1 00:00:00,000 --> 00:00:01,916 [SQUEAKING] 2 00:00:01,916 --> 00:00:03,832 [RUSTLING] 3 00:00:03,832 --> 00:00:05,269 [CLICKING] 4 00:00:10,550 --> 00:00:14,300 FRANK SCHILBACH: So the quizzes are not meant to be difficult. 5 00:00:14,300 --> 00:00:18,170 They're mostly meant to provide some incentives for you to, A, 6 00:00:18,170 --> 00:00:21,620 keep up with the class, B, to do the reading, C, 7 00:00:21,620 --> 00:00:23,455 to come to class. 8 00:00:23,455 --> 00:00:25,330 Are they actually a commitment device or not? 9 00:00:31,600 --> 00:00:33,388 Yes? 10 00:00:33,388 --> 00:00:35,243 AUDIENCE: [INAUDIBLE] 11 00:00:35,243 --> 00:00:37,910 FRANK SCHILBACH: In general, are the quizzes a commitment device 12 00:00:37,910 --> 00:00:40,190 or not by the definition that we had? 13 00:00:40,190 --> 00:00:41,662 AUDIENCE: Yes. 14 00:00:41,662 --> 00:00:42,870 FRANK SCHILBACH: Why is that? 15 00:00:42,870 --> 00:00:45,810 AUDIENCE: Because we [INAUDIBLE].. 16 00:00:49,012 --> 00:00:49,970 FRANK SCHILBACH: Right. 17 00:00:49,970 --> 00:00:53,317 But I guess one condition that we have in the commitment 18 00:00:53,317 --> 00:00:55,650 device-- and that's more like a question of definition-- 19 00:00:55,650 --> 00:01:00,350 is like it's chosen by the agent to actually engage in this, 20 00:01:00,350 --> 00:01:02,070 and I guess you don't have the choice. 21 00:01:02,070 --> 00:01:03,528 So in some sense, if I offer to you 22 00:01:03,528 --> 00:01:06,770 to choose quizzes versus not and you could particularly only 23 00:01:06,770 --> 00:01:08,960 do worse by doing the quizzes, I could offer you 24 00:01:08,960 --> 00:01:11,180 the choice either you'd always get 5 on the quizzes, 25 00:01:11,180 --> 00:01:14,240 or you can opt into saying you do the quizzes, 26 00:01:14,240 --> 00:01:17,250 then you choosing the quizzes would very much be a commitment 27 00:01:17,250 --> 00:01:17,750 device. 28 00:01:17,750 --> 00:01:20,780 But if I sort of just impose it on you, 29 00:01:20,780 --> 00:01:23,060 then, strictly speaking, it is not, 30 00:01:23,060 --> 00:01:25,580 even though it sort of functions in the exact same way 31 00:01:25,580 --> 00:01:28,570 as if you had chosen it. 32 00:01:28,570 --> 00:01:29,750 OK. 33 00:01:29,750 --> 00:01:32,630 Mostly we are discussing empirical questions 34 00:01:32,630 --> 00:01:35,480 related to hyperbolic or quasi-hyperbolic discounting. 35 00:01:35,480 --> 00:01:37,010 I want to have a short intermezzo 36 00:01:37,010 --> 00:01:38,990 very briefly about sort of telling you 37 00:01:38,990 --> 00:01:41,990 about how to solve problems with quasi-hyperbolic discounting 38 00:01:41,990 --> 00:01:44,420 in part because that's relevant to the problems set. 39 00:01:44,420 --> 00:01:46,280 I already talked with you briefly 40 00:01:46,280 --> 00:01:48,800 about how do you deal with fully naive agents 41 00:01:48,800 --> 00:01:51,770 and how do you deal with fully sophisticated agents. 42 00:01:51,770 --> 00:01:54,620 In some sense, it's fairly simple in these two cases, 43 00:01:54,620 --> 00:01:57,990 in part because these are special extreme cases. 44 00:01:57,990 --> 00:02:03,540 Now, the third case is, what do you 45 00:02:03,540 --> 00:02:05,700 do when you have a partially naive decision-maker? 46 00:02:05,700 --> 00:02:08,009 And that's a little trickier, and we 47 00:02:08,009 --> 00:02:10,440 tend to not have that many problems 48 00:02:10,440 --> 00:02:12,790 related to partially naive decision-makers. 49 00:02:12,790 --> 00:02:15,480 [INAUDIBLE] is a little tricky to do. 50 00:02:15,480 --> 00:02:17,670 So essentially, what you do is you kind of 51 00:02:17,670 --> 00:02:20,250 do the same as you do for the fully sophisticated 52 00:02:20,250 --> 00:02:21,060 decision-maker. 53 00:02:21,060 --> 00:02:23,862 You do start by doing backwards induction. 54 00:02:23,862 --> 00:02:25,320 You essentially solve for what does 55 00:02:25,320 --> 00:02:28,950 the person think they're going to do in the future. 56 00:02:28,950 --> 00:02:36,720 For that-- I'm speechless. 57 00:02:41,013 --> 00:02:46,890 For that, what do you do is you use people's beta hat, 58 00:02:46,890 --> 00:02:49,920 and beta hat is essentially what the person believes 59 00:02:49,920 --> 00:02:51,510 they will do in the future. 60 00:02:51,510 --> 00:02:55,050 And then, essentially, solve using backwards induction what 61 00:02:55,050 --> 00:02:56,730 the person thinks they will do. 62 00:02:56,730 --> 00:02:58,290 This is the exact same thing as you 63 00:02:58,290 --> 00:03:00,630 do for a sophisticated person, except for that you're 64 00:03:00,630 --> 00:03:02,460 using now people's beta hat as opposed 65 00:03:02,460 --> 00:03:04,428 to people's actual beta. 66 00:03:04,428 --> 00:03:06,720 The difference is, of course, it's always the beta hat. 67 00:03:06,720 --> 00:03:09,120 It just happens to be for the sophisticated person 68 00:03:09,120 --> 00:03:11,678 that the beta hat actually is the beta. 69 00:03:11,678 --> 00:03:13,470 So then, essentially, you start at the end. 70 00:03:13,470 --> 00:03:14,460 You do backwards induction. 71 00:03:14,460 --> 00:03:16,380 You go back in time, work yourself [INAUDIBLE] 72 00:03:16,380 --> 00:03:18,735 period two, so then the person using the beta hat 73 00:03:18,735 --> 00:03:20,610 knows what they're going to do in the future, 74 00:03:20,610 --> 00:03:22,902 thinks they know what they're going to do in the future 75 00:03:22,902 --> 00:03:24,607 starting in the next period. 76 00:03:24,607 --> 00:03:26,940 Then you're going to start solving for the first period, 77 00:03:26,940 --> 00:03:29,040 like what is the person actually going to do? 78 00:03:29,040 --> 00:03:31,190 For that, what you need is the person's actual beta 79 00:03:31,190 --> 00:03:32,940 because that's essentially what determines 80 00:03:32,940 --> 00:03:35,130 people's actual choices. 81 00:03:35,130 --> 00:03:36,780 And then you go to the next period 82 00:03:36,780 --> 00:03:38,100 and do exactly the same thing. 83 00:03:38,100 --> 00:03:39,850 Again, you do sort of backwards induction. 84 00:03:39,850 --> 00:03:43,950 You already have solved that from the previous step. 85 00:03:43,950 --> 00:03:45,780 But then, essentially, in period two, 86 00:03:45,780 --> 00:03:48,210 then, the person actually uses again their actual beta 87 00:03:48,210 --> 00:03:49,980 for making their decisions. 88 00:03:49,980 --> 00:03:52,260 So the beta hats are guiding people's beliefs 89 00:03:52,260 --> 00:03:54,340 about what they think they will do in the future, 90 00:03:54,340 --> 00:03:56,048 but then you have to go back and actually 91 00:03:56,048 --> 00:03:57,820 look at what does the person actually do. 92 00:03:57,820 --> 00:04:00,900 And that often differs, in particular 93 00:04:00,900 --> 00:04:03,780 when you get to like period two, three, and four in the future. 94 00:04:07,480 --> 00:04:08,604 Sorry, this is not-- 95 00:04:13,928 --> 00:04:15,470 The second thing I wanted to say is-- 96 00:04:15,470 --> 00:04:18,470 and this is mostly relevant for problem set two. 97 00:04:18,470 --> 00:04:21,019 So far, we have essentially looked at continuous choices. 98 00:04:21,019 --> 00:04:22,670 It's like going to a movie versus not, 99 00:04:22,670 --> 00:04:24,680 or at some point you have like one movie that you can go 100 00:04:24,680 --> 00:04:26,388 to which essentially is sort of following 101 00:04:26,388 --> 00:04:29,420 the O'Donaghue-Rabin papers and examples. 102 00:04:29,420 --> 00:04:31,977 You have one problem set you need to do at some point. 103 00:04:31,977 --> 00:04:33,560 This is essentially a discrete choice. 104 00:04:33,560 --> 00:04:35,720 Do you do it, or do you not do it? 105 00:04:35,720 --> 00:04:38,780 Of course, many choices in life so exactly like that, 106 00:04:38,780 --> 00:04:40,880 like, essentially, when do you do the problem sets 107 00:04:40,880 --> 00:04:44,310 and so on and so forth, but other choices in life 108 00:04:44,310 --> 00:04:47,210 are continuous consumption savings or continuous choices. 109 00:04:47,210 --> 00:04:49,820 It's about like how much do you eat, how much do you consume, 110 00:04:49,820 --> 00:04:51,810 and so on and so forth. 111 00:04:51,810 --> 00:04:54,200 And sort of the canonical thing to do in economics 112 00:04:54,200 --> 00:04:55,740 are consumption savings choices. 113 00:04:55,740 --> 00:04:56,990 You have some amount of money. 114 00:04:56,990 --> 00:04:58,100 You think about how much should you 115 00:04:58,100 --> 00:05:00,200 consume in this period, how much should you save. 116 00:05:00,200 --> 00:05:01,040 There's an interest rate. 117 00:05:01,040 --> 00:05:03,082 You're going to get more money than in the future 118 00:05:03,082 --> 00:05:05,360 if you save more and so on and so forth. 119 00:05:05,360 --> 00:05:07,280 The same principles here apply as well 120 00:05:07,280 --> 00:05:10,450 except for now you're not comparing discrete choices, 121 00:05:10,450 --> 00:05:12,950 but now you do, essentially, [INAUDIBLE] conditions the same 122 00:05:12,950 --> 00:05:16,040 what would you do in any optimization problem 123 00:05:16,040 --> 00:05:18,860 as you would do in 1401 or one or any sort 124 00:05:18,860 --> 00:05:25,520 of intro micro class where, essentially, you optimize 125 00:05:25,520 --> 00:05:27,770 using [INAUDIBLE] conditions, et cetera, 126 00:05:27,770 --> 00:05:30,860 subject to some budget constraint. 127 00:05:30,860 --> 00:05:33,200 In the recitation, you get some additional guidance 128 00:05:33,200 --> 00:05:35,253 how to do that, and particularly, the TAs 129 00:05:35,253 --> 00:05:36,920 will walk you through a previous problem 130 00:05:36,920 --> 00:05:39,480 from like a few years ago that essentially 131 00:05:39,480 --> 00:05:40,910 is sort of doing exactly that. 132 00:05:40,910 --> 00:05:43,370 I just want to sort of flag this so you're not surprised. 133 00:05:43,370 --> 00:05:47,240 This is essentially sort of 1401 basics 134 00:05:47,240 --> 00:05:49,100 that we discussed in the first recitation. 135 00:05:49,100 --> 00:05:52,760 There's also some handout on some refreshers for that, 136 00:05:52,760 --> 00:05:54,270 but that'll come as well. 137 00:05:54,270 --> 00:05:55,820 But the same principles apply here. 138 00:05:55,820 --> 00:05:59,600 That's exactly the same thing as we do here in the other choices 139 00:05:59,600 --> 00:06:03,230 that people make except for now we 140 00:06:03,230 --> 00:06:05,990 have, essentially, continuous choices. 141 00:06:05,990 --> 00:06:07,100 Any questions on this? 142 00:06:09,850 --> 00:06:10,570 OK. 143 00:06:10,570 --> 00:06:11,070 All right. 144 00:06:11,070 --> 00:06:13,278 So then we're going to talk about other applications, 145 00:06:13,278 --> 00:06:18,390 starting with credit card teaser rates or credit card behavior. 146 00:06:18,390 --> 00:06:20,430 This is an old paper by Ausubel [INAUDIBLE].. 147 00:06:20,430 --> 00:06:22,680 You can have any concerns about it, but in some sense, 148 00:06:22,680 --> 00:06:27,360 it's sort of nicely illustrates very simple behaviors 149 00:06:27,360 --> 00:06:29,880 that are very nicely fitting into the quasi-hyperbolic 150 00:06:29,880 --> 00:06:31,140 discounting model. 151 00:06:31,140 --> 00:06:35,405 So this is evidence from using real-world credit card usage. 152 00:06:35,405 --> 00:06:36,780 So essentially, there's a company 153 00:06:36,780 --> 00:06:40,360 who has a bunch of data on credit card usages, 154 00:06:40,360 --> 00:06:42,360 and as you might be aware, credit card companies 155 00:06:42,360 --> 00:06:44,443 send out all sorts of mailers about getting people 156 00:06:44,443 --> 00:06:46,080 to sign up for a credit card. 157 00:06:46,080 --> 00:06:49,830 In particular, once you're out of school and gainfully 158 00:06:49,830 --> 00:06:52,920 employed, you get all sorts of mailings 159 00:06:52,920 --> 00:06:55,365 and so on asking you to sign up for credit cards. 160 00:06:58,470 --> 00:07:01,890 Now, in this particular example, the company 161 00:07:01,890 --> 00:07:04,680 mailed out randomized credit card offers 162 00:07:04,680 --> 00:07:06,060 of the following types. 163 00:07:06,060 --> 00:07:07,620 There was like a standard credit card 164 00:07:07,620 --> 00:07:13,500 offer, which was a 6.9% interest rate for six months and 16% 165 00:07:13,500 --> 00:07:16,530 afterwards, so there's already, in some sense, a teaser rate 166 00:07:16,530 --> 00:07:18,810 built-in in the sense it's starting to be looks 167 00:07:18,810 --> 00:07:20,140 like it's really cheap. 168 00:07:20,140 --> 00:07:23,880 You might sort of think, oh, I'm going to just bar for a month, 169 00:07:23,880 --> 00:07:25,762 and then eventually, it gets more expensive. 170 00:07:25,762 --> 00:07:27,220 The hope of the company, of course, 171 00:07:27,220 --> 00:07:29,610 is that you keep borrowing from them once it actually 172 00:07:29,610 --> 00:07:30,900 gets more expensive. 173 00:07:30,900 --> 00:07:33,090 And then there's two randomized conditions. 174 00:07:33,090 --> 00:07:36,780 There's what they call the teaser deal, which 175 00:07:36,780 --> 00:07:41,000 is 4.9% followed by 16%, and then there's 176 00:07:41,000 --> 00:07:42,750 what they call the post-teaser deal, which 177 00:07:42,750 --> 00:07:46,080 is 6.9% followed by 4%. 178 00:07:46,080 --> 00:07:51,570 Notice that if you borrow a given amount for quite a while 179 00:07:51,570 --> 00:07:54,540 and that these two conditions are pretty much similar, 180 00:07:54,540 --> 00:07:56,640 depending a little bit how long you borrow, 181 00:07:56,640 --> 00:07:59,040 if you borrow for something like a year, 182 00:07:59,040 --> 00:08:00,892 then it's about similar or the same. 183 00:08:00,892 --> 00:08:02,350 If you borrow for more than a year, 184 00:08:02,350 --> 00:08:05,280 then you know option three is probably better 185 00:08:05,280 --> 00:08:09,090 than option two for you for a given amount that you borrow. 186 00:08:09,090 --> 00:08:11,880 But what the company then has is response rate 187 00:08:11,880 --> 00:08:17,310 and people's 20-month borrowing history for individuals 188 00:08:17,310 --> 00:08:18,205 who take up the card. 189 00:08:18,205 --> 00:08:19,830 Now, these are kind of selected, so you 190 00:08:19,830 --> 00:08:22,080 want to be a little bit careful how to interpret that. 191 00:08:22,080 --> 00:08:25,452 But the response rates surely are the right thing 192 00:08:25,452 --> 00:08:27,660 to look at because they'd be randomized across people 193 00:08:27,660 --> 00:08:29,040 who read which letter. 194 00:08:29,040 --> 00:08:31,890 What they find is that the response rate 195 00:08:31,890 --> 00:08:36,000 is about 2.5 times larger increase 196 00:08:36,000 --> 00:08:39,240 in take-up in the second relative to the third group. 197 00:08:39,240 --> 00:08:41,580 Now, given subsequent borrowing behavior-- and that's 198 00:08:41,580 --> 00:08:42,480 kind of endogenous. 199 00:08:42,480 --> 00:08:44,397 You want to be a little bit careful with that. 200 00:08:44,397 --> 00:08:46,830 But essentially, if you look at people's behavior, 201 00:08:46,830 --> 00:08:49,687 the latter is a much better deal. 202 00:08:49,687 --> 00:08:51,270 So you'd much rather have-- as I said, 203 00:08:51,270 --> 00:08:55,590 you'd much rather have like deal number three compared 204 00:08:55,590 --> 00:08:56,460 to deal number two. 205 00:08:56,460 --> 00:08:59,160 And particularly, if you borrow for, say, two years 206 00:08:59,160 --> 00:09:02,180 like $1,000, you're going to pay much more interest 207 00:09:02,180 --> 00:09:06,370 in deal number two compared to number three. 208 00:09:06,370 --> 00:09:08,677 Yes? 209 00:09:08,677 --> 00:09:11,010 Now how do we think about this from the quasi-hyperbolic 210 00:09:11,010 --> 00:09:11,790 discounting model? 211 00:09:11,790 --> 00:09:13,123 How do we explain this behavior? 212 00:09:25,690 --> 00:09:26,501 Yes? 213 00:09:26,501 --> 00:09:35,030 AUDIENCE: [INAUDIBLE] the one with the [INAUDIBLE].. 214 00:09:44,588 --> 00:09:45,880 FRANK SCHILBACH: Yeah, exactly. 215 00:09:45,880 --> 00:09:47,600 So let me just repeat what you said. 216 00:09:47,600 --> 00:09:50,200 So you said, people think they will not borrow in the future, 217 00:09:50,200 --> 00:09:51,820 or they're like, I'm going to borrow for a while. 218 00:09:51,820 --> 00:09:52,840 I'm going to borrow for a few months. 219 00:09:52,840 --> 00:09:54,310 I just need some money right now. 220 00:09:54,310 --> 00:09:56,000 In the future, surely I'll be patient. 221 00:09:56,000 --> 00:09:58,840 I'll pay everything back, and so it doesn't really 222 00:09:58,840 --> 00:10:01,540 matter what the interest rate is after six months. 