against Mark’s standardization argument. It might be optimal to have a choice set where plan A has high cost sharing and a broad network / no PCP-gate keepers (demand side utilization control) and plan B has generous cost sharing and a limited network / PCP gatekeeper (e.g., Kaiser). 13:43:09 From Lita-NBER Staff : The live stream volume level here at the NBER works on my desktop computer as well as my laptop computer. Also, I checked with the NBER 13:43:14 From heidi williams : I would be in favor of comments being welcome on the discussant’s remarks, at least in the chat box even if we don’t have time to share verbally :) 13:43:51 From heidi williams : (In response to @neale Mahoney) 13:45:21 From Daria Pelech : You mentioned sick people and people with big families being those who select more generous plans. Does family size matter in part because premiums don’t vary with family size as much as they should? 13:45:29 From Lita-NBER Staff : Also, I checked with the NBER IT person David and his volume level works as well. I was informed the problem is not on the NBER end. If you have headset maybe you can listen to the livestream via headset/earbuds. 13:50:03 From Amanda Kowalski : I’m curious to hear you speculate about the main explanation for backward-bending? Is it income? If so, can you investigate that formally? 13:54:03 From Erzo Luttmer : Your conclusion about choice in this setting being suboptimal gets reinforced if you take decision-making costs of individuals into account (and of employers designing menus of plans). 14:08:05 From Fiona Scott Morton : Exclusions are the only good response for an insurer facing co-pay coupons 14:08:14 From Fiona Scott Morton : ... oops the authors said it! 14:09:54 From Ben Handel : The largest PBMs are vertically integrated with insurers. Do you investigate the extent to which that impacts bargaining leverage / innovation? Obviously hard to do =) 14:09:55 From Amy Finkelstein : what is an observation in this regression? 14:10:18 From Craig Garthwaite : Ben -- those PBMs are only very recently integrated. 14:10:33 From Fiona Scott Morton : So my question was, if the environment shifted with the copay coupons, did the ability to shift share of the PBM change? They always had the co-pay tool and used to use it, but now can't. So I am not understanding for sure that there is a change... 14:10:34 From David Slusky : Where do consumer's rebate coupons fit into this? 14:10:48 From Craig Garthwaite : Would be hard to see an impact on innovation at this point. 14:10:53 From Leemore Dafny : this practice is a reaction to those coupons, which disable PBMS' ability to steer via tiers 14:11:10 From Danielle Li : @Amy — it’s a drug-year 14:11:17 From Leemore Dafny : (above was response to @DAvid) 14:11:25 From Amy Finkelstein : @danielle - drug class? 14:11:37 From Fiona Scott Morton : @leemore, yes so the ability to shift share stays present 14:12:23 From Danielle Li : @Amy this is at the individual drug-year level. We use drug class level for the later analysis 14:12:35 From David Slusky : Thanks @Leemore! I was wondering how much those coupons are coming up in the negotiations between the PBMs and a the manufacturers, and whether the authors have any time series data on them. 14:12:37 From Amy Finkelstein : thanks @danielle 14:13:11 From Danielle Li : @David, unfortunately we don’t have data on coupon use. 14:13:26 From Amy Finkelstein : @danielle - this is very cool but can you do anything to help me / show me that past exclusion practices are predictive of future ones within a class? bc isn't that what matters for innovation incentives? 14:13:30 From Ben Handel : Got it. Thanks Craig. For those who care, here is the figure I'm looking at re: vertical integration. 14:13:32 From Ben Handel : https://www.dropbox.com/s/ijr6cfiaujgrwkg/PBM.jpg?dl=0 14:14:00 From Ben Handel : Upvote Amy question 14:14:08 From Leemore Dafny : @fiona @david - exclusion is still relatively rare and would be preferable for insurers to have less blunt alternatives. They are doing more prior auth and "fail first" requirements now too...pretty cumbersome 14:14:20 From Emily Oster : Do you have a sense of what they substitute R&D effort to? 14:14:23 From David Cutler : If this is a response to copay coupons, is this the effect net of those coupons? So on net coupon + response to coupon --> lower development? 