And now, as promised, a thread on my new *working* paper (feedback welcome): “Promoting Wellness or Waste? Evidence from Antidepressant Advertising” (link) 1/22

https://papers.ssrn.com/abstract=3130327

DTCA is controversial. Opposition has emphasized that it could 1) over-exaggerate benefits, causing people to be prescribed who don’t stand to benefit and 2) because drugs are expensive, this is very costly. Additionally, it could 3) steer consumers to expensive drugs. 2/22

Not everyone agrees. The flip side of the argument is that many diseases (e.g. depression) are poorly understood by patients, and DTCA could inform patients that their symptoms can be treated and give refill reminders. 3/22

From my vantage point, these proposed costs and benefits from each side are both theoretically plausible and empirically measurable. So rather than debate blind, we should try and measure them. I do this in the context of antidepressant advertising. 4/22

Preliminaries: 1) I have 0 industry funding for this study. 2) I do not claim to measure every possible cost and benefit, only that those I measure are important to know. 3) I make no claim that the results hold for other types of drugs. 5/22

To evaluate the magnitude of proposed costs, I estimate the effect of DTCA on 1) New prescriptions (and drug spending), 2) Refills, 3) Price conditional on prescription and 4) Generic penetration rate from individual insurance claims data. 6/22

To evaluate the potential benefits, I estimate the effect of DTCA on labor supply from human resources records, which is the outcome most associated with depression. If advertising attracts people who don’t stand to gain, there should be minimal effect. 7/N

I could be missing things. Extended discussion in the paper. In particular, people might hate watching the ads, which is costly. Also, if employers have market power, they will pay workers less than MPL, so wage benefits miss gains to employers. 8/22

Method: Ads on TV are non-random firm decision. Could be spuriously correlated with both Rx and outcomes. To overcome this hurdle to causal inference, I exploit random variation in TV ads generated by the borders of TV markets (as in Shapiro 2018: https://papers.ssrn.com/abstract=2477877) 9/22

Intuition is as follows- TV markets are relatively large. People close to the borders of these markets but on opposite sides are generally similar (all else equal), but they see different ads. Add to this individual level fixed effects. 10/22

Results: 1) Ads increase new antidepressant prescriptions modestly (ad elasticity=0.03). This is consistent with previous work showing category expansive effects of DTCA. 10% increase in DTCA->~$32 million increase in spending. 11/22

Results: 2) Ads decrease refills/adherence very modestly (elasticity= -0.004). This suggests the proposed benefit of “refill reminders” is not really happening. Also decreases that $32 million estimate a bit. 12/22

Results: 3) While ads decrease adherence slightly, lagged ads are not associated with decreased first month adherence. Point estimate is positive. This suggests that "advertising marginal" from last month are not more likely to discontinue use. 13/22

Results: 4) Ads have precise null effect on a) price of drug chosen, b) co-pay of drug chosen and c) the generic penetration rate. The proposed cost of ads "steering" consumers to expensive drugs is unfounded in this setting. 14/22

Results: 5) It takes time for antidepressants to "work", so immediate effect on outcomes not expected. Lagged advertising (past 6 months) significantly decreases absenteeism. 10% increase in ads-> ~0.4% increase in labor supply 15/22

Results: 6) These labor supply increases are concentrated entirely on people who missed more work than median in the preceding 6 months. Back of the envelope, 10% increase in ads -> $770 million in wage benefits. 16/22

Discussion: 1) Do these labor supply increases come exclusively from taking the antidepressants themselves? I can't say for sure given stat power. Some could come from viewing ad and dealing with depression in an alternative way. 17/22

Discussion: 2) Important to learn this, because if effect comes mainly from something other than the drug, then the benefit could potentially be achieved through other means, such as disease awareness PSAs 18/22

Discussion: 3) Best I can say is it appears that at least *some* of effect is coming from the drugs, as the lag timing (in apx) shows effect kicking in at 2 months, which is consistent with antidep timing of effectiveness from RCTs. 19/22

Overall Summary 1: It appears antidep ads are acting primarily on patients who stand to gain and those individuals are eventually gaining ($770m wages per 10% increase) considerably more than the cost of the drugs ($32m per 10% increase). 20/22

Overall Summary 2: While these specific results only apply to antideps, it is heartening that ads seem to be most effective on those that stand to gain from treatment. That "selection" *might* plausibly extend to other categories 21/22

Overall Summary 3: Additionally, these results suggest that when regulating the potential harms of DTCA, we probably need more nuanced policy than a blanket ban, and it is important to actually measure rather than assume the magnitude of the costs and benefits. 22/22