223 00:10:01,540 --> 00:10:03,880 Therefore I'm just going to optimize 224 00:10:03,880 --> 00:10:08,120 according to the six-month period that I see. 225 00:10:08,120 --> 00:10:11,668 The third deal is cheaper than the second, 226 00:10:11,668 --> 00:10:12,710 so let me just pick that. 227 00:10:12,710 --> 00:10:15,430 And then so, surprise, people actually then keep borrowing, 228 00:10:15,430 --> 00:10:19,360 and it ends up being a lot more expensive than it is. 229 00:10:19,360 --> 00:10:20,600 That's exactly right. 230 00:10:20,600 --> 00:10:22,610 That's what I have here as well. 231 00:10:22,610 --> 00:10:24,430 Now, if people are sophisticated, why would 232 00:10:24,430 --> 00:10:27,790 they choose option two? 233 00:10:32,250 --> 00:10:33,200 Yes? 234 00:10:33,200 --> 00:10:41,187 AUDIENCE: [INAUDIBLE] 235 00:10:41,187 --> 00:10:43,020 FRANK SCHILBACH: So there's sort of the idea 236 00:10:43,020 --> 00:10:45,330 that, essentially, if you don't like your future self 237 00:10:45,330 --> 00:10:48,450 to borrow, you might want to actually choose a high interest 238 00:10:48,450 --> 00:10:51,270 rate to protect you from doing that. 239 00:10:51,270 --> 00:10:54,060 Now, that's a very sophisticated type of behavior. 240 00:10:54,060 --> 00:10:55,540 My guess is, at the end of the day, 241 00:10:55,540 --> 00:10:58,208 that will not be particularly successful 242 00:10:58,208 --> 00:11:00,750 because, at the end of the day, the 2% difference is probably 243 00:11:00,750 --> 00:11:01,650 not making a huge difference. 244 00:11:01,650 --> 00:11:04,230 But in principle, the data are sort of consistent with that. 245 00:11:04,230 --> 00:11:06,150 We cannot rule this out. 246 00:11:06,150 --> 00:11:08,310 Now, there's other issues with like substitution 247 00:11:08,310 --> 00:11:12,360 to other cards and so on, so you might sort of-- 248 00:11:12,360 --> 00:11:14,160 that might backfire in various other ways, 249 00:11:14,160 --> 00:11:18,090 but in principle, at least, that's exactly right. 250 00:11:18,090 --> 00:11:20,340 Now, what evidence can be collect to separate 251 00:11:20,340 --> 00:11:21,170 these stories? 252 00:11:21,170 --> 00:11:23,940 I show you this data from the paper that they have. 253 00:11:23,940 --> 00:11:26,730 Now, as I'm telling you, I can't really 254 00:11:26,730 --> 00:11:29,070 disentangle these stories because there's no other data 255 00:11:29,070 --> 00:11:29,768 available. 256 00:11:29,768 --> 00:11:31,560 But what other data would you want to know, 257 00:11:31,560 --> 00:11:33,560 would you like to collect to sort of disentangle 258 00:11:33,560 --> 00:11:35,940 from these two stories or disentangle these two stories? 259 00:11:39,750 --> 00:11:40,567 Yes? 260 00:11:40,567 --> 00:11:43,770 AUDIENCE: Whether people continued borrowing. 261 00:11:43,770 --> 00:11:46,710 FRANK SCHILBACH: So you could look at their boring behavior, 262 00:11:46,710 --> 00:11:48,710 and then what would you conclude, or what data-- 263 00:11:48,710 --> 00:11:50,845 what would you-- 264 00:11:50,845 --> 00:11:52,470 AUDIENCE: So if they stopped borrowing, 265 00:11:52,470 --> 00:11:57,145 that would imply they are sophisticated and they continue 266 00:11:57,145 --> 00:11:58,067 [INAUDIBLE]. 267 00:12:01,960 --> 00:12:03,880 FRANK SCHILBACH: Yes. 268 00:12:03,880 --> 00:12:05,920 I think a little inconclusive in some sense, 269 00:12:05,920 --> 00:12:08,740 but yes, I think that's right. 270 00:12:08,740 --> 00:12:09,970 What else can we collect? 271 00:12:13,890 --> 00:12:14,740 Yes? 272 00:12:14,740 --> 00:12:15,984 AUDIENCE: [INAUDIBLE]. 273 00:12:19,540 --> 00:12:20,790 FRANK SCHILBACH: Yes, exactly. 274 00:12:20,790 --> 00:12:23,730 So if you ask people, what do you think you're going to do, 275 00:12:23,730 --> 00:12:25,060 that would be helpful. 276 00:12:25,060 --> 00:12:26,910 And then in some ways, it will also 277 00:12:26,910 --> 00:12:29,760 be nice to have some form of like demand for commitment. 278 00:12:29,760 --> 00:12:32,310 We don't quite have this here because neither 279 00:12:32,310 --> 00:12:35,520 of these options is actually dominated. 280 00:12:35,520 --> 00:12:37,600 You could sort of tell stories either way. 281 00:12:37,600 --> 00:12:40,170 But once you have a dominated option, 282 00:12:40,170 --> 00:12:42,540 the naive person would not pick that anymore. 283 00:12:42,540 --> 00:12:45,180 I think it's also right that looking at people's borrowing 284 00:12:45,180 --> 00:12:47,530 behavior-- we can also learn from that in some sense. 285 00:12:47,530 --> 00:12:51,810 It can't be that you keep doing stuff that's bad for you 286 00:12:51,810 --> 00:12:53,360 if you're really sophisticated. 287 00:12:57,090 --> 00:12:59,090 Sorry. 288 00:12:59,090 --> 00:13:03,568 So two things I think that you can potentially 289 00:13:03,568 --> 00:13:05,360 collect clearcut evidence to rule stuff out 290 00:13:05,360 --> 00:13:11,663 is, one, collect beliefs about for borrowing behavior. 291 00:13:11,663 --> 00:13:13,580 So there's one very nice paper, a recent paper 292 00:13:13,580 --> 00:13:16,520 by Hunt Allcott, and Dmitry Taubinsky, 293 00:13:16,520 --> 00:13:19,820 and co-authors that looks at payday borrowers. 294 00:13:19,820 --> 00:13:22,478 This is like payday lending at really extremely high interest 295 00:13:22,478 --> 00:13:24,020 rates where people essentially borrow 296 00:13:24,020 --> 00:13:26,090 against their future paycheck, and they keep 297 00:13:26,090 --> 00:13:27,425 doing this over and over again. 298 00:13:27,425 --> 00:13:29,300 And what they do there-- they actually elicit 299 00:13:29,300 --> 00:13:30,963 people's beliefs and say, what do 300 00:13:30,963 --> 00:13:32,630 you think is the probability that you're 301 00:13:32,630 --> 00:13:34,200 going to borrow again? 302 00:13:34,200 --> 00:13:36,800 You can ask several people and then, essentially, just 303 00:13:36,800 --> 00:13:38,750 look at what are the actual beliefs-- 304 00:13:38,750 --> 00:13:41,000 sorry, what are the actual borrowing rates, about what 305 00:13:41,000 --> 00:13:43,730 fraction of people are actually borrowing again compared 306 00:13:43,730 --> 00:13:46,340 to the perceived fraction. 307 00:13:46,340 --> 00:13:48,440 And what they find, in fact, is payday borrowers 308 00:13:48,440 --> 00:13:50,217 seem to be quite sophisticated. 309 00:13:50,217 --> 00:13:51,800 So these are people who actually know, 310 00:13:51,800 --> 00:13:55,860 I'm going to like borrow again in two weeks from now 311 00:13:55,860 --> 00:13:58,440 and so on, and I'm going to roll it over and over again. 312 00:13:58,440 --> 00:14:02,553 Now, they also find that that is more common among repeat 313 00:14:02,553 --> 00:14:04,220 borrowers, so like somebody who has like 314 00:14:04,220 --> 00:14:05,960 rolled his over five times-- 315 00:14:05,960 --> 00:14:08,717 they actually don't have a lot of illusions of what they're 316 00:14:08,717 --> 00:14:09,800 going to do in the future. 317 00:14:09,800 --> 00:14:11,870 They know they're going to roll it over again. 318 00:14:11,870 --> 00:14:14,453 When you look at people who are like first-time or second-time 319 00:14:14,453 --> 00:14:16,358 borrowers, things are different. 320 00:14:16,358 --> 00:14:17,900 There people tend to think, I'm going 321 00:14:17,900 --> 00:14:20,840 to do this only once or twice, and then 322 00:14:20,840 --> 00:14:24,710 they sort of surprised themselves by borrowing again. 323 00:14:24,710 --> 00:14:26,210 That's consistent with some story 324 00:14:26,210 --> 00:14:30,202 of learning in a sense of they can't really necessarily nail 325 00:14:30,202 --> 00:14:31,910 that, but I think they find, essentially, 326 00:14:31,910 --> 00:14:35,300 that or they interpret their evidence as people do 327 00:14:35,300 --> 00:14:38,280 seem to be learning over time. 328 00:14:38,280 --> 00:14:40,430 They also offer some form of commitment device. 329 00:14:40,430 --> 00:14:45,590 Essentially, they offer him to say, if you don't borrow again, 330 00:14:45,590 --> 00:14:48,020 we'll pay you $100 or versions of that. 331 00:14:48,020 --> 00:14:50,540 Or you have to pay or you can forego money 332 00:14:50,540 --> 00:14:54,560 if you commit to not borrowing again. 333 00:14:54,560 --> 00:14:57,020 They also find some evidence of that sort 334 00:14:57,020 --> 00:14:59,540 of suggesting that, again, these payday borrowers seem 335 00:14:59,540 --> 00:15:01,400 to be quite sophisticated. 336 00:15:01,400 --> 00:15:02,990 Now, of course, it's a different type 337 00:15:02,990 --> 00:15:05,900 of sample, a different type of population. 338 00:15:05,900 --> 00:15:08,010 So credit card people, particularly who 339 00:15:08,010 --> 00:15:10,280 got credit card mailings, in particular, people 340 00:15:10,280 --> 00:15:12,560 who just started their jobs and so on and so forth, 341 00:15:12,560 --> 00:15:15,170 might be not as sophisticated. 342 00:15:15,170 --> 00:15:18,485 Now, if you are a credit card company, what would you do? 343 00:15:18,485 --> 00:15:20,360 How would you react to that kind of behavior? 344 00:15:20,360 --> 00:15:22,318 Suppose you believed what I showed you earlier, 345 00:15:22,318 --> 00:15:25,010 the evidence that I just showed you about from the Ausubel 346 00:15:25,010 --> 00:15:27,380 paper in terms of people seem to be 347 00:15:27,380 --> 00:15:31,340 either naive or sophisticated quasi-hyperbolic discounters. 348 00:15:31,340 --> 00:15:33,020 What do you do as a credit card company? 349 00:15:33,020 --> 00:15:34,728 Whom do you target, and how do you do it? 350 00:15:52,440 --> 00:15:53,160 Yes? 351 00:15:53,160 --> 00:15:55,920 AUDIENCE: Perhaps you could map some kind of characteristic 352 00:15:55,920 --> 00:15:59,610 of the person, maybe demographics like education 353 00:15:59,610 --> 00:16:02,640 or socio-economic characteristics 354 00:16:02,640 --> 00:16:05,310 with possible naive/sophisticated 355 00:16:05,310 --> 00:16:10,015 and then almost price [INAUDIBLE] 356 00:16:10,015 --> 00:16:12,651 target each person based on those characteristics. 357 00:16:12,651 --> 00:16:17,110 So the sophisticated person would have the lower teaser 358 00:16:17,110 --> 00:16:20,094 rate and then the higher interest rate towards the end, 359 00:16:20,094 --> 00:16:23,198 and you could do the same for those who are naive. 360 00:16:23,198 --> 00:16:24,240 FRANK SCHILBACH: Exactly. 361 00:16:24,240 --> 00:16:26,310 So what you kind of want to know in some sense-- 362 00:16:26,310 --> 00:16:28,530 you want to know who are the naive people, who 363 00:16:28,530 --> 00:16:32,105 are the people that we can essentially exploit 364 00:16:32,105 --> 00:16:34,230 because the naive person is the person who is going 365 00:16:34,230 --> 00:16:37,140 to be the person who signs up and then ends up 366 00:16:37,140 --> 00:16:39,390 paying really high interest rates surprisingly. 367 00:16:39,390 --> 00:16:41,280 So you can essentially figure out 368 00:16:41,280 --> 00:16:43,440 some ways of making a lot of money from them. 369 00:16:43,440 --> 00:16:46,380 The exponential discounters or even the sophisticated people-- 370 00:16:46,380 --> 00:16:49,050 they will not be helpful for you because either they 371 00:16:49,050 --> 00:16:51,420 will, essentially, just optimize in some sense. 372 00:16:51,420 --> 00:16:52,620 They use some offers. 373 00:16:52,620 --> 00:16:54,540 They get some bonus points or whatever 374 00:16:54,540 --> 00:16:56,245 for signing up in the first place 375 00:16:56,245 --> 00:16:57,870 and then essentially just drop the card 376 00:16:57,870 --> 00:16:59,190 and not use it anymore. 377 00:16:59,190 --> 00:17:01,230 Even a sophisticated person you can't really 378 00:17:01,230 --> 00:17:03,630 exploit that much because if you could, 379 00:17:03,630 --> 00:17:06,390 the sophisticated person would understand that and then 380 00:17:06,390 --> 00:17:08,849 would just not sign up in the first place. 381 00:17:08,849 --> 00:17:12,750 And this is kind of the answer to what Lauren just said about, 382 00:17:12,750 --> 00:17:15,359 if somebody if you can make a lot of money from that person, 383 00:17:15,359 --> 00:17:17,400 it can't be that person is perfectly 384 00:17:17,400 --> 00:17:19,890 sophisticated because then, otherwise, the person 385 00:17:19,890 --> 00:17:22,359 would never sign up in the first place. 386 00:17:22,359 --> 00:17:25,145 Now, credit card companies-- there's 387 00:17:25,145 --> 00:17:27,270 some evidence, at least, that credit card companies 388 00:17:27,270 --> 00:17:29,820 do exactly what you just said, which essentially 389 00:17:29,820 --> 00:17:32,190 is-- so they don't have a lot of information of people 390 00:17:32,190 --> 00:17:33,600 that they actually mail to. 391 00:17:33,600 --> 00:17:37,560 The one thing that they have is zip code-level information, 392 00:17:37,560 --> 00:17:40,320 very often about education and some other characteristics 393 00:17:40,320 --> 00:17:41,980 from some other data sets. 394 00:17:41,980 --> 00:17:44,160 So there's a very nice paper by Ru 395 00:17:44,160 --> 00:17:48,990 and Antoinette Schoar at Sloan, at MIT that looks at, 396 00:17:48,990 --> 00:17:52,980 analyzes a detailed data set of mailed credit card offers. 397 00:17:52,980 --> 00:17:54,480 So essentially, there are people who 398 00:17:54,480 --> 00:17:56,397 are paid to collect all the mailings that they 399 00:17:56,397 --> 00:18:00,930 get from different companies, and just collect them, and have 400 00:18:00,930 --> 00:18:01,980 them analyzed. 401 00:18:01,980 --> 00:18:05,140 And there's sort of a data set doing exactly that. 402 00:18:05,140 --> 00:18:07,200 And so what they did is they analyzed 403 00:18:07,200 --> 00:18:10,050 this data set specifically and looked 404 00:18:10,050 --> 00:18:13,560 at how some of the details of these offers 405 00:18:13,560 --> 00:18:17,140 relate to things like education, zip code. 406 00:18:17,140 --> 00:18:19,020 So you look at sort of education-level data, 407 00:18:19,020 --> 00:18:21,420 like where do the highly-educated people, 408 00:18:21,420 --> 00:18:23,760 and where do other people live, and what 409 00:18:23,760 --> 00:18:26,370 they find is that education levels are systematically 410 00:18:26,370 --> 00:18:30,420 related to things like backloaded and hidden fees, 411 00:18:30,420 --> 00:18:34,440 teasers, stuff that's written in small font, 412 00:18:34,440 --> 00:18:36,840 having more photos, color print, and so on and so forth, 413 00:18:36,840 --> 00:18:38,310 essentially a bunch of stuff that's 414 00:18:38,310 --> 00:18:43,385 trying to trick people into things that are bad for them. 415 00:18:43,385 --> 00:18:44,760 So what the credit card companies 416 00:18:44,760 --> 00:18:46,718 are doing-- they're essentially sort of saying, 417 00:18:46,718 --> 00:18:48,450 here's some stuff that is exploiting 418 00:18:48,450 --> 00:18:51,210 potentially naive customers. 419 00:18:51,210 --> 00:18:55,020 Now, you only want to send that kind of stuff to naive people 420 00:18:55,020 --> 00:18:59,010 or mostly to naive or less sophisticated 421 00:18:59,010 --> 00:19:02,740 customers, the reason being that, A, it's costly to do it. 422 00:19:02,740 --> 00:19:05,860 So if you send it to people who are exponential discounters 423 00:19:05,860 --> 00:19:08,520 or who are more sophisticated, you're just wasting money, 424 00:19:08,520 --> 00:19:11,160 or they might actually exploit your offer 425 00:19:11,160 --> 00:19:15,125 by optimizing and getting good deals out of it in some ways. 426 00:19:15,125 --> 00:19:16,500 So what do you kind of want to do 427 00:19:16,500 --> 00:19:18,870 is you want to focus on the naive customers 428 00:19:18,870 --> 00:19:21,540 and sort of exploit them as much as you can. 429 00:19:21,540 --> 00:19:23,988 Now, actually, it's not that you want to exploit them 430 00:19:23,988 --> 00:19:24,780 as much as you can. 431 00:19:24,780 --> 00:19:25,710 There are some constraints. 432 00:19:25,710 --> 00:19:26,960 What are the constraints here? 433 00:19:26,960 --> 00:19:30,100 What is a credit card company, in some ways, worried about? 434 00:19:30,100 --> 00:19:30,600 Yes? 435 00:19:30,600 --> 00:19:32,892 AUDIENCE: They want to make sure that they can actually 436 00:19:32,892 --> 00:19:33,920 pay the [INAUDIBLE]. 437 00:19:33,920 --> 00:19:34,962 FRANK SCHILBACH: Exactly. 438 00:19:34,962 --> 00:19:36,890 So this is a pretty cynical view of the world, 439 00:19:36,890 --> 00:19:40,430 but I think it's a pretty correct one in some ways, which 440 00:19:40,430 --> 00:19:43,520 is, the credit card company wants to exploit people 441 00:19:43,520 --> 00:19:46,067 but only to the extent that they can actually repay. 442 00:19:46,067 --> 00:19:47,900 The last thing the credit card company wants 443 00:19:47,900 --> 00:19:51,050 is for people to default, not because a credit card company 444 00:19:51,050 --> 00:19:53,300 actually cares about the people who might default 445 00:19:53,300 --> 00:19:54,380 but rather because the credit card 446 00:19:54,380 --> 00:19:56,172 company has, then, trouble actually getting 447 00:19:56,172 --> 00:19:58,183 their money back. 448 00:19:58,183 --> 00:19:59,600 So there's, essentially, some form 449 00:19:59,600 --> 00:20:01,700 of a participation constraint. 