14:14:30 From Edward Norton : What fraction of national market do these 3 PBMs have? 14:14:39 From Emily Oster : @Edward: 60% 14:14:59 From Craig Garthwaite : Emily -- isn't it higher? I think it is about 75-80% by volumne 14:15:07 From Emily Oster : I was just saying what she said 14:15:16 From Danielle Li : @Amy — yes we have an analysis in the paper that speaks to this. We predict *early* exclusions (2012, 2013) using pre 2012 drug class variables. But we also show that drug classes with high predicted exclusion risk but which did NOT actually see any exclusions in 2012-13, are more likely to see exclusions later 14:15:33 From Amy Finkelstein : cool. thanks @danielle! 14:15:54 From Colleen Carey : Similar to David Cutler's comment, if many molecules in the class predict exclusion, couldn't many molecules in the class represent absence of low-hanging fruit, sense that it would be hard to make an incremental improvement? I know the variation is over time but everything here is slow-moving (including the onset of exclusions) 14:16:50 From Danielle Li : @Emily, Craig, Edward — it was 60% in 2012 but has since risen to 75-80. 14:17:10 From klavetti : is it surprising that drug development activity would respond so quickly to changes in exclusion rates? Is there a way to test for differences in the speed of responses to exclusions in early years, when the trend of increasing exclusion may not yet have been clear, compared to later years? 14:17:23 From Craig Garthwaite : Yea, that reflects ESI/Medco and Optum cateamarran merger. 14:17:44 From Brad Shapiro : upvoting @klavetti 14:17:50 From David Slusky : ↑ Kurt's question 14:17:57 From Ellie Prager : upvote @kurt 14:18:31 From Wes Yin : Reduction in innovation is a cost. Is there any savings benefit or are all reductions in manufacturer rents just shifted to PBMs? 14:18:34 From Ben Handel : Upvote Kurt's question. Is it possible to dive into R & D process and see not only initial development but subsequent promotion to market and through clinical trials? I.e. conditional phase conversion. Maybe you're doing this already. 14:19:08 From Wes Yin : @Kurt The differences by stage kind of get at this 14:19:13 From Ellie Prager : Related to @Kurt's question: can you look at heterogeneity in R&D reductions as a function of how successful the earlier trial was? That way can distinguish between magnitude of innovation and likelihood of success effects 14:19:36 From Ben Handel : Thanks Wes. I just missed that. 14:19:44 From Tim J. Layton : upvote @colleen : If restrictions more likely to be implemented in classes where there was recent innovation, seems natural for development to decline more rapidly in those classes, even absent restrictions, because low-hanging fruit was recently picked. kind of like regression to the mean. 14:20:58 From Danielle Li : @Kurt — most of our effects are found in preclinical drugs (in the paper we break it down by phase 1, phase 2, etc). Especially among larger firms, they would have a stock of early stage drug candidates in these areas. From the conversations I’ve had w pharma ppl, they grade candidates based on both scientific and *market* risk each quarter and make their go/no go decisions based on this 14:21:13 From Ellie Prager : Can you check for differential exit/takeover of firms with a large fraction of their pipeline at risk of exclusion? 14:21:26 From Danielle Li : So my sense is that the market risk for certain drugs increased a lot, and that would have an impact on how internal capital gets allocated 14:22:21 From Danielle Li : @Ellie — that’s a great idea we have not done that. However, many of the drugs we see are pre-clinical, meaning that there won’t be data on that. 14:22:57 From Ellie Prager : @Danielle that's a bummer 14:23:31 From Pierre Azoulay : One interesting margin of response is the quality of the Phase III evidence (better control groups, published trials, and generally less corner cutting?) 14:24:56 From Douglas Staiger : Are there other reasons to think this change effects innovation in low risk drugs (e.g. spillovers on the control group)? e.g., do they face less threat of entry (so more profitable) or if there is entry face lower prices (so less profitable)? Related, it seems like a drug may currently be low risk, but that could change over time (as new competing ways of treating the disease are developed). 14:26:05 From heidi williams : @Danielle: I didn’t understand your response to Ellie’s question. Why does it matter if many of the drugs are pre-clinical? Cortellis would still have firm names, and many would have patents that you could track ownership of too, no? 14:27:06 From Pierre Azoulay : Heidi, the further back in development history you reach, the more you need to worry about disclosure being less than complete. 14:27:50 From heidi williams : @Pierre, I agree with that, but they are already (it sounds like) using the early-stage R&D data, Ellie’s question was just about testing for differential exit from the sample they are already analyzing? 