1 00:00:00,135 --> 00:00:02,490 The following content is provided under a Creative 2 00:00:02,490 --> 00:00:04,059 Commons license. 3 00:00:04,059 --> 00:00:06,330 Your support will help MIT OpenCourseWare 4 00:00:06,330 --> 00:00:10,690 continue to offer high-quality educational resources for free. 5 00:00:10,690 --> 00:00:13,350 To make a donation or view additional materials 6 00:00:13,350 --> 00:00:17,280 from hundreds of MIT courses, visit MIT OpenCourseWare 7 00:00:17,280 --> 00:00:18,480 at ocw.mit.edu. 8 00:00:28,016 --> 00:00:33,070 PROFESSOR: So today is the fourth and last lecture 9 00:00:33,070 --> 00:00:36,640 on the sort of microfounded macro models. 10 00:00:36,640 --> 00:00:39,700 And then next class, we will turn 11 00:00:39,700 --> 00:00:45,640 to a section on measurement, which is like a halfway point, 12 00:00:45,640 --> 00:00:48,190 or not really, in terms of total number of lectures. 13 00:00:48,190 --> 00:00:52,720 But we'll get into them-- to the micro evidence. 14 00:00:52,720 --> 00:00:55,540 I must say, this is sometime-- 15 00:00:55,540 --> 00:00:59,140 I said this last time-- in reverse of what I sometimes do. 16 00:00:59,140 --> 00:01:02,440 I often start with the micro stuff and build it up. 17 00:01:02,440 --> 00:01:04,150 But this isn't a bad idea, because you'll 18 00:01:04,150 --> 00:01:09,070 get to see this way how different assumptions matter 19 00:01:09,070 --> 00:01:11,180 within each of these models as you compare 20 00:01:11,180 --> 00:01:12,430 the predicted solutions. 21 00:01:12,430 --> 00:01:17,590 So it hopefully heightens your sensitivity to assumptions. 22 00:01:17,590 --> 00:01:19,990 And then we'll get into the microdata 23 00:01:19,990 --> 00:01:22,570 and really scrutinize what we see there. 24 00:01:22,570 --> 00:01:28,510 This paper, in particular, is part of that conversation. 25 00:01:28,510 --> 00:01:30,580 The way it's organized in terms of the lectures 26 00:01:30,580 --> 00:01:34,840 is this paper features mostly something 27 00:01:34,840 --> 00:01:37,540 other than limited commitment. 28 00:01:37,540 --> 00:01:42,340 Almost every single model has had sort of very limited credit 29 00:01:42,340 --> 00:01:46,360 or financing constraints, collateral constraints, 30 00:01:46,360 --> 00:01:52,060 basically putting a damper on sudden transitions 31 00:01:52,060 --> 00:01:55,150 and generating growth paths. 32 00:01:55,150 --> 00:02:00,460 And here, we're going to be looking at transition 33 00:02:00,460 --> 00:02:02,362 somewhat-- also, steady states. 34 00:02:02,362 --> 00:02:03,820 The main thing is we're going to be 35 00:02:03,820 --> 00:02:09,940 featuring on other alternative underpinnings of the models-- 36 00:02:09,940 --> 00:02:12,850 costly state verification, adverse selection, 37 00:02:12,850 --> 00:02:14,530 moral hazard. 38 00:02:14,530 --> 00:02:19,600 Again, the irony, for those of you coming down 39 00:02:19,600 --> 00:02:21,370 from the advanced macro class, is 40 00:02:21,370 --> 00:02:26,410 that Alp was featuring costly state verification up there 41 00:02:26,410 --> 00:02:27,570 early in a lecture. 42 00:02:27,570 --> 00:02:32,200 And these things are somehow related, 43 00:02:32,200 --> 00:02:37,860 in the sense that the dynamics that are created with the debt 44 00:02:37,860 --> 00:02:41,760 contracts and so on have to do with what you assume 45 00:02:41,760 --> 00:02:45,840 about the information structure, as well as adverse selection 46 00:02:45,840 --> 00:02:48,280 and so on. 47 00:02:48,280 --> 00:02:53,460 So you'll be perhaps a little relieved 48 00:02:53,460 --> 00:02:55,500 to know there isn't as much material 49 00:02:55,500 --> 00:03:00,090 on the front end of this lecture as there was last time. 50 00:03:00,090 --> 00:03:03,210 That's kind of a sad comment on the literature. 51 00:03:03,210 --> 00:03:06,390 Almost everybody assumes limited commitment, 52 00:03:06,390 --> 00:03:08,930 so there really wasn't much to choose from. 53 00:03:08,930 --> 00:03:14,640 We sort of scoured the earth to find some relevant papers 54 00:03:14,640 --> 00:03:17,250 in this genre. 55 00:03:17,250 --> 00:03:20,610 So there's three, and the third one 56 00:03:20,610 --> 00:03:24,090 is what I'm going to focus my attention on, the tree, so 57 00:03:24,090 --> 00:03:25,770 to speak, of the lecture. 58 00:03:25,770 --> 00:03:28,770 And now we'll look at several trees. 59 00:03:28,770 --> 00:03:31,420 The first two are a bit different. 