450 00:20:01,700 --> 00:20:06,470 And so what Ru and Schoar find is that, at times-- and maybe 451 00:20:06,470 --> 00:20:09,650 this is suggestive evidence, but it's pretty neat in some way. 452 00:20:09,650 --> 00:20:15,650 At times when unemployment insurance from the government 453 00:20:15,650 --> 00:20:17,780 increases, essentially, as more people have 454 00:20:17,780 --> 00:20:21,170 more money available, a lot of these fees, et cetera 455 00:20:21,170 --> 00:20:24,140 tend to be higher, and so one way 456 00:20:24,140 --> 00:20:26,060 to interpret that is, essentially, there's 457 00:20:26,060 --> 00:20:28,520 a participation constrained if you want an optimization 458 00:20:28,520 --> 00:20:29,020 problem. 459 00:20:29,020 --> 00:20:31,190 That is to say, they try to exploit the customers, 460 00:20:31,190 --> 00:20:33,075 but it's subject to the constraint 461 00:20:33,075 --> 00:20:34,700 that the person doesn't default or just 462 00:20:34,700 --> 00:20:36,410 doesn't use the offers at all. 463 00:20:36,410 --> 00:20:38,630 Now, that participation constraint 464 00:20:38,630 --> 00:20:42,530 gets relaxed now that people, due to unemployment 465 00:20:42,530 --> 00:20:47,190 and other types of insurance, have more money available, 466 00:20:47,190 --> 00:20:50,240 so essentially, you can exploit them more. 467 00:20:50,240 --> 00:20:52,970 This is sort of a pattern, I think, 468 00:20:52,970 --> 00:20:54,900 that's general for many cases. 469 00:20:54,900 --> 00:20:57,002 I think when you, for example, think about gyms, 470 00:20:57,002 --> 00:20:58,460 you might sort of say, well, people 471 00:20:58,460 --> 00:21:03,470 choose in gyms the offers that gyms give to people. 472 00:21:03,470 --> 00:21:06,127 In a way, they are meant to exploit customers. 473 00:21:06,127 --> 00:21:08,210 Or do you really want what I showed you last time? 474 00:21:08,210 --> 00:21:12,620 You really want to offer people these monthly fees or whatever? 475 00:21:12,620 --> 00:21:14,870 Like you want people to sign up but then actually 476 00:21:14,870 --> 00:21:16,310 not show up very much to your gym 477 00:21:16,310 --> 00:21:18,950 because you have a certain capacity. 478 00:21:18,950 --> 00:21:21,020 You'll then hit your gym capacity. 479 00:21:21,020 --> 00:21:23,330 But some people might say, well, why not just 480 00:21:23,330 --> 00:21:26,060 ask people to give them even more and more expensive 481 00:21:26,060 --> 00:21:29,180 contracts that they sign up so they really sort of-- to screw 482 00:21:29,180 --> 00:21:30,920 them over more and more? 483 00:21:30,920 --> 00:21:34,190 The problem is there's again a participation constraint. 484 00:21:34,190 --> 00:21:36,890 If the person realizes, I'm spending way too much money 485 00:21:36,890 --> 00:21:39,140 on this and I'm not actually going to the gym 486 00:21:39,140 --> 00:21:41,685 ever, that person is just going to quit. 487 00:21:41,685 --> 00:21:42,560 So you kind of want-- 488 00:21:42,560 --> 00:21:44,477 there's this game where, essentially, you want 489 00:21:44,477 --> 00:21:45,920 your customer to be engaged. 490 00:21:45,920 --> 00:21:48,740 You want your customer to come back sometimes and feel 491 00:21:48,740 --> 00:21:50,870 good about going to the gym sometimes and not 492 00:21:50,870 --> 00:21:52,940 feel terrible about not coming, so that's 493 00:21:52,940 --> 00:21:55,400 why the gyms and other companies are always 494 00:21:55,400 --> 00:21:56,750 trying to be very motivating. 495 00:21:56,750 --> 00:21:58,280 We choose the best contract for you. 496 00:21:58,280 --> 00:22:00,500 We're trying to help you with this and that. 497 00:22:00,500 --> 00:22:03,080 But really, it's, in some sense, making the appearance 498 00:22:03,080 --> 00:22:05,570 of trying to help the customers, keeping them happy 499 00:22:05,570 --> 00:22:07,575 while you can exploit them at the same time. 500 00:22:07,575 --> 00:22:09,450 Now, that's a very cynical view of the world, 501 00:22:09,450 --> 00:22:11,000 and I'm sure there are some companies who are actually 502 00:22:11,000 --> 00:22:12,205 care about their customers. 503 00:22:12,205 --> 00:22:13,580 But in some sense, I think it can 504 00:22:13,580 --> 00:22:17,193 explain quite a few patterns that we see in the world. 505 00:22:17,193 --> 00:22:18,110 Any questions on this? 506 00:22:21,190 --> 00:22:21,790 Or comments? 507 00:22:25,670 --> 00:22:26,540 OK. 508 00:22:26,540 --> 00:22:27,680 So then next is-- 509 00:22:27,680 --> 00:22:31,040 and this is sort of, again, almost like a classic paper now 510 00:22:31,040 --> 00:22:32,900 because it's from quite a while ago. 511 00:22:32,900 --> 00:22:35,570 It's a paper by Nava Ashraf and colleagues in the Philippines. 512 00:22:35,570 --> 00:22:38,330 And these are clients of a Philippine bank, 513 00:22:38,330 --> 00:22:44,870 and they are randomized into three different groups 514 00:22:44,870 --> 00:22:49,940 to be receiving three different kinds of savings product. 515 00:22:49,940 --> 00:22:53,620 The first one is essentially like a very simple-- 516 00:22:53,620 --> 00:22:54,120 sorry. 517 00:22:54,120 --> 00:22:56,210 The first one, number three, is like a control treatment, 518 00:22:56,210 --> 00:22:58,080 or it's just a standard savings account. 519 00:22:58,080 --> 00:22:59,660 Here's an account that you can open, 520 00:22:59,660 --> 00:23:03,170 and you're going to get some interest rate 521 00:23:03,170 --> 00:23:05,580 for saving in this account. 522 00:23:05,580 --> 00:23:07,430 The second one is like a marketing treatment 523 00:23:07,430 --> 00:23:10,010 which essentially is encouraging people to sign up. 524 00:23:10,010 --> 00:23:11,930 Otherwise, it's like the same, essentially 525 00:23:11,930 --> 00:23:14,353 trying to get people to do it. 526 00:23:14,353 --> 00:23:15,770 Number one-- and this is, I guess, 527 00:23:15,770 --> 00:23:17,312 the treatment of interest-- is what's 528 00:23:17,312 --> 00:23:19,790 called the SEED Treatment, which is encouraging people 529 00:23:19,790 --> 00:23:21,770 to sign up, plus a commitment offer, which 530 00:23:21,770 --> 00:23:25,220 is clients can restrict their access to the deposits 531 00:23:25,220 --> 00:23:26,970 upon opening the account. 532 00:23:26,970 --> 00:23:30,330 So what you can essentially do is you can set some goals. 533 00:23:30,330 --> 00:23:32,780 These goals are either date-based or amount-based. 534 00:23:32,780 --> 00:23:37,370 So you can say, I want to save $1,000 or some amount, 535 00:23:37,370 --> 00:23:39,710 and until I've reached that amount I cannot withdraw any 536 00:23:39,710 --> 00:23:40,490 money. 537 00:23:40,490 --> 00:23:42,110 Alternatively, you can say, I would 538 00:23:42,110 --> 00:23:45,710 like to save until my husband's or my wife's 539 00:23:45,710 --> 00:23:47,390 birthday or some other-- 540 00:23:47,390 --> 00:23:48,860 child's graduation and whatever you 541 00:23:48,860 --> 00:23:50,960 want to save for, Christmas, et cetera, 542 00:23:50,960 --> 00:23:54,120 and only then you can withdraw the money. 543 00:23:54,120 --> 00:23:56,270 As you can see, most people seem to be choosing 544 00:23:56,270 --> 00:23:58,820 date-based types of goals, but others also 545 00:23:58,820 --> 00:24:01,310 have a amount-based. 546 00:24:01,310 --> 00:24:04,220 You can see about 70% here are date-based, 547 00:24:04,220 --> 00:24:06,800 and 30% are amount-based. 548 00:24:06,800 --> 00:24:08,270 Now, what they find is, A, there's 549 00:24:08,270 --> 00:24:10,220 quite a bit of take-up of this account, 550 00:24:10,220 --> 00:24:12,380 and again, when you think about this, 551 00:24:12,380 --> 00:24:14,300 there's not a good reason to take up 552 00:24:14,300 --> 00:24:19,560 this account unless you want to restrict your future self. 553 00:24:19,560 --> 00:24:24,200 So there's some evidence of demand for commitment. 554 00:24:24,200 --> 00:24:27,500 Second, offering just kinds of account 555 00:24:27,500 --> 00:24:30,300 increases people's savings, so on average, people save more. 556 00:24:30,300 --> 00:24:31,160 I'm going to tell you a little bit 557 00:24:31,160 --> 00:24:32,452 about the distribution of that. 558 00:24:32,452 --> 00:24:34,620 Is actually everybody better off from doing that? 559 00:24:34,620 --> 00:24:36,330 But on average, people do save more, 560 00:24:36,330 --> 00:24:37,940 so essentially, it's kind of similar to the evidence 561 00:24:37,940 --> 00:24:39,140 that I showed you earlier. 562 00:24:39,140 --> 00:24:41,030 Commitment devices, at least on average, 563 00:24:41,030 --> 00:24:44,390 can improve outcomes or affect outcomes in some way. 564 00:24:44,390 --> 00:24:45,770 And then number three-- 565 00:24:45,770 --> 00:24:47,270 and that's a little bit suggestive-- 566 00:24:47,270 --> 00:24:49,957 but essentially saying, if you look at money now versus later, 567 00:24:49,957 --> 00:24:52,040 a question that I showed you earlier in the class, 568 00:24:52,040 --> 00:24:55,490 they actually do predict commitment take-up consistent 569 00:24:55,490 --> 00:24:56,900 with some form of-- 570 00:24:56,900 --> 00:24:58,850 this is really about time preferences 571 00:24:58,850 --> 00:25:01,520 as opposed to maybe other things like demand effects, 572 00:25:01,520 --> 00:25:03,725 social-desirability bias, essentially 573 00:25:03,725 --> 00:25:06,350 customers maybe just wanting to make the bank happy in some way 574 00:25:06,350 --> 00:25:07,880 because they feel compelled to do it 575 00:25:07,880 --> 00:25:10,850 because they get some marketing treatment and a particularly 576 00:25:10,850 --> 00:25:12,300 strong one. 577 00:25:12,300 --> 00:25:13,430 Any questions on this? 578 00:25:13,430 --> 00:25:13,930 Yeah? 579 00:25:13,930 --> 00:25:14,900 AUDIENCE: [INAUDIBLE]? 580 00:25:22,190 --> 00:25:26,040 FRANK SCHILBACH: No, I think it's taking it-- 581 00:25:26,040 --> 00:25:28,500 I think it's taking it up and putting some money in it. 582 00:25:28,500 --> 00:25:31,010 Yeah, the take-up is put in like, 583 00:25:31,010 --> 00:25:33,890 once you put in some money, it's counted as using or taking up 584 00:25:33,890 --> 00:25:34,753 the account. 585 00:25:34,753 --> 00:25:36,170 Otherwise, it would be like, yeah, 586 00:25:36,170 --> 00:25:39,110 that's not really anything now, essentially restricting 587 00:25:39,110 --> 00:25:41,690 your choices in some way by putting in some money 588 00:25:41,690 --> 00:25:42,460 into this account. 589 00:25:45,980 --> 00:25:46,790 OK. 590 00:25:46,790 --> 00:25:49,610 So next is some work that I've done in India on alcohol 591 00:25:49,610 --> 00:25:53,750 consumption, and this is a very common issue 592 00:25:53,750 --> 00:25:57,050 among low-income workers, in South India, in Tamil Nadu, 593 00:25:57,050 --> 00:25:59,370 and Kerala, and other places. 594 00:25:59,370 --> 00:26:03,300 It's also an issue in other countries, in Uganda and so on. 595 00:26:03,300 --> 00:26:05,810 So large fractions of low-income workers 596 00:26:05,810 --> 00:26:09,292 drink large amounts of alcohol, often daily. 597 00:26:09,292 --> 00:26:10,250 What do I mean by that? 598 00:26:10,250 --> 00:26:14,240 They drink something like five, six standard drinks per day. 599 00:26:14,240 --> 00:26:15,240 What's a standard drink? 600 00:26:15,240 --> 00:26:19,190 It's like a small beer or like a shot of liquor, so five, 601 00:26:19,190 --> 00:26:22,200 six standard drinks are a lot of alcohol. 602 00:26:22,200 --> 00:26:23,750 Now, they also spend large amounts 603 00:26:23,750 --> 00:26:26,060 of their money of their income on alcohol 604 00:26:26,060 --> 00:26:28,850 on the order of magnitude of 20%, 30%, 40%, 605 00:26:28,850 --> 00:26:32,510 so think of these as like families of five or six. 606 00:26:32,510 --> 00:26:35,480 Sorry, families of four or five often. 607 00:26:35,480 --> 00:26:39,020 What they earn is something like 300, 400 rupees per day. 608 00:26:39,020 --> 00:26:40,880 That's usually, often, the family income. 609 00:26:40,880 --> 00:26:43,100 Some earn a little bit more, but that's something 610 00:26:43,100 --> 00:26:44,780 like $5, $6, $7 per day. 611 00:26:44,780 --> 00:26:48,140 And often they spend $1 or $2 on alcohol. 612 00:26:48,140 --> 00:26:52,250 And so when you ask people, when you ask women, 613 00:26:52,250 --> 00:26:56,030 when you ask men about, how does alcohol affect your lives, 614 00:26:56,030 --> 00:26:58,280 you get sort of this uniform response of people saying 615 00:26:58,280 --> 00:27:01,700 like, well, if only liquor stores closed, 616 00:27:01,700 --> 00:27:04,130 my life would be so much better. 617 00:27:04,130 --> 00:27:06,800 I would love for liquor prices to be higher, 618 00:27:06,800 --> 00:27:09,800 and I would like to find some way to reduce my drinking. 619 00:27:09,800 --> 00:27:12,230 But I have trouble doing so. 620 00:27:12,230 --> 00:27:14,292 When you ask people, OK, now you would 621 00:27:14,292 --> 00:27:15,500 like to reduce your drinking. 622 00:27:15,500 --> 00:27:17,875 It would be great for you the liquor stores to be closed. 623 00:27:17,875 --> 00:27:20,990 Why is it difficult for you to reduce your drinking? 624 00:27:20,990 --> 00:27:22,940 The common response is always, A, addiction, 625 00:27:22,940 --> 00:27:24,560 which makes things tricky anyway, 626 00:27:24,560 --> 00:27:28,190 but B, people say they have physical pain, 627 00:27:28,190 --> 00:27:29,210 physical pain from work. 628 00:27:29,210 --> 00:27:33,170 Particularly, these are people with very low levels 629 00:27:33,170 --> 00:27:35,680 of education, so often they're illiterate and so on. 630 00:27:35,680 --> 00:27:38,500 So it's very hard for them to get any job or the like, 631 00:27:38,500 --> 00:27:40,190 so what instead they have to do is, 632 00:27:40,190 --> 00:27:42,500 essentially, do physical labor, hard physical labor. 633 00:27:42,500 --> 00:27:44,000 These are like construction workers, 634 00:27:44,000 --> 00:27:46,262 rickshaw drivers, sewage workers, 635 00:27:46,262 --> 00:27:47,220 and so on and so forth. 636 00:27:47,220 --> 00:27:49,730 This is really hard work, and they 637 00:27:49,730 --> 00:27:52,980 tend to have, then, lots of physical pain from their work. 638 00:27:52,980 --> 00:27:55,790 Now, why is it that, then, the pain exacerbates, 639 00:27:55,790 --> 00:27:57,627 potentially, people's self-control problems? 640 00:27:57,627 --> 00:27:58,960 Why is that making things worse? 641 00:28:09,730 --> 00:28:11,563 Yes? 642 00:28:11,563 --> 00:28:23,970 AUDIENCE: [INAUDIBLE] you might have been drinking for instance 643 00:28:23,970 --> 00:28:28,900 social reasons or alcoholism. 644 00:28:28,900 --> 00:28:32,162 Now you're also self-medicating for [INAUDIBLE].. 645 00:28:32,162 --> 00:28:33,120 FRANK SCHILBACH: Right. 646 00:28:33,120 --> 00:28:34,620 So what are you saying, essentially, 647 00:28:34,620 --> 00:28:38,302 is you can do self-medicating physical pain, in addition 648 00:28:38,302 --> 00:28:40,510 to maybe some other reason, social reasons and so on. 649 00:28:40,510 --> 00:28:43,800 It turns out, in this setting, at least, specifically, 650 00:28:43,800 --> 00:28:46,140 people tend to not drink so much for social reasons. 651 00:28:46,140 --> 00:28:47,670 There's some social function. 652 00:28:47,670 --> 00:28:51,510 There's like funerals or other sort of festivals where 653 00:28:51,510 --> 00:28:53,280 lots of people drink, but they're 654 00:28:53,280 --> 00:28:55,110 drinking-- in particular, in rickshaw 655 00:28:55,110 --> 00:28:58,020 is often people drink on their own, 656 00:28:58,020 --> 00:29:00,900 essentially to numb their pain in some ways. 657 00:29:00,900 --> 00:29:03,480 Now, when you think about self-control problems, 658 00:29:03,480 --> 00:29:06,240 often where self-control problems arise 659 00:29:06,240 --> 00:29:10,328 is when you have short-run benefits and long-run costs. 660 00:29:10,328 --> 00:29:12,120 Essentially, that's sort of the consumption 661 00:29:12,120 --> 00:29:13,710 good that people over consume. 662 00:29:13,710 --> 00:29:15,870 Now, when you think about physical pain, what does 663 00:29:15,870 --> 00:29:16,900 physical pain do? 664 00:29:16,900 --> 00:29:19,290 It very much increases the short-run benefits 665 00:29:19,290 --> 00:29:22,110 of drinking while keeping long-run costs or anything else 666 00:29:22,110 --> 00:29:22,947 constant. 