14:28:15 From Danielle Li : @Heidi — Sorry, I was responding to Ellie’s earlier question about whether I could use information on the past success of a drug to see if it that impacts its likelihood of continuation. We could do that for later stage drugs, but wouldn’t have info for the earlier stage drugs. 14:28:49 From heidi williams : @Danielle, ah — got it. I guess we need to reference Ellie (2020a) and Ellie (2020b) :) 14:29:09 From Danielle Li : @Ellie, Heidi. To Ellie 2020b, yes, I think we could do that! 14:29:10 From Fiona Scott Morton : Question: if coupons increase bargaining power but the exclusion raises it, how do the authors know there is an increase in plan bargaining power? 14:29:21 From Ellie Prager : @Heidi @Danielle thanks for clarifying. Not a pharma person so I don't understand the data landscape! 14:29:52 From Ellie Prager : (Also thanks for the attribution of two new NBER WPs) 14:30:46 From Seth A. Seabury : Have you tried a specification that excludes cancer drugs? Anecdotally, there seemed to be a big push in cancer following the first successful immunotherapy drugs coming out in 2011. My sense is that these are low exclusion risk. 14:31:41 From David Slusky : Can you link this up with the drug detailing data that @Colleen @Sarah @Ethan have used? Did drug companies increase detailing in response to exclusions? 14:32:43 From Colleen Carey : Open Payments starts reporting in August 2013, so data is not great for the period before 2012 14:33:33 From David Slusky : @Colleen right. I wonder if there's enough variation across drug classes to still study it 14:35:15 From Danielle Li : @Seth — we have not, but could try. And, yes, cancer is a low exclusion risk category. 14:36:02 From Danielle Li : @Fiona — coupons were used prior to our policy change, so we would argue that plan bargaining power in affected classes increased in a relative sense. 14:36:50 From Fiona Scott Morton : Craig - can you stop sharing? 14:42:14 From Ben Handel : I have Amitabh's number too if anyone else wants it 14:42:55 From Pierre Azoulay : Not in cortellis unfortunately 14:43:52 From Danielle Li : Calling out @Jennifer Kao, who has collected this data and is working on a project about clinical trial design! 14:45:43 From Fiona Scott Morton : I am raising my hand 14:45:45 From Wes Yin : Following on Amitabh's q on bargaining, how much of the PBM's increased bargaining power (and manufacturer revenue reduction) is retained or transferred to consumers? 14:46:17 From Craig Garthwaite : They didn't talk about "formulary risk" but they definitely talked about copayment/coninsurane risk. 14:48:26 From Colleen Carey : If drug A is on the market, and drug B is an incremental improvement, exclusion is not purely a risk for drug B. It will suddenly become a risk for drug A as well. In a world without exclusions, these might have been in a duopoly where they are both covered universally by plans. In a world with exclusions, they are still in a duopoly, but coverage is only supplied by one or the other PBMs. 14:49:16 From Colleen Carey : One thing that makes me concerned about this was that the chart of exclusions over time had a y-axis going to 150, and a total exclusions of 300, with mostly two PBMs excluding. This would suggest one PBM is excluding drug A, the other drug B. 14:50:16 From Danielle Li : @colleen, yes there are some PBMs that exclude one drug in a class and other PBMs that exclude the other. But I think that speaks to the fact that all drugs in that *class* are at risk, and our analysis is at the class level. 14:50:17 From Craig Garthwaite : Colleen -- my sense is that you often have different drugs for each PBM. That "sense" might be biased by a few high profile cases. 14:51:05 From Ashley Swanson : Could catastrophic phase shenanigans be interesting here, at least for Part D? I.e., PBMs being more willing to keep pricey drugs on formulary (with big rebates) when the govt is going to pick up a lot of the tab (post catastrophic threshold), and moreso after 2012 when rebates take off. Not sure if there’s variation in drugs’ exposure to catastrophic phase… 14:51:16 From Colleen Carey : @Danielle, that makes sense, I think I was just hearing it discussed the other way a lot 14:51:36 From Fiona Scott Morton : @Ashley. yes, the higher sticker price gets people to the max quicker and offloads expense onto te government 14:53:05 From Naoki Aizawa : Hi moderators: I (Naoki) will present the next paper and You Suk will reply in the chat. I will share my screen around pm 2:58. 15:05:30 From Maria Polyakova : Minor, but seems it would be useful to be clear that the government is not just advertising and private firms providing coverage, but the government is actually paying a high fraction of costs for providing coverage. 15:05:52 From You Suk Kim : @maria. Well noted. thanks 15:06:44 From Ellie Prager : Does gov't advertise more through non-TV channels (esp via community organizations) than private insurers? And if so, do you have a sense of how closely that lines up with TV market boundaries? 15:07:24 From neale mahoney : If firms can form an industry group to engage in joint advertising, do we need to worry about business steeling / spillovers. The industry group can internalize the externalities. 15:07:48 From Amy Finkelstein : @neale: why would they want to? 15:07:57 From Amy Finkelstein : don't they want to business steal? 15:08:21 From Brad Shapiro : @amy @neale they would want to in the case of positive spillover to shut off the externality. In business stealing, they might want to collude not to advertise 15:08:29 From neale mahoney : Privately optimal, but a prisoners dilemma. 15:08:42 From You Suk Kim : @Ellie we do not have great info on non-TV channels. There are some papers looking at the effect of direct mailings sent by the government based on individuals characteristics. If such activities are based on individual characteristics, it is not less likely to be correlated with TV market borders. 15:08:46 From Amy Finkelstein : right thanks @brad / @neale 15:09:59 From Amy Finkelstein : @neale: but by same argument if we all form an industry gorup, don't I have incentive to cheat? how does industry group enforce the cooperative equilibrium? 15:10:14 From Ellie Prager : @You Suk thanks. DMAs are based on county boundaries, right? If you can do something simple in the data like overlaying county-level pre-ACA uninsured/Mcaid shares on your DMA borders, I think that would assuage my concern 15:10:27 From David Slusky : Building on @Ellie's comment, I think that the non-TV channels are really important, especially as more and more people are getting their ads from podcasts, youtube, Twitter, etc. 15:10:32 From Wes Yin : @amy there's also a lot of above brand advertising, likely to boost political and public optics of pharma (esp in light of pricing). Just a specific eg of externalities that Ben mentioned. 15:11:00 From Brad Shapiro : @Ellie- a DMA is usually ~20-30 counties. Boundaries don't neatly line up with much else. They don't typically coincide with state boundaries, for example. 15:11:13 From Leila Agha : Does per capita advertising make sense as an exposure measure? Is this because this is ad $ spent, which scales with market size? Just trying to understand the exposure variable 15:11:19 From You Suk Kim : @Ellie we do something similar in the paper. We have a table that compares characteristics of either side of a border, and we find characteristics are quite similar 15:11:57 From Ellie Prager : @Brad @You Suk perfect. If you don't specifically have uninsured/Medicaid share in the table, I think worth adding 15:12:24 From Maria Polyakova : Is geo variation in advertising positively or negatively correlated with how subsidized consumer are locally? 15:12:39 From Brad Shapiro : @Leila- the standard measure in marketing literature is GRP which is exposures/capita. I've found that expenditures/capita in kantar is highly correlated with GRP. This is because expenditures are "list prices" that are derived from exposure data. So it's reasonably mechanical. 15:12:39 From You Suk Kim : @David I agree. But it is not really clear a prior whether exposure to ads from podcasts or the internet in general is systematically correlated with TV ad spending across TV market borders. 15:12:56 From Leila Agha : thanks @brad 15:13:25 From Brad Shapiro : @Leila could alternatively just use number of occurrences, which is also reasonably correlated, but noisier 15:13:49 From David Slusky : How sharp is this discontinuity? It's not like school zones - if I'm on the border of two media markets shouldn't I be able to get both channels? 15:14:07 From You Suk Kim : @Maria, we have not checked the subsidy level. We do include the rating area x year FE. Maybe this address your point to some extent? We can check it explicitly. Thanks 15:15:00 From neale mahoney : If private spend is endogenous to state and fed spend, how do we interpret this regression. 15:15:03 From Brad Shapiro : @David- it's very sharp if you have cable or dish. Not as much if you're watching over the air. Not that many people watch over the air, especially away from cities. 