667 00:29:22,947 --> 00:29:25,530 And so if you think about, you have lots of short-run benefits 668 00:29:25,530 --> 00:29:28,440 now of drinking because there's lots of pain that alcohol helps 669 00:29:28,440 --> 00:29:29,160 numbing-- 670 00:29:29,160 --> 00:29:32,390 and alcohol is a very powerful anesthetic as it turns out. 671 00:29:32,390 --> 00:29:34,140 So then, essentially, it sort of increases 672 00:29:34,140 --> 00:29:35,557 the short-run benefits, and that's 673 00:29:35,557 --> 00:29:38,190 sort of the short-run desire to drink. 674 00:29:38,190 --> 00:29:40,680 And then any long-run costs about health 675 00:29:40,680 --> 00:29:43,368 and so on and so forth are very much in the future, 676 00:29:43,368 --> 00:29:45,660 and that's where, essentially, the self-control problem 677 00:29:45,660 --> 00:29:48,510 potentially arises. 678 00:29:48,510 --> 00:29:51,870 Now, people not only drink a lot overall 679 00:29:51,870 --> 00:29:54,873 in Chennai and in surrounding areas, 680 00:29:54,873 --> 00:29:57,540 but they drink also quite a bit, in particular rickshaw drivers, 681 00:29:57,540 --> 00:29:58,600 during the day. 682 00:29:58,600 --> 00:30:01,680 So what I was doing there is essentially just randomly 683 00:30:01,680 --> 00:30:04,860 walking around in Chennai-- our RAs would do that-- 684 00:30:04,860 --> 00:30:07,500 walking around in Chennai and just breathalyzing people 685 00:30:07,500 --> 00:30:10,500 more or less randomly, offering them, essentially, would 686 00:30:10,500 --> 00:30:13,920 you like to do a breathalyzer test for some small amount, 687 00:30:13,920 --> 00:30:15,720 regardless of the result. 688 00:30:15,720 --> 00:30:18,210 And about half of these rickshaw drivers-- and these 689 00:30:18,210 --> 00:30:20,640 are all cycle rickshaws, so it's not particularly 690 00:30:20,640 --> 00:30:22,470 dangerous to do that. 691 00:30:22,470 --> 00:30:25,300 But about half of them are showing positive breathalyzer 692 00:30:25,300 --> 00:30:25,800 scores. 693 00:30:25,800 --> 00:30:28,383 So essentially, half of them are at least somewhat inebriated. 694 00:30:28,383 --> 00:30:32,100 About 20%, 25% are pretty drunk, as 695 00:30:32,100 --> 00:30:36,280 in their blood alcohol content is about 0.1 or higher, 696 00:30:36,280 --> 00:30:40,440 which is above the US legal driving limit overall. 697 00:30:40,440 --> 00:30:42,240 Now, this is during the regular work hours. 698 00:30:42,240 --> 00:30:44,910 These are people who are actually at work. 699 00:30:44,910 --> 00:30:47,485 Now, then sort of seeing that and sort 700 00:30:47,485 --> 00:30:49,860 of having people report about their self-control problem, 701 00:30:49,860 --> 00:30:51,235 I was then sort of wondering, OK, 702 00:30:51,235 --> 00:30:53,160 now, what kinds of commitment devices 703 00:30:53,160 --> 00:30:55,500 could be perhaps offered to people to reduce 704 00:30:55,500 --> 00:30:57,990 their drinking given that they tend to report 705 00:30:57,990 --> 00:31:00,480 self-control problems, or physical pain, 706 00:31:00,480 --> 00:31:02,850 or self-control problems related to physical pain 707 00:31:02,850 --> 00:31:05,940 are a key issue and preventing them 708 00:31:05,940 --> 00:31:08,640 from reducing their drinking if they desire to do so, 709 00:31:08,640 --> 00:31:09,240 of course? 710 00:31:09,240 --> 00:31:10,170 What could we do? 711 00:31:27,400 --> 00:31:28,243 Yes? 712 00:31:28,243 --> 00:31:29,618 AUDIENCE: We could just tell them 713 00:31:29,618 --> 00:31:31,951 you're going to randomly administer a breathalyzer test, 714 00:31:31,951 --> 00:31:36,743 and if they haven't been drinking [INAUDIBLE].. 715 00:31:36,743 --> 00:31:39,160 FRANK SCHILBACH: Yeah, so we could, essentially-- exactly. 716 00:31:39,160 --> 00:31:42,280 So you could essentially say, I'm going to try and find you. 717 00:31:42,280 --> 00:31:44,890 I'm going to call you sometimes randomly, 718 00:31:44,890 --> 00:31:47,860 and you get some reward if you're sober, 719 00:31:47,860 --> 00:31:49,960 so essentially incentivizing people to be sober. 720 00:31:49,960 --> 00:31:50,835 That's exactly right. 721 00:31:50,835 --> 00:31:51,850 You could do that. 722 00:31:51,850 --> 00:31:53,450 Now, is that a-- 723 00:31:53,450 --> 00:31:55,160 so what would be a commitment device? 724 00:31:55,160 --> 00:31:57,792 What would a commitment device for that look like? 725 00:31:57,792 --> 00:31:59,500 So what you just mentioned is essentially 726 00:31:59,500 --> 00:32:01,708 just providing people incentives, incentivizing them, 727 00:32:01,708 --> 00:32:03,460 and that could truly work. 728 00:32:03,460 --> 00:32:06,280 How would I show that these are, in fact, self-control problems 729 00:32:06,280 --> 00:32:11,318 or people have quasi-hyperbolic or similar preferences? 730 00:32:11,318 --> 00:32:13,708 AUDIENCE: So if they choose to enter into this study, 731 00:32:13,708 --> 00:32:17,054 they're [INAUDIBLE] for example maybe [INAUDIBLE] 732 00:32:17,054 --> 00:32:20,022 they've been drinking [INAUDIBLE].. 733 00:32:20,022 --> 00:32:20,980 FRANK SCHILBACH: Right. 734 00:32:20,980 --> 00:32:23,020 Essentially, you have to give people choices, 735 00:32:23,020 --> 00:32:25,430 and some choices are potentially dominating. 736 00:32:25,430 --> 00:32:26,600 So what does that mean? 737 00:32:26,600 --> 00:32:28,760 I'm giving you some choices. 738 00:32:28,760 --> 00:32:30,970 You can choose to either receive the incentives, 739 00:32:30,970 --> 00:32:33,640 and you only receive these incentives if you're sober. 740 00:32:33,640 --> 00:32:36,220 Alternatively, I'm going to just pay you some amount, 741 00:32:36,220 --> 00:32:39,340 and that amount, ideally, is like higher than the amount 742 00:32:39,340 --> 00:32:41,615 that you could get from these incentives. 743 00:32:41,615 --> 00:32:43,240 Now, then if you choose the incentives, 744 00:32:43,240 --> 00:32:44,210 what does that mean? 745 00:32:44,210 --> 00:32:46,043 Essentially, that means you want to increase 746 00:32:46,043 --> 00:32:47,470 your future prices of drinking. 747 00:32:47,470 --> 00:32:50,950 That would be, essentially, demand for commitment. 748 00:32:50,950 --> 00:32:52,930 Now, the other thing we could potentially do 749 00:32:52,930 --> 00:32:54,515 is what I showed you earlier, which 750 00:32:54,515 --> 00:32:56,890 is the drug that's called Antabuse which, essentially, is 751 00:32:56,890 --> 00:32:59,800 a drug that you can take to prevent yourself from being 752 00:32:59,800 --> 00:33:01,620 able to metabolize alcohol. 753 00:33:01,620 --> 00:33:04,420 I was actually thinking about that for quite a while. 754 00:33:04,420 --> 00:33:09,070 The reason why I wasn't doing that was the tricky part 755 00:33:09,070 --> 00:33:10,870 here is that, once you actually start 756 00:33:10,870 --> 00:33:14,820 drinking while you're on this drug, that's very dangerous. 757 00:33:14,820 --> 00:33:16,570 And particularly if you have been drinking 758 00:33:16,570 --> 00:33:19,600 for quite a while, withdrawal symptoms related to alcohol 759 00:33:19,600 --> 00:33:22,000 are very dangerous, so if you have been drinking alcohol 760 00:33:22,000 --> 00:33:25,360 for 10, 20 years like many of the people in the sample have, 761 00:33:25,360 --> 00:33:27,910 abruptly reducing your drinking is dangerous 762 00:33:27,910 --> 00:33:29,530 because you can essentially get what's 763 00:33:29,530 --> 00:33:35,290 called delirium tremens, very severe withdrawal symptoms that 764 00:33:35,290 --> 00:33:38,080 could even-- you could even die from that. 765 00:33:38,080 --> 00:33:40,000 Now, one thing you could do is when 766 00:33:40,000 --> 00:33:42,790 you get withdrawal symptoms, you can actually start drinking, 767 00:33:42,790 --> 00:33:43,810 and that helps. 768 00:33:43,810 --> 00:33:46,150 Now, the problem is, when you have taken Antabuse, 769 00:33:46,150 --> 00:33:46,942 you can't do that. 770 00:33:46,942 --> 00:33:48,400 So essentially, there's the issues, 771 00:33:48,400 --> 00:33:51,070 and then you get this alcohol-Antabuse reaction 772 00:33:51,070 --> 00:33:52,850 which is really bad. 773 00:33:52,850 --> 00:33:55,420 So I ended up not doing that for various ethical and other 774 00:33:55,420 --> 00:33:56,170 reasons. 775 00:33:56,170 --> 00:33:58,840 But what I did do is essentially give people the option 776 00:33:58,840 --> 00:34:01,397 of dominated choices. 777 00:34:01,397 --> 00:34:02,980 So what does the experiment look like? 778 00:34:02,980 --> 00:34:05,800 It's a three-week field experiment with the rickshaw 779 00:34:05,800 --> 00:34:08,050 drivers in Chennai. 780 00:34:08,050 --> 00:34:11,650 We didn't do the random breathalyzing in the field, 781 00:34:11,650 --> 00:34:13,983 in part because it's really hard to actually find people 782 00:34:13,983 --> 00:34:16,150 if they're sort of roaming around on their rickshaw. 783 00:34:16,150 --> 00:34:18,070 Not everybody has their phone, or answers 784 00:34:18,070 --> 00:34:19,252 their phone, and so on. 785 00:34:19,252 --> 00:34:20,710 So instead, we asked people to come 786 00:34:20,710 --> 00:34:24,639 to the study office between 6:00 and 10:00 PM every day. 787 00:34:24,639 --> 00:34:27,520 It's between 6:00 and 10:00 PM because breathalyzer tests can 788 00:34:27,520 --> 00:34:30,340 tell you about drinking during the previous something like six 789 00:34:30,340 --> 00:34:33,117 to 10 hours, so if you breathalyze me now, 790 00:34:33,117 --> 00:34:34,659 what you can tell is what I have been 791 00:34:34,659 --> 00:34:37,090 drinking in the last, let's say, eight hours or something. 792 00:34:37,090 --> 00:34:39,909 You cannot really tell whether I have been drinking yesterday. 793 00:34:39,909 --> 00:34:41,739 So if we did the breathalyzer test in the morning, 794 00:34:41,739 --> 00:34:44,139 that would not be particularly helpful because, really, it 795 00:34:44,139 --> 00:34:45,931 can't tell you very much about the drinking 796 00:34:45,931 --> 00:34:48,070 on the previous day. 797 00:34:48,070 --> 00:34:51,580 And then we provided financial incentives for sobriety 798 00:34:51,580 --> 00:34:52,810 for a subset of individuals. 799 00:34:52,810 --> 00:34:55,179 What do these financial incentives look like? 800 00:34:55,179 --> 00:34:56,860 They look very simple and something 801 00:34:56,860 --> 00:34:59,590 very similar to what you just suggested. 802 00:34:59,590 --> 00:35:01,780 So option A-- these are the financial incentives-- 803 00:35:01,780 --> 00:35:06,040 are people are given 60 rupees, which is about $1, just 804 00:35:06,040 --> 00:35:06,700 for showing up. 805 00:35:06,700 --> 00:35:08,782 So just come to the office, and show up. 806 00:35:08,782 --> 00:35:10,990 Do a breathalyzer test, and do some survey questions. 807 00:35:10,990 --> 00:35:13,820 And you get like $1 for doing that. 808 00:35:13,820 --> 00:35:16,480 And then if your breathalyzer score is 0, 809 00:35:16,480 --> 00:35:19,630 you get another dollar, 120 rupees. 810 00:35:19,630 --> 00:35:23,110 That's essentially the incentives for sobriety. 811 00:35:23,110 --> 00:35:28,540 Now, option B-- and people were given choices for that. 812 00:35:28,540 --> 00:35:31,060 Option B was-- there was like three different versions 813 00:35:31,060 --> 00:35:32,050 of that. 814 00:35:32,050 --> 00:35:35,030 The first one is where option B was 90 rupees, which, 815 00:35:35,030 --> 00:35:36,640 essentially, is, come to the office, 816 00:35:36,640 --> 00:35:38,230 and you can choose to get 90 rupees. 817 00:35:38,230 --> 00:35:39,430 Just come to the office. 818 00:35:39,430 --> 00:35:42,730 Whenever you show up, you get 90 rupees, regardless 819 00:35:42,730 --> 00:35:44,530 of what your breathalyzer score is. 820 00:35:44,530 --> 00:35:46,930 You are breathalyzed, but it doesn't matter what 821 00:35:46,930 --> 00:35:48,700 your scores for your payment. 822 00:35:48,700 --> 00:35:50,890 Now, when you choose the incentives over option B, 823 00:35:50,890 --> 00:35:52,525 is that a commitment device, or how 824 00:35:52,525 --> 00:35:53,860 do we think about this choice? 825 00:35:59,163 --> 00:36:00,580 What do we learn from that choice? 826 00:36:06,220 --> 00:36:07,640 AUDIENCE: [INAUDIBLE] 827 00:36:07,640 --> 00:36:08,515 FRANK SCHILBACH: Why? 828 00:36:10,640 --> 00:36:20,702 AUDIENCE: [INAUDIBLE] 829 00:36:20,702 --> 00:36:22,160 FRANK SCHILBACH: So let me give you 830 00:36:22,160 --> 00:36:25,020 an example of a person who might choose this. 831 00:36:25,020 --> 00:36:28,250 Suppose there's a person who is actually mostly sober, 832 00:36:28,250 --> 00:36:30,350 is an exponential discounter, and wants 833 00:36:30,350 --> 00:36:31,860 to be drunk just once a week. 834 00:36:31,860 --> 00:36:33,830 So-- sorry-- these are choices for six days 835 00:36:33,830 --> 00:36:35,240 for like the following week. 836 00:36:35,240 --> 00:36:37,100 So the person is perfectly rational. 837 00:36:37,100 --> 00:36:39,650 The person has perfect foresight and knows 838 00:36:39,650 --> 00:36:41,780 that they're going to be drunk once next week, 839 00:36:41,780 --> 00:36:45,080 and they're going to choose option A because they 840 00:36:45,080 --> 00:36:47,430 think it's a great deal. 841 00:36:47,430 --> 00:36:50,955 And so they actually don't want any commitment. 842 00:36:50,955 --> 00:36:52,580 The only reason why they choose that is 843 00:36:52,580 --> 00:36:55,520 because the expected value of the money that they will get 844 00:36:55,520 --> 00:37:00,830 will be higher if they are, at least half of the time, sober. 845 00:37:00,830 --> 00:37:03,540 So that choice, in fact, is not a commitment device. 846 00:37:03,540 --> 00:37:07,460 However, moving towards choices two and three, 847 00:37:07,460 --> 00:37:10,220 now we get into territory of dominated contracts 848 00:37:10,220 --> 00:37:12,600 that, in fact, are commitment devices. 849 00:37:12,600 --> 00:37:14,630 So the key part here is that, when 850 00:37:14,630 --> 00:37:17,630 you look at the second choice, when you choose, now, 851 00:37:17,630 --> 00:37:21,610 option A over option B, you can actually only do worse. 852 00:37:21,610 --> 00:37:23,540 The only reason why you choose option A 853 00:37:23,540 --> 00:37:26,570 is because you want to have incentives. 854 00:37:26,570 --> 00:37:31,190 Now, you risk showing up drunk on some days, in which case 855 00:37:31,190 --> 00:37:34,392 you get less money, but the benefit of that 856 00:37:34,392 --> 00:37:36,350 now is that then now you have these incentives. 857 00:37:36,350 --> 00:37:38,053 So essentially, you want incentives, 858 00:37:38,053 --> 00:37:40,220 but what you take into account is that, potentially, 859 00:37:40,220 --> 00:37:41,887 if you don't show up sober all the time, 860 00:37:41,887 --> 00:37:43,380 you might lose some money. 861 00:37:43,380 --> 00:37:46,610 So this kind of contract or this type of choice 862 00:37:46,610 --> 00:37:50,150 we cal a weekly-dominated contract or choice 863 00:37:50,150 --> 00:37:51,710 in terms of steady payments. 864 00:37:51,710 --> 00:37:55,640 So if people choose option A and the second choice, 865 00:37:55,640 --> 00:37:58,268 that is some demand for commitment. 866 00:37:58,268 --> 00:37:59,810 Now, when I look at the third choice, 867 00:37:59,810 --> 00:38:02,970 now option A is strictly-dominated. 868 00:38:02,970 --> 00:38:07,640 So now, if you choose option A, you will do worse for sure 869 00:38:07,640 --> 00:38:09,710 in terms of how much you will be paid. 870 00:38:09,710 --> 00:38:12,080 Every day, you got like 30 rupees, which is about 10% 871 00:38:12,080 --> 00:38:13,640 of people's daily incomes. 872 00:38:13,640 --> 00:38:16,880 Every day, you got 30 rupees less, 873 00:38:16,880 --> 00:38:19,615 and the benefit, again, is like incentives for sobriety. 874 00:38:19,615 --> 00:38:21,740 So if you want to choose these incentives that will 875 00:38:21,740 --> 00:38:23,870 be costly for you for sure-- 876 00:38:23,870 --> 00:38:27,690 and now you have a sort of weigh whether you like those or not. 877 00:38:27,690 --> 00:38:29,690 So what I find is and what you see here 878 00:38:29,690 --> 00:38:32,780 is, essentially, the fraction of people choosing incentives 879 00:38:32,780 --> 00:38:35,330 for choice one, two, and three. 880 00:38:35,330 --> 00:38:38,900 So in black you see choice one, which is a 90 rupees choice. 881 00:38:38,900 --> 00:38:42,950 In dark blue you see the 120 rupees choices. 882 00:38:42,950 --> 00:38:44,840 You see the weekly-dominated choices, 883 00:38:44,840 --> 00:38:50,240 and then you see, in light blue, the 150 rupee choices. 884 00:38:50,240 --> 00:38:53,412 These are the strictly-dominated choices, 885 00:38:53,412 --> 00:38:54,620 and there's like three weeks. 