15:15:15 From Shooshan Danagoulian : @You Suk The federal advertising is relatively small compared to private markets. Is it acting as a signal about future of health exchanges and profitability of these markets? 15:15:22 From Jason Abaluck : Do you have a sense of the absolute magnitude of $ in fed advertising per marginal enrollee? 15:15:39 From Ellie Prager : Upvote @Jason 15:16:03 From David Slusky : Thanks @Brad! (This is what happens when spend my entire life either without cable/dish or without a TV.) 15:16:12 From You Suk Kim : @Neale we assume that all types of ads (fed or private) are close to random within the border pair. 15:16:33 From neale mahoney : To follow up on my point, we have “1 instrument” (borders). How can we learn about multiple parameters. 15:17:42 From Amy Finkelstein : @you suk - i didn't understand your answer to @neale? 15:17:53 From David Slusky : What about the content of the ads? Are they all pro-enrollment? Are some of them pro/anti Medicaid expansion (in non expansion states)? What about intentional misinformation ones from ACA opponents? 15:19:23 From You Suk Kim : @Jason We have not calculated the number. That sounds worth doing it in the next iteration of the paper. Thanks 15:19:36 From neale mahoney : @you suk: In the model, shouldn’t a firm advertise less if the fed gov’t advertises more (competitive substitutes). 15:20:40 From Maria Polyakova : @Neale, couldn't they also be complements? If gov't ads make more people generally interested in enrollment, then I want to advertise more for my specific firm to get those new people to enroll in my plan? 15:21:29 From Brad Shapiro : @Neale @Maria both are right I think- could go either way, and very difficult to tell the difference given the amount of noise in ad data. The most relevant cross partial is in the appx. Imprecisely estimated. 15:21:30 From You Suk Kim : @neale @any With the border strategy, we just assume any variation in advertising variables is random within a small set of counties along the TV border. If this still does not answer your question, we can follow up in Q&A 15:21:45 From Haizhen Lin : do you model government decision on advertising? 15:21:49 From neale mahoney : Yes, we should chat more during the Q&A. 15:22:15 From You Suk Kim : @haizhen No. We do not.. 15:23:07 From Haizhen Lin : how do you see that government might be strategic to firms' decision? 15:23:51 From Amy Finkelstein : try studying health economics @brad :) 15:24:40 From You Suk Kim : @haizhen That indeed might be possible. In this paper, we try to keep things simple by simulating only private behaviors with exogenous changes in gov ads 15:26:20 From Ellie Prager : @Brad I want this slide for teaching our PhD students! 15:28:51 From Sarah Miller : lol 15:36:44 From heidi williams : Let me advertise one more — ReStat’s new “short paper policy” — just to be fair to the competition for Amy’s AER:Insights :). https://www.mitpressjournals.org/journals/rest/sub?mobileUi=0 15:37:11 From Joshua Gottlieb : Great idea @Brad to expand AER:Insights style. Journal of Public Economics is doing exactly what you say at least in a limited context. All of you should submit your short and excellent COVID-19 papers! https://www.journals.elsevier.com/journal-of-public-economics/call-for-papers/call-for-papers-the-public-economics-of-covid-19 15:37:14 From Ben Handel : Yes. As an Editor at ReStat: advertising our short paper format! 15:37:27 From Ben Handel : We are really looking for shorter crisp paper like the one Brad is proposing here. 15:37:34 From Peter Hull : but what is the effect of such advertising.....? 15:37:41 From Joshua Gottlieb : 50% 15:37:51 From Brad Shapiro : @everyone :: facepalm :: 15:39:13 From Brad Shapiro : @Ellie- I posted the slides to the conference website. Feel free to use them as you like (with attribution ;-)) 15:42:56 From Amy Finkelstein : if others had comments in the chat or not yet chatted they should raise their hands! i didn't mean to ignore comments i just have trouble findgn them all! 15:45:20 From Brad Shapiro : Re: chat being super active-- There is easily 5X as much volume in the chat as there was at the IO meetings last week. And with fewer participants! 15:45:32 From Brad Shapiro : (it's amazing) 15:46:43 From Amy Finkelstein : @ brad - i'm sure it's due to the far superior leaders ofhealth care than of io (@liran :))) 15:47:11 From Ellie Prager : @Brad @Amy my bet was on the lack of greek letter functionality in zoom 15:47:30 From Peter Hull : β 15:48:15 From David Slusky : @yoo suk given you have the video of the ads, and given how good Zoom's auto transcription is, I wonder if you could transcribe them and then use some kind of machine learning algorithm to see which phrases are in them and which are more useful 15:49:41 From Wes Yin : State advertising and other state initiatives (outreach to community organization, DMV, radio, etc). How do we think about the effect of state coefficient and interactions in light of this? 15:50:03 From Wes Yin : These are big issues in CA (which the authors highlight in their paper) 15:52:07 From Wes Yin : State initiatives are determined jointly 15:52:36 From Ben Handel : @Wes, very interesting. 