886 00:38:54,620 --> 00:38:55,620 And people will do this. 887 00:38:55,620 --> 00:38:57,798 So each week, they choose for the following week 888 00:38:57,798 --> 00:38:59,840 what incentives they would like to set themselves 889 00:38:59,840 --> 00:39:01,940 for the next six days. 890 00:39:01,940 --> 00:39:04,700 So these are the three options again, 891 00:39:04,700 --> 00:39:07,290 and you see the choices over time. 892 00:39:07,290 --> 00:39:09,590 So what we find is or what I find is that, 893 00:39:09,590 --> 00:39:12,170 in fact, quite a few people, about half of the individuals, 894 00:39:12,170 --> 00:39:14,420 choose the incentives when they're weekly-dominated, 895 00:39:14,420 --> 00:39:16,880 which have some evidence of demand for commitment, 896 00:39:16,880 --> 00:39:19,310 but perhaps more strikingly, about a third of individuals 897 00:39:19,310 --> 00:39:21,560 choose the incentives when they're strictly-dominated. 898 00:39:21,560 --> 00:39:25,040 So these are people who are willing to forgo three times 899 00:39:25,040 --> 00:39:27,438 in a row, not just once but in the second week 900 00:39:27,438 --> 00:39:29,480 and also in the third week, and then if anything, 901 00:39:29,480 --> 00:39:31,580 the fraction of people choosing these incentives 902 00:39:31,580 --> 00:39:32,880 goes up over time. 903 00:39:32,880 --> 00:39:34,850 People are willing to choose these incentives 904 00:39:34,850 --> 00:39:38,660 and willing to give up about 10% of their daily incomes 905 00:39:38,660 --> 00:39:42,200 for like an entire week to receive incentives to be sober. 906 00:39:42,200 --> 00:39:45,890 So that's very strong evidence of demand for commitment. 907 00:39:45,890 --> 00:39:49,230 Now, what we also find is, when you sort of randomize 908 00:39:49,230 --> 00:39:52,580 just the incentives themselves for some people 909 00:39:52,580 --> 00:39:54,680 versus unconditional payments overall, 910 00:39:54,680 --> 00:39:56,540 these incentives are, in fact, successful 911 00:39:56,540 --> 00:39:58,010 in reducing people's day drinking. 912 00:39:58,010 --> 00:40:01,520 So people actually show up more sober 913 00:40:01,520 --> 00:40:02,930 when they receive incentives. 914 00:40:02,930 --> 00:40:05,250 However, as you might have expected as well, 915 00:40:05,250 --> 00:40:08,690 the problem is the breathalyzer incentivizes 916 00:40:08,690 --> 00:40:11,360 during the day, the time before 6:00 and 10:00 917 00:40:11,360 --> 00:40:13,640 PM, before people come to the study office. 918 00:40:13,640 --> 00:40:16,988 It is not incentivizing the time after that's in the evening. 919 00:40:16,988 --> 00:40:18,780 And as you sort of-- this is some work by-- 920 00:40:18,780 --> 00:40:22,280 Holmstrom and others would say, if you have multitasking 921 00:40:22,280 --> 00:40:24,752 at different times of the day or different actions 922 00:40:24,752 --> 00:40:26,960 that you can take, if you incentivize one thing, what 923 00:40:26,960 --> 00:40:29,270 often happens-- people follow these incentives, 924 00:40:29,270 --> 00:40:32,130 but then they substitute to other things in their life. 925 00:40:32,130 --> 00:40:34,550 So then what, in fact, happens-- 926 00:40:34,550 --> 00:40:36,200 overall, people's overall drinking 927 00:40:36,200 --> 00:40:38,257 does not decrease very much. 928 00:40:38,257 --> 00:40:40,340 So while people seem to have quite a bit of demand 929 00:40:40,340 --> 00:40:43,430 for these incentives, in fact, it 930 00:40:43,430 --> 00:40:45,890 is helpful for increasing their daytime sobriety 931 00:40:45,890 --> 00:40:49,550 but not helpful for decreasing their overall drinking. 932 00:40:49,550 --> 00:40:51,252 So then I was looking at the impact 933 00:40:51,252 --> 00:40:53,210 of that treatment in the sense like, now people 934 00:40:53,210 --> 00:40:55,315 are more sober during the day. 935 00:40:55,315 --> 00:40:56,690 In fact, somewhat surprisingly, I 936 00:40:56,690 --> 00:40:59,120 find very little evidence on labor market outcomes. 937 00:40:59,120 --> 00:41:01,400 It's not like actually people are working better, 938 00:41:01,400 --> 00:41:03,650 but I do find people make different savings decisions. 939 00:41:03,650 --> 00:41:06,293 People are willing to save a lot more when they're sober, 940 00:41:06,293 --> 00:41:08,210 even when you take into account that they also 941 00:41:08,210 --> 00:41:11,180 might have more money overall, or at least more money 942 00:41:11,180 --> 00:41:11,720 available. 943 00:41:11,720 --> 00:41:12,892 Yes? 944 00:41:12,892 --> 00:41:26,065 AUDIENCE: I was wondering if there was [INAUDIBLE] 945 00:41:26,065 --> 00:41:31,312 but they may drink at night 946 00:41:31,312 --> 00:41:33,270 FRANK SCHILBACH: So I spent quite a bit of time 947 00:41:33,270 --> 00:41:36,060 thinking about demand effect and making sure that people are-- 948 00:41:36,060 --> 00:41:38,130 some of the people in the setting 949 00:41:38,130 --> 00:41:42,063 have much less concerns about what other people think 950 00:41:42,063 --> 00:41:42,730 when they drink. 951 00:41:42,730 --> 00:41:48,010 They're very happy to tell you, yesterday I drank six drinks. 952 00:41:48,010 --> 00:41:50,010 They're also very happy to tell you, yesterday I 953 00:41:50,010 --> 00:41:52,090 spent a third of my income on alcohol and so on. 954 00:41:52,090 --> 00:41:54,423 People are very open about this, so it's not like people 955 00:41:54,423 --> 00:41:57,000 misreport in surveys and so on. 956 00:41:57,000 --> 00:42:00,570 People's reported drinking is also very much consistent 957 00:42:00,570 --> 00:42:02,270 with the actual breathalyzer scores, 958 00:42:02,270 --> 00:42:03,270 so if you sort of matched that up 959 00:42:03,270 --> 00:42:04,920 in terms of how much have you been drinking today 960 00:42:04,920 --> 00:42:06,930 versus what the breathalyzer score says, 961 00:42:06,930 --> 00:42:09,190 that seems consistent as well. 962 00:42:09,190 --> 00:42:10,680 Now, there are some studies looking 963 00:42:10,680 --> 00:42:13,200 at social-desirability bias and demand effects, 964 00:42:13,200 --> 00:42:15,928 and often what you find is that when choices are not 965 00:42:15,928 --> 00:42:17,970 incentivized, when they are very weak incentives, 966 00:42:17,970 --> 00:42:19,920 there is often evidence for demand effect 967 00:42:19,920 --> 00:42:21,510 and social-desirability bias. 968 00:42:21,510 --> 00:42:23,940 But when you make things very costly in a sense of like, 969 00:42:23,940 --> 00:42:25,398 this is a very costly thing for you 970 00:42:25,398 --> 00:42:27,510 to do to just make the experimenter happy-- 971 00:42:27,510 --> 00:42:28,927 we spent a lot of time essentially 972 00:42:28,927 --> 00:42:31,302 being very clear about like, you can do whatever you want 973 00:42:31,302 --> 00:42:32,290 and so on and so forth. 974 00:42:32,290 --> 00:42:34,830 And what we tend to find is that, in the first few days, 975 00:42:34,830 --> 00:42:36,730 people seem slightly uncomfortable in this. 976 00:42:36,730 --> 00:42:38,313 There's more to the study, but there's 977 00:42:38,313 --> 00:42:40,230 like a baseline period of about a week, 978 00:42:40,230 --> 00:42:41,970 and the first few days, people seem 979 00:42:41,970 --> 00:42:44,980 to be slightly uncomfortable of being drunk in the office 980 00:42:44,980 --> 00:42:46,750 and so on and so forth and sort of 981 00:42:46,750 --> 00:42:49,320 having positive breathalyzer scores. 982 00:42:49,320 --> 00:42:50,910 That goes away very quickly, so people 983 00:42:50,910 --> 00:42:54,670 are very happy to show up as drunk as they are. 984 00:42:54,670 --> 00:42:57,707 So overall, I would think, if this is one choice, 985 00:42:57,707 --> 00:42:59,790 and you're like, maybe people are making mistakes, 986 00:42:59,790 --> 00:43:01,710 or people maybe want to make you happy, 987 00:43:01,710 --> 00:43:03,710 once you make that choice once, for a whole week 988 00:43:03,710 --> 00:43:05,293 you get, essentially, a lot less money 989 00:43:05,293 --> 00:43:06,645 than you would get otherwise. 990 00:43:06,645 --> 00:43:07,770 You make that choice again. 991 00:43:07,770 --> 00:43:09,270 Again, you get, like for a whole week a lot 992 00:43:09,270 --> 00:43:10,870 less money than you would otherwise, 993 00:43:10,870 --> 00:43:12,370 and then people choose it yet again. 994 00:43:12,370 --> 00:43:16,170 So while I can't entirely rule out this explanation-- 995 00:43:16,170 --> 00:43:18,108 there's a discussion of that in the paper. 996 00:43:18,108 --> 00:43:19,650 While I can't rule out this entirely, 997 00:43:19,650 --> 00:43:22,990 it seems unlikely, at least in the second and third choice, 998 00:43:22,990 --> 00:43:26,065 given that it's really, really costly for people. 999 00:43:26,065 --> 00:43:27,690 AUDIENCE: My question was also related. 1000 00:43:27,690 --> 00:43:29,977 If they know that the [INAUDIBLE] device is not 1001 00:43:29,977 --> 00:43:31,769 working because they are drinking at night, 1002 00:43:31,769 --> 00:43:32,640 why do they keep-- 1003 00:43:32,640 --> 00:43:33,450 FRANK SCHILBACH: Right. 1004 00:43:33,450 --> 00:43:34,830 So the second question is like, OK, 1005 00:43:34,830 --> 00:43:36,520 now why are people making these choices? 1006 00:43:36,520 --> 00:43:41,280 So one broad question is like, OK, so-- 1007 00:43:41,280 --> 00:43:44,280 when talking to people, a lot of people would say, well-- 1008 00:43:44,280 --> 00:43:47,670 they would come, be like, last week it was really difficult. 1009 00:43:47,670 --> 00:43:49,410 I didn't really succeed and so on. 1010 00:43:49,410 --> 00:43:51,060 And then they would be like, from tomorrow onwards, 1011 00:43:51,060 --> 00:43:52,030 things will be different. 1012 00:43:52,030 --> 00:43:53,970 So there is lots of people who, in fact, seem 1013 00:43:53,970 --> 00:43:56,130 naive in the sense of they really think, 1014 00:43:56,130 --> 00:43:59,610 going forward, they will change their behavior. 1015 00:43:59,610 --> 00:44:01,980 Often, I think what they are reasonably 1016 00:44:01,980 --> 00:44:03,840 clear about is how often they will actually 1017 00:44:03,840 --> 00:44:05,577 be sober going forward. 1018 00:44:05,577 --> 00:44:07,410 What they have trouble with is, essentially, 1019 00:44:07,410 --> 00:44:09,300 predicting their behavior later in the day. 1020 00:44:09,300 --> 00:44:10,830 So they seem to think like, OK, tomorrow 1021 00:44:10,830 --> 00:44:12,538 I'm not going to drink, and then I'm also 1022 00:44:12,538 --> 00:44:13,860 not going to drink at nights. 1023 00:44:13,860 --> 00:44:15,810 That doesn't seem to change. 1024 00:44:15,810 --> 00:44:17,185 Now, there's a second explanation 1025 00:44:17,185 --> 00:44:19,393 that's also in the paper and discussed in more detail 1026 00:44:19,393 --> 00:44:21,180 that I can't really necessarily rule out. 1027 00:44:21,180 --> 00:44:24,648 There are actually some benefits of reducing people's drinking. 1028 00:44:24,648 --> 00:44:26,190 These are all kind of small benefits, 1029 00:44:26,190 --> 00:44:27,960 but they actually sort of add up. 1030 00:44:27,960 --> 00:44:29,302 So one part of that is savings. 1031 00:44:29,302 --> 00:44:30,510 People save a lot more money. 1032 00:44:30,510 --> 00:44:32,010 They get more interest from saving. 1033 00:44:32,010 --> 00:44:35,490 That's actually quite a bit of money that they make from that, 1034 00:44:35,490 --> 00:44:37,540 so they potentially make better decisions. 1035 00:44:37,540 --> 00:44:40,200 Second, they have like-- so while it's not significant, 1036 00:44:40,200 --> 00:44:42,360 people have somewhat higher earnings. 1037 00:44:42,360 --> 00:44:43,890 People also have somewhat reduced 1038 00:44:43,890 --> 00:44:45,967 costs of drinking and so on. 1039 00:44:45,967 --> 00:44:47,550 So there's some, essentially, evidence 1040 00:44:47,550 --> 00:44:50,820 of several benefits in the order of magnitude of like 5, 1041 00:44:50,820 --> 00:44:52,180 10 rupees per day. 1042 00:44:52,180 --> 00:44:54,120 So when you have somewhat increased earnings, 1043 00:44:54,120 --> 00:44:57,500 somewhat reduce expenditures on alcohol, increased savings-- 1044 00:44:57,500 --> 00:44:59,250 and then on top of that, people would also 1045 00:44:59,250 --> 00:45:01,680 tell you they like to be what they say, like a sober man. 1046 00:45:01,680 --> 00:45:02,430 These are all men. 1047 00:45:02,430 --> 00:45:04,950 They like to be a sober man, and so they have some value of, 1048 00:45:04,950 --> 00:45:07,357 actually, sobriety during the day. 1049 00:45:07,357 --> 00:45:08,440 So I think it's a mixture. 1050 00:45:08,440 --> 00:45:10,410 My guess is-- when we've had to best explain the evidence 1051 00:45:10,410 --> 00:45:12,420 that we have, it's a mixture of there actually 1052 00:45:12,420 --> 00:45:16,560 being some evidence of benefits of being sober that you 1053 00:45:16,560 --> 00:45:22,140 can weigh against like 30 rupees forgone benefits 1054 00:45:22,140 --> 00:45:23,770 from the study. 1055 00:45:23,770 --> 00:45:26,790 And there is also some naivete in a sense of people desires, 1056 00:45:26,790 --> 00:45:32,280 people's beliefs about wanting to drink glass also 1057 00:45:32,280 --> 00:45:35,910 in the evening, and that's just not necessarily working out. 1058 00:45:35,910 --> 00:45:37,890 I want to say another comment to this, 1059 00:45:37,890 --> 00:45:40,350 which is when you see people choosing commitment 1060 00:45:40,350 --> 00:45:44,250 devices that fail, you can't necessarily rule out 1061 00:45:44,250 --> 00:45:46,770 that people are reasonably sophisticated, 1062 00:45:46,770 --> 00:45:47,760 the reason being-- 1063 00:45:47,760 --> 00:45:49,320 for example, if I were a smoker that 1064 00:45:49,320 --> 00:45:52,200 smoked a lot and the value of me stopping to smoke 1065 00:45:52,200 --> 00:45:55,110 is really high-- suppose I think if I just could stop smoking 1066 00:45:55,110 --> 00:45:56,610 I would live like five years longer. 1067 00:45:56,610 --> 00:45:59,160 That's a huge value in terms of the benefits. 1068 00:45:59,160 --> 00:46:01,440 If somebody came to me and said, there's 1069 00:46:01,440 --> 00:46:03,840 a commitment device with a 1% chance it's going to work. 1070 00:46:03,840 --> 00:46:05,710 Would you like to buy it? 1071 00:46:05,710 --> 00:46:07,200 And it costs $100. 1072 00:46:07,200 --> 00:46:10,500 I might actually choose this commitment device 1073 00:46:10,500 --> 00:46:12,750 because the expected value of that 1074 00:46:12,750 --> 00:46:16,110 is actually higher, even if 99% of the cases I'm 1075 00:46:16,110 --> 00:46:17,500 actually going to fail. 1076 00:46:17,500 --> 00:46:19,230 So then you're looking at the data. 1077 00:46:19,230 --> 00:46:21,272 You might say, look, these people are essentially 1078 00:46:21,272 --> 00:46:22,483 making a bunch of mistakes. 1079 00:46:22,483 --> 00:46:23,900 But I think from their perspective 1080 00:46:23,900 --> 00:46:25,790 it's rather-- they have been drinking for so long. 1081 00:46:25,790 --> 00:46:27,165 They struggle with this, and they 1082 00:46:27,165 --> 00:46:28,430 try to reduce their drinking. 1083 00:46:28,430 --> 00:46:30,263 There is a chance that maybe this will work. 1084 00:46:30,263 --> 00:46:31,050 Maybe it will not. 1085 00:46:31,050 --> 00:46:33,500 So let me really try and do this and see. 1086 00:46:33,500 --> 00:46:36,440 Unfortunately, as it turns out, it's, in fact, not successful, 1087 00:46:36,440 --> 00:46:39,740 at least when it comes to their drinking overall. 1088 00:46:43,000 --> 00:46:45,220 Second, speaking of smoking, as a second paper 1089 00:46:45,220 --> 00:46:48,850 on commitment devices related to sort of addictive products, 1090 00:46:48,850 --> 00:46:53,200 this is a voluntary commitment product for smoking cessation. 1091 00:46:53,200 --> 00:46:55,630 Smokers are offered a nonstandard savings account. 1092 00:46:55,630 --> 00:46:59,170 This is the same setting as in the Philippines 1093 00:46:59,170 --> 00:47:01,960 where these are clients of a bank. 1094 00:47:01,960 --> 00:47:05,080 These are smokers selecting on actually somebody 1095 00:47:05,080 --> 00:47:06,220 being a smoker. 1096 00:47:06,220 --> 00:47:09,190 And they are offered the following commitment device. 1097 00:47:09,190 --> 00:47:11,920 You can deposit funds for six months, 1098 00:47:11,920 --> 00:47:14,490 and then you can take a urine test for nicotine and cotinine, 1099 00:47:14,490 --> 00:47:16,990 which apparently tests pretty well, whether you haven't been 1100 00:47:16,990 --> 00:47:19,990 drinking-- sorry, not been smoking for the last week 1101 00:47:19,990 --> 00:47:21,640 or so or a bit longer. 1102 00:47:21,640 --> 00:47:24,130 If you pass, you get the money back that's in your account. 1103 00:47:24,130 --> 00:47:27,483 If you fail, the money is gone, essentially. 1104 00:47:27,483 --> 00:47:29,650 Now, do we think this is a good commitment contract? 1105 00:47:29,650 --> 00:47:30,820 Or do we like this or not? 1106 00:47:30,820 --> 00:47:32,903 Or how do we think about this commitment contract? 1107 00:47:40,260 --> 00:47:41,073 Yes? 1108 00:47:41,073 --> 00:47:49,290 AUDIENCE: [INAUDIBLE] 1109 00:47:49,290 --> 00:47:50,367 FRANK SCHILBACH: Yeah. 1110 00:47:50,367 --> 00:47:52,200 Exactly, you could lose quite a bit of money 1111 00:47:52,200 --> 00:47:57,122 and make people quite unhappy, but do we think-- 1112 00:47:57,122 --> 00:47:58,830 actually, suppose you were sophisticated. 1113 00:47:58,830 --> 00:48:03,756 Would you sign up, or how would you behave? 1114 00:48:03,756 --> 00:48:05,560 Like suppose I offer you the contract. 1115 00:48:05,560 --> 00:48:06,520 You have it now. 1116 00:48:06,520 --> 00:48:09,505 What would you do? 1117 00:48:09,505 --> 00:48:12,012 AUDIENCE: [INAUDIBLE] 1118 00:48:12,012 --> 00:48:12,970 FRANK SCHILBACH: Right. 1119 00:48:12,970 --> 00:48:15,930 So you put money in it, but I think the one key part here 1120 00:48:15,930 --> 00:48:19,710 is that the actual quitting only occurs 1121 00:48:19,710 --> 00:48:21,330 in like six months from now. 1122 00:48:21,330 --> 00:48:24,300 It's to say, it's sort of a little bit of a funny thing 1123 00:48:24,300 --> 00:48:26,100 in the sense of saying, look, you really 1124 00:48:26,100 --> 00:48:27,160 want to quit smoking. 1125 00:48:27,160 --> 00:48:29,340 It's like, in six months from now, 1126 00:48:29,340 --> 00:48:31,870 you already have lots of long-run benefits-- sorry, 1127 00:48:31,870 --> 00:48:33,910 lots of long-run costs of smoking, 1128 00:48:33,910 --> 00:48:37,660 which is lung cancer, health issues, and so on and so forth. 1129 00:48:37,660 --> 00:48:40,110 Now I'm sort of adding long-run incentives 1130 00:48:40,110 --> 00:48:43,590 in like six months from now, which is a little bit far away. 1131 00:48:43,590 --> 00:48:45,960 So what it sort of needs is-- 1132 00:48:45,960 --> 00:48:48,525 in a way, you would not start quitting to smoke right away. 1133 00:48:48,525 --> 00:48:50,400 You would say, well, in five months from now, 1134 00:48:50,400 --> 00:48:51,810 I can sort of stop. 1135 00:48:51,810 --> 00:48:54,570 And now, of course, you can then build your own commitment 1136 00:48:54,570 --> 00:48:56,700 device by putting in a lot of money 1137 00:48:56,700 --> 00:48:59,185 and then really ramping up these incentives. 1138 00:48:59,185 --> 00:49:01,810 But in some sense, what you kind of want is a commitment device 1139 00:49:01,810 --> 00:49:02,070 that-- 1140 00:49:02,070 --> 00:49:02,970 I sign up for it. 1141 00:49:02,970 --> 00:49:04,860 I'm like really excited to stop smoking, 1142 00:49:04,860 --> 00:49:07,500 and now you want to provide incentives almost right away 1143 00:49:07,500 --> 00:49:08,775 to get people going. 1144 00:49:08,775 --> 00:49:10,650 But instead, it's this commitment device that 1145 00:49:10,650 --> 00:49:12,840 requires quite a bit of sophistication, in a sense, 1146 00:49:12,840 --> 00:49:15,510 of saying, now, I know that in like six months from now 1147 00:49:15,510 --> 00:49:16,470 I want to stop smoking. 1148 00:49:16,470 --> 00:49:18,428 I kind of don't want to stop smoking right now. 1149 00:49:18,428 --> 00:49:20,920 I can always do it tomorrow, because people like to delay, 1150 00:49:20,920 --> 00:49:23,310 depending on how naive they are and so on. 1151 00:49:23,310 --> 00:49:25,967 And so I can see lots of people then saying, 1152 00:49:25,967 --> 00:49:28,050 yeah, in three months from now I'm going to do it. 1153 00:49:28,050 --> 00:49:30,300 And then they're going to put in some money and so on, 1154 00:49:30,300 --> 00:49:32,520 but they end up not doing it eventually. 1155 00:49:32,520 --> 00:49:35,340 It turns out, empirically, it works reasonably well 1156 00:49:35,340 --> 00:49:35,910 in a sense. 1157 00:49:35,910 --> 00:49:37,410 There's like 10% of take-up. 1158 00:49:37,410 --> 00:49:39,810 Like 10% of people actually put in some money. 1159 00:49:39,810 --> 00:49:44,790 There is also like an impact on smoking of 3 percentage points. 1160 00:49:44,790 --> 00:49:47,490 That sounds kind of small, but that's actually 1161 00:49:47,490 --> 00:49:49,200 pretty sizable in terms of getting people 1162 00:49:49,200 --> 00:49:51,240 to stop smoking is pretty hard. 1163 00:49:51,240 --> 00:49:53,880 Also, these effects persist in surprise tests 1164 00:49:53,880 --> 00:49:54,983 like 12 months later. 1165 00:49:54,983 --> 00:49:57,400 So when you sort of surprise people later and sort of say, 1166 00:49:57,400 --> 00:50:00,380 hey, surprise, here's a test, you 1167 00:50:00,380 --> 00:50:02,100 get some money for doing it, turns out 1168 00:50:02,100 --> 00:50:07,890 that people who offered this commitment, smoking a contract 1169 00:50:07,890 --> 00:50:11,040 or product are, in fact, less likely to smoke. 1170 00:50:14,520 --> 00:50:18,700 Now, a second question related to that is to say, 1171 00:50:18,700 --> 00:50:22,160 if people are unhappy about their smoking in the future, 1172 00:50:22,160 --> 00:50:27,180 if people would like to reduce their future smoking, 1173 00:50:27,180 --> 00:50:29,040 there's some arguments that taxes 1174 00:50:29,040 --> 00:50:31,380 can be some form of not quite commitment devices 1175 00:50:31,380 --> 00:50:35,513 but ways in which you could reduce your smoking. 1176 00:50:35,513 --> 00:50:37,680 So the rickshaw drivers in Chennai would of tell me, 1177 00:50:37,680 --> 00:50:40,080 I would love for alcohol prices to be 1178 00:50:40,080 --> 00:50:41,595 more expensive because that would 1179 00:50:41,595 --> 00:50:42,720 help me reduce my drinking. 1180 00:50:42,720 --> 00:50:45,570 So if they could set prices, quite a few people would say, 1181 00:50:45,570 --> 00:50:47,610 I would set prices really high, so that would 1182 00:50:47,610 --> 00:50:49,180 help me reduce my drinking. 1183 00:50:49,180 --> 00:50:51,148 So if people could choose their own prices 1184 00:50:51,148 --> 00:50:53,190 and if they would set high prices, that would be, 1185 00:50:53,190 --> 00:50:54,648 essentially, a form of a commitment 1186 00:50:54,648 --> 00:50:58,140 device to reduce your smoking or drinking. 1187 00:50:58,140 --> 00:51:00,960 So when you look at smokers, there's 1188 00:51:00,960 --> 00:51:03,750 lots of people who say they would like to quit. 1189 00:51:03,750 --> 00:51:06,930 About eight out of 10 US smokers say that. 1190 00:51:06,930 --> 00:51:08,880 If you have friends who smoke, you 1191 00:51:08,880 --> 00:51:10,870 will get this common patterns. 1192 00:51:10,870 --> 00:51:12,900 So lots of people say, I would love to quit, 1193 00:51:12,900 --> 00:51:16,510 lots of quit attempts that often are not successful. 1194 00:51:16,510 --> 00:51:18,570 Now, one way you can sort of reduce smoking 1195 00:51:18,570 --> 00:51:21,690 potentially or accomplish your goal is to say, 1196 00:51:21,690 --> 00:51:22,950 prices might be higher. 1197 00:51:22,950 --> 00:51:25,290 Again, higher prices-- essentially, 1198 00:51:25,290 --> 00:51:27,390 that's the law of demand. 1199 00:51:27,390 --> 00:51:30,000 Like if prices go up, people smoke less, 1200 00:51:30,000 --> 00:51:32,050 and that's true for smoking, drinking as well. 1201 00:51:32,050 --> 00:51:35,580 And so that might help curb people's self-control problems. 1202 00:51:35,580 --> 00:51:38,685 Now, one simple question that John Gruber, who 1203 00:51:38,685 --> 00:51:42,300 was in the economics department, and Sendhil Mullainathan we're 1204 00:51:42,300 --> 00:51:45,780 asking is like, well, once we increase taxes, 1205 00:51:45,780 --> 00:51:48,030 now smokers are actually happier. 1206 00:51:48,030 --> 00:51:53,402 And so if it's true that people have self-control problems, 1207 00:51:53,402 --> 00:51:55,860 people might be happier because it helps them, essentially, 1208 00:51:55,860 --> 00:51:56,910 reduce their smoking. 1209 00:51:56,910 --> 00:51:59,790 It helps them deal with their self-control problems. 1210 00:51:59,790 --> 00:52:02,927 You could even say, once prices become infinitely expensive, 1211 00:52:02,927 --> 00:52:05,010 as long as people can't substitute in certain ways 1212 00:52:05,010 --> 00:52:08,368 or get around these rules, they might even 1213 00:52:08,368 --> 00:52:10,410 be happy about banning smoking because that would 1214 00:52:10,410 --> 00:52:12,785 be the simplest way of getting rid of their smoking habit 1215 00:52:12,785 --> 00:52:14,670 if they really want to do so. 1216 00:52:14,670 --> 00:52:17,820 Now, Gruber and Mullainathan have an empirical test 1217 00:52:17,820 --> 00:52:21,250 of this in the US and Canada, and in fact, 1218 00:52:21,250 --> 00:52:22,930 when you look at variation over time, 1219 00:52:22,930 --> 00:52:26,250 higher cigarette taxes are associated 1220 00:52:26,250 --> 00:52:28,140 with higher self-reported well-being, 1221 00:52:28,140 --> 00:52:31,830 in particular among those who are likely smokers. 1222 00:52:31,830 --> 00:52:34,260 Now, you could also say, well, other people are happier 1223 00:52:34,260 --> 00:52:36,593 as well, but that's not that interesting because there's 1224 00:52:36,593 --> 00:52:37,990 externalities from smoking. 1225 00:52:37,990 --> 00:52:39,780 And if you are not a smoker, you should 1226 00:52:39,780 --> 00:52:42,072 be happy about smoking taxes because, A, they're 1227 00:52:42,072 --> 00:52:43,530 going to reduce your externalities. 1228 00:52:43,530 --> 00:52:45,363 People are going to smoke less in your face, 1229 00:52:45,363 --> 00:52:46,980 and you might sort of like that. 1230 00:52:46,980 --> 00:52:49,110 Second, there's going to be revenues from taxation. 1231 00:52:49,110 --> 00:52:51,360 You're also going to be very happy if other people pay 1232 00:52:51,360 --> 00:52:52,955 higher taxes. 1233 00:52:52,955 --> 00:52:54,330 But what they do is they look at, 1234 00:52:54,330 --> 00:52:56,670 essentially, households particularly 1235 00:52:56,670 --> 00:53:00,810 that are likely smokers, and that helps then identify-- 1236 00:53:00,810 --> 00:53:03,720 or that's much more interesting a result than households 1237 00:53:03,720 --> 00:53:05,070 who are non-smokers. 1238 00:53:05,070 --> 00:53:07,380 There's some caveat that some of the respondents 1239 00:53:07,380 --> 00:53:09,953 might be spouses of the actual smokers. 1240 00:53:09,953 --> 00:53:11,370 Again, that's, in some sense, less 1241 00:53:11,370 --> 00:53:14,580 interesting from the beta-delta perspective 1242 00:53:14,580 --> 00:53:17,310 because if the husband or wife wants to get their spouse 1243 00:53:17,310 --> 00:53:19,710 to smoke less, they might be quite happy about it, 1244 00:53:19,710 --> 00:53:23,550 but it's more like because the spouse thinks their husband 1245 00:53:23,550 --> 00:53:26,460 or wife have a self-control problem as opposed to they 1246 00:53:26,460 --> 00:53:28,980 have one themselves. 1247 00:53:28,980 --> 00:53:31,570 Or it could be even-- like for lack of self-control problems, 1248 00:53:31,570 --> 00:53:33,510 it could just be the husband likes smoking, 1249 00:53:33,510 --> 00:53:34,950 the wife does not. 1250 00:53:34,950 --> 00:53:38,580 If prices go up, the husband might still smoke less, 1251 00:53:38,580 --> 00:53:41,170 and the wife might be happier or the other way around 1252 00:53:41,170 --> 00:53:43,003 and nothing to do with self-control problem. 1253 00:53:43,003 --> 00:53:47,400 Just like if prices go up people, consume stuff less. 1254 00:53:47,400 --> 00:53:49,080 To be clear, such effects are not 1255 00:53:49,080 --> 00:53:54,720 possible and not consistent with exponential discounting 1256 00:53:54,720 --> 00:53:56,700 because if you're an exponential discounter, 1257 00:53:56,700 --> 00:53:59,370 you would either smoke because you like smoking 1258 00:53:59,370 --> 00:54:00,690 or you would not smoke. 1259 00:54:00,690 --> 00:54:02,760 There's no scenario in which you are smoking 1260 00:54:02,760 --> 00:54:04,880 but you really would like to quit. 1261 00:54:04,880 --> 00:54:07,380 If you would like to quit and if it's worth for you to quit, 1262 00:54:07,380 --> 00:54:10,440 you would actually do that, so you would not 1263 00:54:10,440 --> 00:54:12,720 be happier about higher prices the same way 1264 00:54:12,720 --> 00:54:15,810 as you would not be happy about higher prices for apples 1265 00:54:15,810 --> 00:54:16,747 and bananas. 1266 00:54:16,747 --> 00:54:19,080 People don't like high prices because, essentially, they 1267 00:54:19,080 --> 00:54:23,250 have to pay more, and that sort of reduces their budget. 1268 00:54:23,250 --> 00:54:26,360 Any questions on this? 1269 00:54:26,360 --> 00:54:27,209 Yes? 1270 00:54:27,209 --> 00:54:39,175 AUDIENCE: [INAUDIBLE] 1271 00:54:39,175 --> 00:54:40,050 FRANK SCHILBACH: Yes. 1272 00:54:40,050 --> 00:54:44,280 So I think you're bringing up a very good point, which is often 1273 00:54:44,280 --> 00:54:47,700 a big discussion for so-called "sin taxes," which 1274 00:54:47,700 --> 00:54:52,920 is essentially taxes on either cigarettes, on alcohol, 1275 00:54:52,920 --> 00:54:56,650 on sodas, and so on. 1276 00:54:56,650 --> 00:54:59,040 So there's different objectives often. 1277 00:54:59,040 --> 00:55:01,440 One is essentially helping people who potentially 1278 00:55:01,440 --> 00:55:02,950 have self-control problems. 1279 00:55:02,950 --> 00:55:07,260 Second, there is sort of redistribution as a goal 1280 00:55:07,260 --> 00:55:09,070 or a problem in some sense. 1281 00:55:09,070 --> 00:55:12,040 And that's the same case in India, in some sense. 1282 00:55:12,040 --> 00:55:14,700 If you increased the alcohol prices in India, on the one 1283 00:55:14,700 --> 00:55:18,870 hand, people would drink less, but the elasticity of drinking 1284 00:55:18,870 --> 00:55:23,070 tends to be fairly low, meaning that if you increase 1285 00:55:23,070 --> 00:55:25,890 the price of alcohol, or cigarettes, or the like by 1%, 1286 00:55:25,890 --> 00:55:29,873 usually the consumption only reduces by less than 1%, 1287 00:55:29,873 --> 00:55:32,040 which means that overall expenditures-- while people 1288 00:55:32,040 --> 00:55:35,490 actually drink or smoke less, overall expenditures actually 1289 00:55:35,490 --> 00:55:36,463 go up. 1290 00:55:36,463 --> 00:55:38,130 And so if you're worried about the poor, 1291 00:55:38,130 --> 00:55:40,140 then, in particular, A, they tend 1292 00:55:40,140 --> 00:55:42,660 to consume these types of goods more 1293 00:55:42,660 --> 00:55:45,120 when it comes to smoking, drinking, but also sodas, 1294 00:55:45,120 --> 00:55:47,580 in particular in the US-- 1295 00:55:47,580 --> 00:55:51,312 when you're worried, then, about increasing taxes and then 1296 00:55:51,312 --> 00:55:52,770 expenditures and then, essentially, 1297 00:55:52,770 --> 00:55:56,130 having a particular high burden among the poor-- 1298 00:55:56,130 --> 00:55:59,640 So then there's questions on how to deal with that. 1299 00:55:59,640 --> 00:56:04,620 So one potential "why did you do this" would be, in some sense, 1300 00:56:04,620 --> 00:56:06,600 to refund, then, the taxes that are actually 1301 00:56:06,600 --> 00:56:09,900 incurred, in particular, then in some form of transfers. 1302 00:56:09,900 --> 00:56:11,730 That's often not necessarily feasible, 1303 00:56:11,730 --> 00:56:13,920 but that's exactly the problem when you think 1304 00:56:13,920 --> 00:56:15,960 about, for example, soda taxes. 1305 00:56:15,960 --> 00:56:18,202 There's a very nice paper by Dmitry Taubinsky, 1306 00:56:18,202 --> 00:56:19,410 and Hunt Allcott, and so on-- 1307 00:56:19,410 --> 00:56:21,630 I can add this to the slides-- 1308 00:56:21,630 --> 00:56:24,210 that essentially looking at exactly that in the US. 1309 00:56:24,210 --> 00:56:26,070 When you think about soda taxes, how 1310 00:56:26,070 --> 00:56:27,870 do you deal with two different objectives? 1311 00:56:27,870 --> 00:56:31,680 One is helping people to consume less sodas if they desire 1312 00:56:31,680 --> 00:56:35,010 to do so and second, redistribution 1313 00:56:35,010 --> 00:56:36,570 dealing with the fact that the burden 1314 00:56:36,570 --> 00:56:38,580 is particularly high among the poor in terms 1315 00:56:38,580 --> 00:56:41,345 of how much money to pay. 1316 00:56:41,345 --> 00:56:42,970 You kind of want it to be the opposite. 1317 00:56:42,970 --> 00:56:46,650 You want the rich to pay more taxes rather than the poor, 1318 00:56:46,650 --> 00:56:48,420 and some of these taxes do the opposite. 1319 00:56:48,420 --> 00:56:49,590 And then there's a question of, how 1320 00:56:49,590 --> 00:56:50,940 do you weigh these two objectives, 1321 00:56:50,940 --> 00:56:52,920 and how you can maybe undo some of the effects 1322 00:56:52,920 --> 00:56:55,800 that you're doing? 1323 00:56:55,800 --> 00:56:56,612 Yes? 1324 00:56:56,612 --> 00:56:57,968 AUDIENCE: [INAUDIBLE]? 1325 00:57:02,722 --> 00:57:06,080 FRANK SCHILBACH: I think these are simple-- 1326 00:57:06,080 --> 00:57:09,680 these are just simple surveys about people's happiness 1327 00:57:09,680 --> 00:57:10,730 and so on and so forth. 1328 00:57:10,730 --> 00:57:11,896 Is that your question? 