15:53:00 From Ellie Prager : Upvoting @Doug's idea to benchmark against subsidy spending (if you can do it) 15:53:55 From You Suk Kim : @Wes that is an interesting point. Unless these activities are coordinated around the TV market border, that won’t affect our estimates for state ads. But that could potentially interact with effects of state ads. We do find some evidence of heterogeneity in state ads effects. Especially CA has larger effects 15:55:50 From Wes Yin : @You Suk But this also relates to Doug's point about the marginal dollar identified in your studied versus the inframarginal or average advertising dollar. Baseline initiatives in TV and other efforts seem to be important and will affect interpretation of your estimated coefficiencts estimated at borders, no 15:55:51 From Wes Yin : ? 15:57:40 From Brad Shapiro : @Wes - since the advertising is targeting the average of the whole DMA rather than the borders, you can end up with big deviations from the "optimal microtargeted" advertising, even if there isn't huge overall variance in residual advertising. Sometimes there is a lot of advertising on the border of Cleveland DMA despite low demand just because Cleveland proper has high demand, for example, and vice versa. This can give variation all along the curve, in principle. I think the effect they estimate is closer to the "average" effect than to the "marginal" effect, though it is difficult to prove, as it were. 15:57:50 From Brad Shapiro : Something I've thought about a lot, though, so happy to discuss offline 16:08:06 From Amitabh Chandra : I do like that Harvard Medical School is getting credit for me without employing me. 16:08:14 From Amitabh Chandra : I really hope that this tweet makes it into the literature-review section of the published paper. 16:08:21 From heidi williams : Not central to this paper, but it would be interesting if you could document descriptively how, within a university, research in the same “field” differs across departments housed in the “main” university versus the medical school on the basic vs. bedside margin (e.g. GSAS vs. Harvard Medical School). 16:09:13 From heidi williams : In case anyone missed @Amitabh’s joke, please see @Pierre’s zoom background :) 16:09:23 From David Cutler : @Heidi most biomed research at the med school. Only a few people do in a biology department (e.g., ours studies birds a lot). 16:11:27 From Pierre Azoulay : David just canceled MIT and Berkeley... 16:12:09 From Shooshan Danagoulian : @Heidi In this period there is also an increased collaboration in clinical trials between device manufacturers/pharmaceutical companies and academic medical centers, especially in centers which have less federal research support. What is the role of these private industry contracts in this? 16:14:00 From Pierre Azoulay : @Shooshan, actually, the story is a bit more complicated. Lots of competition from for-profit dedicated study sites. I have another paper on this topic. We will show the CT results separately (but only deals with published CTs). 16:14:10 From Ellie Prager : I think I'm missing something obvious: What does the DSH variation buy you here? Aren't DSH hospitals and AMCs non-overlapping (so DSH aren't great controls to begin with)? 16:14:41 From Pierre Azoulay : There is some DSH in our sample. But you are correct, IME is where the action's at. 16:14:47 From Douglas Staiger : DSH interacts with the teaching adjustment (hospitals get both, multiplied) 16:15:00 From Shooshan Danagoulian : @Pierre Thank you! 16:15:35 From Pierre Azoulay : @Doug, that is true, and we did our best to capture this. Outlier payments is the stuff that is really hard to deal with cost report data. 16:15:55 From Adam Sacarny : If there is enough of it, the DSH variation could be helpful as it creates variation in the payment shock even conditional on teaching intensity 16:16:36 From Pierre Azoulay : Jen is the DSH-head in the coauthorship team, so I will let her address this later... 