1329 00:57:11,896 --> 00:57:13,294 AUDIENCE: [INAUDIBLE]? 1330 00:57:16,192 --> 00:57:17,900 FRANK SCHILBACH: Oh, I think this is just 1331 00:57:17,900 --> 00:57:23,390 trying to predict it using the demographics. 1332 00:57:23,390 --> 00:57:25,970 So I think one of the key-- 1333 00:57:25,970 --> 00:57:28,640 the key weakness of the study is two things. 1334 00:57:28,640 --> 00:57:31,010 They just do not have data on like, 1335 00:57:31,010 --> 00:57:34,460 does the respondent actually smoke, so they only know, 1336 00:57:34,460 --> 00:57:36,292 essentially, household characteristics, how 1337 00:57:36,292 --> 00:57:38,000 rich are they are, and so on and so forth 1338 00:57:38,000 --> 00:57:40,310 from, essentially, a bunch of survey questions. 1339 00:57:40,310 --> 00:57:44,370 But there's no actual question of, do you smoke or not? 1340 00:57:44,370 --> 00:57:46,220 And so that's why, in part, this paper, 1341 00:57:46,220 --> 00:57:49,190 in fact, is not published in a top journal, the reason 1342 00:57:49,190 --> 00:57:51,080 being that exactly that's the issue. 1343 00:57:51,080 --> 00:57:54,050 A, you don't actually know whether the person smokes. 1344 00:57:54,050 --> 00:57:57,562 B, it could be that it's a spouse of a smoker that's 1345 00:57:57,562 --> 00:57:58,520 answering the question. 1346 00:57:58,520 --> 00:58:01,010 So we kind of know that the propensity of that household 1347 00:58:01,010 --> 00:58:03,980 to smoke from other data sets when we try to just predict 1348 00:58:03,980 --> 00:58:07,280 smoking using these demographics is high, 1349 00:58:07,280 --> 00:58:09,290 but we don't actually know in the specific cases 1350 00:58:09,290 --> 00:58:10,730 are people smoking, which makes it 1351 00:58:10,730 --> 00:58:16,170 sort of vulnerable to these kinds of concerns, yeah. 1352 00:58:16,170 --> 00:58:16,830 OK. 1353 00:58:16,830 --> 00:58:20,400 Let me talk briefly about a fertilizer 1354 00:58:20,400 --> 00:58:24,100 and then summarize a little bit the evidence that we saw. 1355 00:58:24,100 --> 00:58:26,820 So this is a paper by Esther Duflo and Michael Kremer, 1356 00:58:26,820 --> 00:58:30,390 who recently won a Nobel Prize in economics. 1357 00:58:30,390 --> 00:58:33,220 Esther Duflo is in the Economics Department at MIT as well, 1358 00:58:33,220 --> 00:58:35,020 and Michael Kremer is at Harvard. 1359 00:58:35,020 --> 00:58:40,230 It's a paper with John Robinson from 2011 in the AR. 1360 00:58:40,230 --> 00:58:41,970 What this paper looks at is fertilizer 1361 00:58:41,970 --> 00:58:44,250 used in Western Kenya, and this is 1362 00:58:44,250 --> 00:58:46,597 one of these technologies in developing countries 1363 00:58:46,597 --> 00:58:48,180 where there's quite a bit of evidence, 1364 00:58:48,180 --> 00:58:51,630 and people think the returns to using fertilizer are high. 1365 00:58:51,630 --> 00:58:53,550 That's a thing that farmers could use. 1366 00:58:53,550 --> 00:58:54,780 Essentially, it would increase your yields. 1367 00:58:54,780 --> 00:58:57,030 The returns are on the order of magnitude of like 50%, 1368 00:58:57,030 --> 00:59:01,210 100% often over the course of like a few months. 1369 00:59:01,210 --> 00:59:05,230 So it's a really highly profitable type of investment. 1370 00:59:05,230 --> 00:59:07,140 It's also like a divisible investment. 1371 00:59:07,140 --> 00:59:09,510 It's not like a cow, or like a tractor, or something 1372 00:59:09,510 --> 00:59:11,910 where you have to save a lot of money to actually buy it. 1373 00:59:11,910 --> 00:59:14,130 This is a thing where you can buy like a kilo or 5 1374 00:59:14,130 --> 00:59:15,360 kilos of fertilizer. 1375 00:59:15,360 --> 00:59:21,853 You can lose very small amounts of money and then start saving. 1376 00:59:21,853 --> 00:59:23,520 And if the returns are really that high, 1377 00:59:23,520 --> 00:59:27,140 people are just using it and then become richer over time. 1378 00:59:27,140 --> 00:59:29,573 Now, there's a persistent finding 1379 00:59:29,573 --> 00:59:30,990 that fertilizer use, in particular 1380 00:59:30,990 --> 00:59:32,880 in sub-Saharan Africa, tends to be low, 1381 00:59:32,880 --> 00:59:36,360 including in Western Kenya and the area of this study. 1382 00:59:36,360 --> 00:59:38,880 Now, one explanation that people often say 1383 00:59:38,880 --> 00:59:42,345 is that farmers would like to purchase fertilizer. 1384 00:59:42,345 --> 00:59:44,220 When you ask them around the time of harvest, 1385 00:59:44,220 --> 00:59:45,390 which is when they have a lot of money-- 1386 00:59:45,390 --> 00:59:47,040 so there are these harvest cycles. 1387 00:59:47,040 --> 00:59:49,777 Farmers farm maize in this area. 1388 00:59:49,777 --> 00:59:51,610 When you ask them around the time of harvest 1389 00:59:51,610 --> 00:59:53,970 when people have lots of money and ask them about their plans 1390 00:59:53,970 --> 00:59:55,560 to purchase fertilizer, they would say, oh, yeah, 1391 00:59:55,560 --> 00:59:57,393 I'd love to put this fertilizer, it's great, 1392 00:59:57,393 --> 00:59:58,600 and so on and so forth. 1393 00:59:58,600 --> 01:00:00,480 But then lots of stuff comes up. 1394 01:00:00,480 --> 01:00:02,610 Lots of expenses pile up, school fees, 1395 01:00:02,610 --> 01:00:06,090 and friends ask for money, and other festivals 1396 01:00:06,090 --> 01:00:07,352 and so on and so forth. 1397 01:00:07,352 --> 01:00:09,060 So then when the actual new season comes, 1398 01:00:09,060 --> 01:00:11,018 when the planting time or the top dressing time 1399 01:00:11,018 --> 01:00:12,810 comes when the maize is a little bit higher 1400 01:00:12,810 --> 01:00:15,060 and you're supposed to add fertilizer to your maize, 1401 01:00:15,060 --> 01:00:17,670 then people don't have any money left. 1402 01:00:17,670 --> 01:00:20,468 Now, one way to overcome this issue 1403 01:00:20,468 --> 01:00:23,010 was-- essentially, the best kind of commitment and some sense 1404 01:00:23,010 --> 01:00:25,930 is to just do the thing that you're planning to do, 1405 01:00:25,930 --> 01:00:28,690 and that's exactly what they're doing in this experiment. 1406 01:00:28,690 --> 01:00:31,380 So what they're doing is they vary the timing of the purchase 1407 01:00:31,380 --> 01:00:34,290 decision and offer people time-limited discounts 1408 01:00:34,290 --> 01:00:36,030 around the time of harvest. 1409 01:00:36,030 --> 01:00:38,880 So around the time of harvest, they give a very small discount 1410 01:00:38,880 --> 01:00:41,490 and say, you can essentially buy fertilizer, 1411 01:00:41,490 --> 01:00:43,230 and that's going to be delivered to you 1412 01:00:43,230 --> 01:00:45,480 around the time of harvest. 1413 01:00:45,480 --> 01:00:47,250 But you have to buy it right now. 1414 01:00:47,250 --> 01:00:50,610 And they get these different forms of discounts, 1415 01:00:50,610 --> 01:00:52,695 but essentially, the discounts tend to be small. 1416 01:00:52,695 --> 01:00:54,570 And the key part here is that these discounts 1417 01:00:54,570 --> 01:00:56,370 are time-limited, as in like it's only 1418 01:00:56,370 --> 01:00:58,980 around the time of harvest, so you can only do it right now. 1419 01:00:58,980 --> 01:01:01,770 You cannot procrastinate it in any way and say, oh, next week, 1420 01:01:01,770 --> 01:01:04,390 or in two weeks, or three weeks from now, I'm going to do it. 1421 01:01:04,390 --> 01:01:06,390 And so that's essentially a form of a commitment 1422 01:01:06,390 --> 01:01:07,990 device in the sense of, in some ways, 1423 01:01:07,990 --> 01:01:10,260 the best commitment device of, when you want something 1424 01:01:10,260 --> 01:01:12,260 to get done, you can just do it, and then you're 1425 01:01:12,260 --> 01:01:13,020 essentially done. 1426 01:01:13,020 --> 01:01:16,650 So people do these early purchases of fertilizer, 1427 01:01:16,650 --> 01:01:18,833 and nobody resells it or the like. 1428 01:01:18,833 --> 01:01:20,250 That's just not a thing people do. 1429 01:01:20,250 --> 01:01:22,042 In principle, they could obviously do that, 1430 01:01:22,042 --> 01:01:24,510 but that tends to be not the case. 1431 01:01:24,510 --> 01:01:26,070 And what they find, in a sense-- they 1432 01:01:26,070 --> 01:01:29,820 find very large effects of this type of commitment device 1433 01:01:29,820 --> 01:01:31,470 or the early purchase option. 1434 01:01:31,470 --> 01:01:34,350 Even when the discount is very small, very minor-- often 1435 01:01:34,350 --> 01:01:36,330 it's just like delivery to people's homes 1436 01:01:36,330 --> 01:01:38,902 from the fertilizer store, which is not very hard to do. 1437 01:01:38,902 --> 01:01:40,860 So even when this discount is relatively small, 1438 01:01:40,860 --> 01:01:43,770 there's large effects on people fertilizer use. 1439 01:01:43,770 --> 01:01:46,920 And the effects are much larger than like offering people 1440 01:01:46,920 --> 01:01:48,840 discounts around the time of planting. 1441 01:01:48,840 --> 01:01:50,340 So they have another condition where 1442 01:01:50,340 --> 01:01:51,990 they say, around the time of planting, 1443 01:01:51,990 --> 01:01:55,510 when the time comes to actually use fertilizer, 1444 01:01:55,510 --> 01:01:57,648 they have a 50% discount around that time. 1445 01:01:57,648 --> 01:01:59,190 But essentially, that discount almost 1446 01:01:59,190 --> 01:02:01,080 does nothing, the reason being that, when 1447 01:02:01,080 --> 01:02:02,747 people have run out of money, it doesn't 1448 01:02:02,747 --> 01:02:03,930 matter what the price is. 1449 01:02:03,930 --> 01:02:07,230 You can't buy any fertilizer because you're cash-constrained 1450 01:02:07,230 --> 01:02:08,640 and can't buy it. 1451 01:02:08,640 --> 01:02:11,820 Now, that's a very interesting finding, in part for policy, 1452 01:02:11,820 --> 01:02:14,040 because there's also quite a bit of concerns 1453 01:02:14,040 --> 01:02:16,620 about over-adoption, about fertilizer subsidies being 1454 01:02:16,620 --> 01:02:18,790 very expensive, and so on and so forth. 1455 01:02:18,790 --> 01:02:21,330 And now, when you provide time-limited discounts 1456 01:02:21,330 --> 01:02:23,670 around the time of harvest that are very small, 1457 01:02:23,670 --> 01:02:25,140 you can potentially, essentially, 1458 01:02:25,140 --> 01:02:27,480 not spend a lot of money, have large increases 1459 01:02:27,480 --> 01:02:29,010 in fertilizer use among the people 1460 01:02:29,010 --> 01:02:31,177 who have self-control problems, among the people who 1461 01:02:31,177 --> 01:02:33,090 would actually like to use fertilizer. 1462 01:02:33,090 --> 01:02:38,970 If instead you provide discounts around the time of planting 1463 01:02:38,970 --> 01:02:41,820 of 50% or higher, A, that's really expensive to do, 1464 01:02:41,820 --> 01:02:45,660 and governments, in particular in Kenya or sub-Saharan Africa 1465 01:02:45,660 --> 01:02:47,640 in general, tend to not have a lot of money. 1466 01:02:47,640 --> 01:02:49,620 So it's a very expensive thing to do. 1467 01:02:49,620 --> 01:02:52,020 Plus, you're also distorting a lot of decisions 1468 01:02:52,020 --> 01:02:53,760 among exponential discounters. 1469 01:02:53,760 --> 01:02:55,510 What you're doing is, now I'm giving you-- 1470 01:02:55,510 --> 01:02:57,493 I'm making fertilizer much cheaper for you. 1471 01:02:57,493 --> 01:02:59,160 If you don't have a self-control problem 1472 01:02:59,160 --> 01:03:01,950 to start with, you're going to overuse fertilizer because now, 1473 01:03:01,950 --> 01:03:05,020 essentially, fertilizer is like artificially cheap for you, 1474 01:03:05,020 --> 01:03:05,520 essentially. 1475 01:03:05,520 --> 01:03:07,770 So you push some people to overuse fertilizer 1476 01:03:07,770 --> 01:03:11,160 using an actual discount around the time of planting. 1477 01:03:11,160 --> 01:03:13,110 Instead, what this policy does-- essentially, 1478 01:03:13,110 --> 01:03:15,690 it gives very small discounts that are very cheap, 1479 01:03:15,690 --> 01:03:20,167 and it mostly affects people with self-control problems. 1480 01:03:20,167 --> 01:03:21,750 Those are the people who, essentially, 1481 01:03:21,750 --> 01:03:23,710 would procrastinate buying fertilizer. 1482 01:03:23,710 --> 01:03:24,840 They buy it right now. 1483 01:03:24,840 --> 01:03:27,372 They don't spend a lot of money on it, 1484 01:03:27,372 --> 01:03:29,580 and you could have potentially large effects, as they 1485 01:03:29,580 --> 01:03:32,460 have shown in this paper. 1486 01:03:32,460 --> 01:03:34,840 Any questions on this? 1487 01:03:34,840 --> 01:03:35,660 Yeah? 1488 01:03:35,660 --> 01:03:37,758 AUDIENCE: Did they do a part where 1489 01:03:37,758 --> 01:03:40,535 they show that farmers become more sophisticated? 1490 01:03:40,535 --> 01:03:41,920 FRANK SCHILBACH: No, they do not. 1491 01:03:41,920 --> 01:03:45,010 They have, essentially, one-shot decisions for farmers, 1492 01:03:45,010 --> 01:03:48,700 and they sort of have some effects over the course 1493 01:03:48,700 --> 01:03:49,660 of several seasons. 1494 01:03:49,660 --> 01:03:54,560 But it's essentially not the same choices and so on. 1495 01:03:54,560 --> 01:03:57,610 So this is, in some sense, an early paper that doesn't quite 1496 01:03:57,610 --> 01:03:58,480 look into that. 1497 01:03:58,480 --> 01:04:00,910 Some of their work on sophistication and naivete 1498 01:04:00,910 --> 01:04:03,130 comes a little later. 1499 01:04:03,130 --> 01:04:05,770 So let me briefly summarize and talk a little bit about people 1500 01:04:05,770 --> 01:04:08,570 feeling when it comes to using commitment devices. 1501 01:04:08,570 --> 01:04:10,330 So I've shown you several examples 1502 01:04:10,330 --> 01:04:12,040 and there are quite a few more examples 1503 01:04:12,040 --> 01:04:14,700 of demand for commitment in different research studies. 1504 01:04:14,700 --> 01:04:16,450 These are all research studies that people 1505 01:04:16,450 --> 01:04:18,610 have done one way or the other. 1506 01:04:18,610 --> 01:04:20,290 Often, there's a significant share 1507 01:04:20,290 --> 01:04:22,877 of people who demand commitment one way or the other. 1508 01:04:22,877 --> 01:04:25,210 That's [INAUDIBLE] there's a significant share of people 1509 01:04:25,210 --> 01:04:28,330 who struggle with self-control problems in some way, 1510 01:04:28,330 --> 01:04:30,940 and they also are at least partially sophisticated 1511 01:04:30,940 --> 01:04:33,730 because they sort of understand it and demand commitment. 1512 01:04:33,730 --> 01:04:37,810 Now, there's also like a very high variation in the fraction 1513 01:04:37,810 --> 01:04:39,280 of people demanding commitment. 1514 01:04:39,280 --> 01:04:41,470 In some places, it's very high, as I showed you 1515 01:04:41,470 --> 01:04:44,680 in the alcohol examples and in some other examples 1516 01:04:44,680 --> 01:04:46,450 that I haven't necessarily shown you. 1517 01:04:46,450 --> 01:04:49,010 In other examples, like, for example, in the smoking example 1518 01:04:49,010 --> 01:04:51,052 and so on, the demand for commitment tends to be, 1519 01:04:51,052 --> 01:04:52,940 actually, quite low. 1520 01:04:52,940 --> 01:04:53,860 So why is that? 1521 01:04:53,860 --> 01:04:56,500 Well, we don't actually know necessarily. 1522 01:04:56,500 --> 01:05:00,155 There's a few explanations that people have. 1523 01:05:00,155 --> 01:05:01,280 One of them is uncertainty. 1524 01:05:01,280 --> 01:05:03,380 People might be really worried about the risk. 1525 01:05:03,380 --> 01:05:05,110 So if you have lots of uncertainty in your life, 1526 01:05:05,110 --> 01:05:06,910 if you don't know whether next month you're 1527 01:05:06,910 --> 01:05:09,805 going to be smoking or not, or you might have a really-- you 1528 01:05:09,805 --> 01:05:11,770 might have a really stressful month next month 1529 01:05:11,770 --> 01:05:12,810 and you don't probably know whether that's 1530 01:05:12,810 --> 01:05:14,590 going be stressful or not, you might not 1531 01:05:14,590 --> 01:05:16,548 want to choose a commitment device because it's 1532 01:05:16,548 --> 01:05:17,988 very costly for you, potentially, 1533 01:05:17,988 --> 01:05:19,780 because you might start smoking again or do 1534 01:05:19,780 --> 01:05:22,930 other things for reasons that are not necessarily related 1535 01:05:22,930 --> 01:05:23,860 to self-control. 1536 01:05:23,860 --> 01:05:25,450 It's just out of your control. 1537 01:05:25,450 --> 01:05:28,042 Similarly, for example, in the data entry example 1538 01:05:28,042 --> 01:05:29,500 that I showed you, people might not 1539 01:05:29,500 --> 01:05:32,470 choose to have a positive target because they're really 1540 01:05:32,470 --> 01:05:34,960 worried about risk, and in particular, 1541 01:05:34,960 --> 01:05:36,520 one characteristic that was very much 1542 01:05:36,520 --> 01:05:39,310 predictive of whether people choose these commitment 1543 01:05:39,310 --> 01:05:41,290 devices, whether they have kids-- 1544 01:05:41,290 --> 01:05:43,390 the reason being that the kids-- 1545 01:05:43,390 --> 01:05:44,290 they might be sick. 1546 01:05:44,290 --> 01:05:46,970 They might have really bad nights and so on and so forth, 1547 01:05:46,970 --> 01:05:49,345 and so that's an increase-- that's 1548 01:05:49,345 --> 01:05:53,630 the risk of data entry productivity in the future. 