16:18:31 From Ellie Prager : The grant results are cool! Can you see heterogeneity by the AMC's NIH institutional score? That's probably the best predictor of grant success (correct me if I'm wrong) so it would tell you something about how costly it is to submit a grant (lots of work goes into them) 16:19:12 From Pierre Azoulay : @Ellie, we can certainly do this! Only a question of statistical power. 16:19:59 From David Slusky : Can you get demographic data on the grant recipients? My colleague Donna Ginther (who is also Misty's coauthor) has done lots of work on this. 16:20:00 From Ellie Prager : @Pierre would be very cool if you're powered for it! 16:20:58 From Misty : There is demographic data within the IMPACII data...unclear whether we can get access to it at this point since none of us currently work at NIH. :) 16:21:12 From Ellen Meara : But since every institution I’ve ever worked at reminds us that research grants, even federal grants, don’t cover overhead, why/how does giving out “hunting licenses” really help? 16:22:25 From Adam Sacarny : Do you have a sense of the translation from BBA bite to total hospital payment rate (or just Medicare payment rate) over time, as a kind of first stage? I think this would help to scale the effect, and could also help show that the BBA cuts persisted / weren’t undone by subsequent payment changes. 16:22:26 From Pierre Azoulay : @Ellen: one explanation is that they lie. Another is that they do not understand their costs. Or at least no everyone in the institution does. 16:22:28 From Joshua Gottlieb : @Ellen they surely cover marginal overhead 16:22:37 From Maria Polyakova : Seems like another thing that could be going on is that people are diverting their grant application plans from various internal "grant" resources (which is a way to tap clinical revenue for research) into external grant applications. So not clear if they are really doing more or doing more things that are publicly measurable. 16:23:13 From Ellie Prager : Upvote @Maria 16:23:44 From Craig Garthwaite : I don't understand why David thinks people can't multi-task and enjoy a quarantini during his talk? 16:24:53 From heidi williams : One positive benefit of zoom is that the comments that e.g. Craig would have otherwise whispered to the person next to him are now enjoyed by all :) 16:25:29 From Amy Finkelstein : it's certainly an empirically correct descriptive statement. not sure i'd rush to normative interpretation @heidi @craig :) 16:25:30 From Fiona Scott Morton : Nice! @craig and @Heidi. I almost feel like I am at the Sonesta 16:25:39 From Pierre Azoulay : apocryphal or apocalyptic? 16:26:08 From Craig Garthwaite : That comment doesn't apply to Fiona (or Amy) who don't really whisper. 16:26:37 From Ellen Meara : @Amy & @Heidi I think we might need to preserve the chat even in a post-coved world. Not that I can keep up, but it’s like listening in on every side chat and allows us to hear more questions. 16:28:26 From Fiona Scott Morton : Any and I are just broadly appreciated ;-) 16:29:20 From Daria Pelech : Budget surpluses! What a weird concept. 16:30:42 From Mireille Jacobson : Standard errors are large but didn’t the figure suggest bench science declined? I may have missed this. 16:31:28 From Joshua Gottlieb : @David how accurately do you think costs would be allocated between research and clinical in the cost reports? 16:33:29 From Pierre Azoulay : I apologize...listening to David but keep these questions coming! 16:34:40 From David Cutler : @Josh, I don't think you can separate them out. I think one would have to look at costs as a whole, or perhaps costs by department. 16:37:36 From Mark Shepard : Can you say more about how you separate the causal effect of the BBA cuts versus differential trends in which types of hospitals are doing research (e.g., research shifting to bigger, more teaching-heavy institutions)? 16:45:35 From David Slusky : What about changes in individual physician professor's protected research time and time spent on clinical work? 16:45:52 From David Slusky : You obviously can't see the individual contract and agreements, with the NPIs on the Medicare claims data you could potentially infer it. 16:45:53 From Ellie Prager : @David super interesting but sounds impossible to measure 16:46:18 From David Slusky : @Ellie does that help? 16:46:41 From Ellie Prager : @David I wouldn't trust that inferred measure without looots of validation using contract data 16:46:56 From David Slusky : Meaning, specialty adjusted charges per physician. 16:47:44 From David Cutler : Lose money on every grant but make it up on volume! 16:48:11 From Joshua Gottlieb : They make it up by losing money on patient care! 16:48:41 From Craig Garthwaite : Shots fired at MD's … 16:48:51 From Joshua Gottlieb : Not by me 16:56:12 From Sarah Miller : Thank you organizers