1549 01:05:53,630 --> 01:05:55,450 So one part is uncertainty. 1550 01:05:55,450 --> 01:05:56,800 A second part is naivete. 1551 01:05:56,800 --> 01:06:01,180 Partial naivete can get people to not purchase or demand 1552 01:06:01,180 --> 01:06:03,850 certain commitment devices, and naivete might vary it 1553 01:06:03,850 --> 01:06:05,380 quite a bit across settings. 1554 01:06:05,380 --> 01:06:08,920 It might be that you might just be 1555 01:06:08,920 --> 01:06:11,590 like a first-time customer or a first-time user 1556 01:06:11,590 --> 01:06:13,572 of either payday loans or of drinking, 1557 01:06:13,572 --> 01:06:14,530 and smoking, and so on. 1558 01:06:14,530 --> 01:06:15,970 You might be quite naive. 1559 01:06:15,970 --> 01:06:18,430 In contrast, in my setting in Chennai 1560 01:06:18,430 --> 01:06:20,483 when I looked at drinkers, there people 1561 01:06:20,483 --> 01:06:22,150 have been drinking for many, many years. 1562 01:06:22,150 --> 01:06:24,760 In some sense, they're not as naive about their drinking, 1563 01:06:24,760 --> 01:06:27,370 at least overall. 1564 01:06:27,370 --> 01:06:29,742 Another aspect is experience with commitment devices, 1565 01:06:29,742 --> 01:06:31,450 and this is what your question was about. 1566 01:06:31,450 --> 01:06:33,450 Like when you provide people commitment devices, 1567 01:06:33,450 --> 01:06:35,485 maybe it requires some learning over time. 1568 01:06:35,485 --> 01:06:37,360 And again, if you look at different settings, 1569 01:06:37,360 --> 01:06:39,550 people have different types of experience. 1570 01:06:39,550 --> 01:06:41,950 Overall, we don't know the answer to these questions. 1571 01:06:41,950 --> 01:06:43,992 In a sense, we don't actually know exactly what's 1572 01:06:43,992 --> 01:06:46,060 going on why is demand for commitment quite 1573 01:06:46,060 --> 01:06:47,720 different across settings. 1574 01:06:47,720 --> 01:06:50,332 A second question that we don't really know is like, well, 1575 01:06:50,332 --> 01:06:51,790 so there are these research studies 1576 01:06:51,790 --> 01:06:53,975 that look at commitment device and say, OK, 1577 01:06:53,975 --> 01:06:55,600 commitment devices, at least in theory, 1578 01:06:55,600 --> 01:06:58,180 should help some people. 1579 01:06:58,180 --> 01:07:01,090 But we see actually very few real-world examples 1580 01:07:01,090 --> 01:07:03,478 that actually help. 1581 01:07:03,478 --> 01:07:05,020 Yesterday, you mentioned some of them 1582 01:07:05,020 --> 01:07:07,490 that seem to be at least helpful for some of you. 1583 01:07:07,490 --> 01:07:09,475 By the way, if you would email me those, 1584 01:07:09,475 --> 01:07:11,590 I'd like to collect them and report back 1585 01:07:11,590 --> 01:07:16,090 to David Laibson who says uncertainty is the issue. 1586 01:07:16,090 --> 01:07:18,560 But we don't have it really figured out in some sense. 1587 01:07:18,560 --> 01:07:20,443 If commitment devices were really that great, 1588 01:07:20,443 --> 01:07:22,360 you would think that it really would take off. 1589 01:07:22,360 --> 01:07:24,318 There would be a company that would really make 1590 01:07:24,318 --> 01:07:25,610 a lot of money with using them. 1591 01:07:25,610 --> 01:07:27,693 There are some companies that we see in the world, 1592 01:07:27,693 --> 01:07:29,290 but it's not really-- it can't really 1593 01:07:29,290 --> 01:07:31,330 say it has really fully taken off, 1594 01:07:31,330 --> 01:07:33,730 perhaps some companies related to cell phone 1595 01:07:33,730 --> 01:07:35,740 usage, or computer usage, and so on, 1596 01:07:35,740 --> 01:07:39,430 but it's not as prevalent and widely used as we might expect, 1597 01:07:39,430 --> 01:07:42,400 given the theory literature that we have. 1598 01:07:42,400 --> 01:07:45,250 Now, one important part that's really 1599 01:07:45,250 --> 01:07:49,840 another potential contributor to [INAUDIBLE] demand 1600 01:07:49,840 --> 01:07:54,040 for commitment is that people might not only 1601 01:07:54,040 --> 01:07:56,200 be naive in a sense that they might not understand 1602 01:07:56,200 --> 01:07:59,140 the value of commitment devices and then that might prevent 1603 01:07:59,140 --> 01:08:02,410 them from taking them up, but even conditional demand 1604 01:08:02,410 --> 01:08:04,540 commitments they might fail. 1605 01:08:04,540 --> 01:08:06,460 And there's a couple of examples here. 1606 01:08:06,460 --> 01:08:12,310 For example, in the Philippines savings accounts example 1607 01:08:12,310 --> 01:08:15,760 that I showed you, the SEED commitment savings account-- 1608 01:08:15,760 --> 01:08:18,279 and John has a very nice paper that shows that. 1609 01:08:18,279 --> 01:08:22,000 In fact, when you look at the same setting, 1610 01:08:22,000 --> 01:08:25,149 a similar experiment, offering people this commitment contract 1611 01:08:25,149 --> 01:08:27,100 has positive effects on people's savings. 1612 01:08:27,100 --> 01:08:29,020 So people do actually save more on average 1613 01:08:29,020 --> 01:08:30,680 compared to a control group. 1614 01:08:30,680 --> 01:08:32,560 However, there's also a pretty large fraction 1615 01:08:32,560 --> 01:08:35,819 of people who fail in reaching their goals, 1616 01:08:35,819 --> 01:08:40,060 and then often they have to pay some penalty for doing that. 1617 01:08:40,060 --> 01:08:42,643 Now, that's really bad marketing for your product. 1618 01:08:42,643 --> 01:08:44,310 If you think about selling this product, 1619 01:08:44,310 --> 01:08:47,460 if half of your customers are really unhappy about it, 1620 01:08:47,460 --> 01:08:49,182 that's going to be very tricky to do. 1621 01:08:49,182 --> 01:08:51,390 So there's lots of questions about how to best design 1622 01:08:51,390 --> 01:08:54,180 commitment devices, how can we perhaps target 1623 01:08:54,180 --> 01:08:57,310 customers who might not fail and perhaps design them differently 1624 01:08:57,310 --> 01:08:59,310 for different people, but we haven't necessarily 1625 01:08:59,310 --> 01:09:01,290 figured out how to do that. 1626 01:09:01,290 --> 01:09:04,200 Liang Bai and Gautam Rao and co-authors 1627 01:09:04,200 --> 01:09:08,490 have a similar example of a health commitment product 1628 01:09:08,490 --> 01:09:10,229 in India where, essentially, they're 1629 01:09:10,229 --> 01:09:14,590 trying to incentivize people to go to hypertension doctor visit 1630 01:09:14,590 --> 01:09:17,473 to get checkups for their hypertension. 1631 01:09:17,473 --> 01:09:19,140 Lots of people say they would like to go 1632 01:09:19,140 --> 01:09:20,370 and so on and so forth. 1633 01:09:20,370 --> 01:09:22,260 You offer them commitment devices 1634 01:09:22,260 --> 01:09:25,649 to essentially provide themselves incentives to go. 1635 01:09:25,649 --> 01:09:28,540 Pretty few people actually sign up in the first place, 1636 01:09:28,540 --> 01:09:30,729 which is also consistent with some form of naivete 1637 01:09:30,729 --> 01:09:34,050 and so on and so forth, but even among the people who sign up, 1638 01:09:34,050 --> 01:09:35,910 lots of people actually fail. 1639 01:09:35,910 --> 01:09:38,729 So you end up having people who essentially set themselves 1640 01:09:38,729 --> 01:09:40,029 incentives to go to the doctor. 1641 01:09:40,029 --> 01:09:42,390 They're willing to pay for going in the future 1642 01:09:42,390 --> 01:09:43,979 or setting themselves incentives to go 1643 01:09:43,979 --> 01:09:46,423 to the doctor in the future, and then they 1644 01:09:46,423 --> 01:09:47,340 have these incentives. 1645 01:09:47,340 --> 01:09:49,507 And then the incentives are actually not successful, 1646 01:09:49,507 --> 01:09:50,790 and they end up not going. 1647 01:09:50,790 --> 01:09:54,870 Again, that's very much consistent 1648 01:09:54,870 --> 01:09:57,330 with naivete or at least partial naivete 1649 01:09:57,330 --> 01:10:00,510 or partial sophistication. 1650 01:10:00,510 --> 01:10:02,700 And there's many questions now about, 1651 01:10:02,700 --> 01:10:06,000 if that's the case, how do we design better commitment 1652 01:10:06,000 --> 01:10:08,040 devices that work for more people that 1653 01:10:08,040 --> 01:10:09,960 are more successful for people, and how do we 1654 01:10:09,960 --> 01:10:12,270 avoid that some people sign up for them when they, 1655 01:10:12,270 --> 01:10:14,730 in fact, are doing worse because, again, that's really 1656 01:10:14,730 --> 01:10:16,260 bad marketing for your product. 1657 01:10:16,260 --> 01:10:18,757 And of course, it's bad for the people who are doing worse. 1658 01:10:18,757 --> 01:10:21,090 They would be, essentially, better off if they had never 1659 01:10:21,090 --> 01:10:27,510 been offered this device, at least ex post. 1660 01:10:27,510 --> 01:10:30,810 Any questions on this? 1661 01:10:30,810 --> 01:10:33,158 So then, very briefly, there's a second view 1662 01:10:33,158 --> 01:10:35,700 that you might have is to say, well, maybe commitment devices 1663 01:10:35,700 --> 01:10:36,810 are just not that helpful. 1664 01:10:36,810 --> 01:10:39,060 In some sense, we have not figured out how to do this. 1665 01:10:39,060 --> 01:10:41,610 Companies have had lots of time to figure this out. 1666 01:10:41,610 --> 01:10:43,223 Researchers have not figured this out 1667 01:10:43,223 --> 01:10:44,640 in a sense of like-- well, we have 1668 01:10:44,640 --> 01:10:46,380 lots of evidence of demand for commitment 1669 01:10:46,380 --> 01:10:48,795 in terms of people wanting commitment devices. 1670 01:10:48,795 --> 01:10:51,420 We have, actually, not that much evidence of commitment devices 1671 01:10:51,420 --> 01:10:55,770 being useful in the real world in the sense of persistently 1672 01:10:55,770 --> 01:10:58,188 showing that they make things better. 1673 01:10:58,188 --> 01:10:59,730 There are some examples, to be clear. 1674 01:10:59,730 --> 01:11:01,920 For example, the data entry example 1675 01:11:01,920 --> 01:11:04,200 does show that people are more productive at work, 1676 01:11:04,200 --> 01:11:06,618 and probably employers should do that. 1677 01:11:06,618 --> 01:11:08,160 But another view would be like, well, 1678 01:11:08,160 --> 01:11:10,290 if commitment devices are not really helping-- 1679 01:11:10,290 --> 01:11:13,230 so think of commitment devices as a way of providing people 1680 01:11:13,230 --> 01:11:16,080 some choice options for given self-control problems or given 1681 01:11:16,080 --> 01:11:17,280 naivete. 1682 01:11:17,280 --> 01:11:19,750 Now, there's different things you could do. 1683 01:11:19,750 --> 01:11:21,240 You could do some things in terms 1684 01:11:21,240 --> 01:11:23,153 of trying to address people's naivete, 1685 01:11:23,153 --> 01:11:25,570 and the problems ask you a little bit to think about that. 1686 01:11:25,570 --> 01:11:28,740 How do you actually get people to be 1687 01:11:28,740 --> 01:11:31,920 less naive if that depresses demand for commitment? 1688 01:11:31,920 --> 01:11:34,890 Another way to say is you could try to essentially try 1689 01:11:34,890 --> 01:11:36,540 to reduce people's present bias or get 1690 01:11:36,540 --> 01:11:38,923 at the sources of people's present bias overall. 1691 01:11:38,923 --> 01:11:40,590 There's different ways, and we have only 1692 01:11:40,590 --> 01:11:42,720 started thinking about or coming up 1693 01:11:42,720 --> 01:11:45,240 with some potential explanations, 1694 01:11:45,240 --> 01:11:49,380 but for example, a paper that we have on sleep 1695 01:11:49,380 --> 01:11:51,240 finds that when people are napping 1696 01:11:51,240 --> 01:11:55,260 in the afternoon, which some of you do also in my class-- 1697 01:11:55,260 --> 01:12:00,180 that reduces your present bias and [INAUDIBLE] task and also 1698 01:12:00,180 --> 01:12:01,020 savings. 1699 01:12:01,020 --> 01:12:03,900 So that's to say, perhaps sleep, perhaps alcohol consumption, 1700 01:12:03,900 --> 01:12:06,540 perhaps other sort of factors might 1701 01:12:06,540 --> 01:12:09,540 affect people's present bias, so if you could understand 1702 01:12:09,540 --> 01:12:12,300 these underlying factors and if you could tackle them, 1703 01:12:12,300 --> 01:12:15,120 that would be another way of getting around, potentially, 1704 01:12:15,120 --> 01:12:17,280 self-control problems causing trouble 1705 01:12:17,280 --> 01:12:20,400 by getting at the root of this particular issue 1706 01:12:20,400 --> 01:12:22,690 if you think that's an issue overall. 1707 01:12:22,690 --> 01:12:25,480 There's some other research that I find quite interesting, 1708 01:12:25,480 --> 01:12:31,530 which is the idea that present bias is very much related 1709 01:12:31,530 --> 01:12:33,150 to attention in the sense that people 1710 01:12:33,150 --> 01:12:36,333 pay too much attention to the present compared to the future. 1711 01:12:36,333 --> 01:12:38,250 So now, if you could sort make the future more 1712 01:12:38,250 --> 01:12:40,750 salient in certain ways, if you could make people think more 1713 01:12:40,750 --> 01:12:42,750 about the future, how things will be like, 1714 01:12:42,750 --> 01:12:44,880 they might behave more patiently. 1715 01:12:44,880 --> 01:12:47,760 So Hal Hershfield has a series of papers, 1716 01:12:47,760 --> 01:12:51,030 and Hal Hershfield is the guy on the left, the picture that you 1717 01:12:51,030 --> 01:12:56,893 see, who used computer software to make people look older. 1718 01:12:56,893 --> 01:12:58,310 So on the right of the screen, you 1719 01:12:58,310 --> 01:13:01,680 see Hal Hershfield like another 40 years old or something, 1720 01:13:01,680 --> 01:13:04,050 and so what they did is they tried to show people, 1721 01:13:04,050 --> 01:13:06,360 essentially, pictures of their old ourselves. 1722 01:13:06,360 --> 01:13:08,910 And this is how you're going to look like in 40, 50 years 1723 01:13:08,910 --> 01:13:11,370 from now, and then think about how 1724 01:13:11,370 --> 01:13:14,320 things might be like in terms of your health 1725 01:13:14,320 --> 01:13:15,297 and so on and so forth. 1726 01:13:15,297 --> 01:13:17,130 And then get people to make savings choices. 1727 01:13:17,130 --> 01:13:18,630 And there's some evidence, at least, 1728 01:13:18,630 --> 01:13:21,330 that increasing people's focus and attention to the future 1729 01:13:21,330 --> 01:13:24,020 might actually make them more patient, 1730 01:13:24,020 --> 01:13:25,860 and there's quite a bit of work of people 1731 01:13:25,860 --> 01:13:28,080 trying to do that right now. 1732 01:13:28,080 --> 01:13:29,880 So just to give you a brief summary-- 1733 01:13:29,880 --> 01:13:33,090 I think I said already almost everything that's to be said-- 1734 01:13:33,090 --> 01:13:36,240 exponential discounting is like a workhorse model 1735 01:13:36,240 --> 01:13:37,230 of modern economics. 1736 01:13:37,230 --> 01:13:38,490 It's extremely important. 1737 01:13:38,490 --> 01:13:41,070 And just to be clear, it's helpful for many applications, 1738 01:13:41,070 --> 01:13:42,828 and it's been a very successful model. 1739 01:13:42,828 --> 01:13:45,120 There's some limitations and some things that we cannot 1740 01:13:45,120 --> 01:13:48,270 explain, so quasi-hyperbolic discounting or some forms 1741 01:13:48,270 --> 01:13:51,690 of different discounting models might be able to improve 1742 01:13:51,690 --> 01:13:55,163 the predictability of our models. 1743 01:13:55,163 --> 01:13:56,580 Commitment devices, in particular, 1744 01:13:56,580 --> 01:13:59,118 may help improve outcomes, but we haven't quite 1745 01:13:59,118 --> 01:14:00,660 figured out how to make them actually 1746 01:14:00,660 --> 01:14:01,980 most useful for the world. 1747 01:14:01,980 --> 01:14:05,740 There's lots of research going on in this area. 1748 01:14:05,740 --> 01:14:08,230 Next week, we're going to talk about risk preferences. 1749 01:14:08,230 --> 01:14:10,260 In particular, I'd like you to read the paper 1750 01:14:10,260 --> 01:14:12,360 by Rabin and Thaler, 2001. 1751 01:14:12,360 --> 01:14:14,160 In recitation this week, there will be more 1752 01:14:14,160 --> 01:14:16,230 on time preferences, in particular related 1753 01:14:16,230 --> 01:14:19,380 to the next problem set, and then 1754 01:14:19,380 --> 01:14:21,390 sort of an introduction to risk preferences. 1755 01:14:21,390 --> 01:14:24,270 If you don't remember what risk preferences are, 1756 01:14:24,270 --> 01:14:26,280 you should very much come to recitation and get 1757 01:14:26,280 --> 01:14:27,490 a refresher of that. 1758 01:14:27,490 --> 01:14:29,480 Thank you very much.