60 00:03:31,420 --> 00:03:33,510 This one is Greenwood, Sanchez, and Wang-- 61 00:03:33,510 --> 00:03:37,440 Jeremy Greenwood, as of Greenwood and Jovanavic, 62 00:03:37,440 --> 00:03:43,170 which was a key feature of one of the first two models 63 00:03:43,170 --> 00:03:47,490 we talked about in lecture 2. 64 00:03:47,490 --> 00:03:51,480 And he's going to focus on Levine's comment. 65 00:03:51,480 --> 00:03:55,170 It's kind of cool the way these things come together. 66 00:03:55,170 --> 00:03:57,870 Levine worried that the finance causes growth. 67 00:03:57,870 --> 00:04:00,200 Empirical literature doesn't have 68 00:04:00,200 --> 00:04:06,510 a microfounded view of what really impedes intermediation. 69 00:04:10,440 --> 00:04:13,650 So that's what this paper is about. 70 00:04:13,650 --> 00:04:16,589 And it's going to be sort of a costly state verification 71 00:04:16,589 --> 00:04:22,140 point of view with diminishing returns and exogenous, 72 00:04:22,140 --> 00:04:23,250 technological progress. 73 00:04:26,190 --> 00:04:28,320 There's a telling sentence. 74 00:04:28,320 --> 00:04:30,090 In the paper, Uganda could more than 75 00:04:30,090 --> 00:04:32,040 double its output if it would adopt best 76 00:04:32,040 --> 00:04:36,150 practice in the financial sector, which is something 77 00:04:36,150 --> 00:04:38,610 like the US or Luxembourg. 78 00:04:38,610 --> 00:04:40,740 Although, there are other things in the model, 79 00:04:40,740 --> 00:04:43,950 so that that's not enough to get it up to, 80 00:04:43,950 --> 00:04:47,190 quote unquote, "its total potential output." 81 00:04:47,190 --> 00:04:49,170 Again, you see this language of gaps 82 00:04:49,170 --> 00:04:52,170 here that we focused on earlier in one 83 00:04:52,170 --> 00:04:55,920 of those financial possibilities frontier. 84 00:04:55,920 --> 00:04:58,380 So this is very much trying to get 85 00:04:58,380 --> 00:05:03,970 at gaps in terms of the underlying model. 86 00:05:03,970 --> 00:05:06,345 AUDIENCE: What are these other factors? 87 00:05:06,345 --> 00:05:06,970 PROFESSOR: Hmm? 88 00:05:06,970 --> 00:05:08,595 AUDIENCE: What are these other factors? 89 00:05:08,595 --> 00:05:11,280 PROFESSOR: There's going to be some technological progress 90 00:05:11,280 --> 00:05:14,430 that varies across countries, which is pretty standard stuff. 91 00:05:17,310 --> 00:05:19,140 And then this is kind of layered in, 92 00:05:19,140 --> 00:05:21,450 or it's actually more of the heart of the model, 93 00:05:21,450 --> 00:05:23,496 given that other. 94 00:05:23,496 --> 00:05:27,270 Well, we've seen that was done in the China paper as well. 95 00:05:29,820 --> 00:05:34,530 So another interesting thing, 29% of US growth 96 00:05:34,530 --> 00:05:38,610 is due to improvements in financial intermediation. 97 00:05:38,610 --> 00:05:42,090 So this paper both has transition paths in it, 98 00:05:42,090 --> 00:05:43,770 as well as steady states. 99 00:05:43,770 --> 00:05:46,750 This is a comment about steady state. 100 00:05:46,750 --> 00:05:50,830 There's exogenous technological progress going on, 101 00:05:50,830 --> 00:05:53,490 not at the level of the sort of TFP shocks 102 00:05:53,490 --> 00:05:55,530 that are in front of the production function, 103 00:05:55,530 --> 00:05:59,400 but in terms of the monitoring technologies which 104 00:05:59,400 --> 00:06:04,410 intermediaries can use to verify some private information. 105 00:06:04,410 --> 00:06:07,060 And that's where the technological progress 106 00:06:07,060 --> 00:06:11,970 is that's driving, say, steady state US growth rate according 107 00:06:11,970 --> 00:06:17,001 to their calibrated version of the model. 108 00:06:17,001 --> 00:06:20,550 Now, there are two key variables here that they focus a lot 109 00:06:20,550 --> 00:06:24,075 on-- the interest rate spread and the capital output ratio. 110 00:06:28,590 --> 00:06:30,950 So let's look at them. 111 00:06:34,370 --> 00:06:37,385 These slides are interest rate spreads. 112 00:06:40,490 --> 00:06:41,600 This is the US. 113 00:06:41,600 --> 00:06:44,000 That's Taiwan. 114 00:06:44,000 --> 00:06:47,450 This is the spread in the US, maybe 3%, 115 00:06:47,450 --> 00:06:50,300 maybe declining, but not so clear-- 116 00:06:50,300 --> 00:06:52,010 modestly so. 117 00:06:52,010 --> 00:06:57,150 And this is sort of the capital to GDP ratio. 118 00:06:57,150 --> 00:07:01,340 And it's paired to Taiwan, where it's a bit more-- 119 00:07:01,340 --> 00:07:02,870 at first at least-- 120 00:07:02,870 --> 00:07:04,640 easier to interpret. 121 00:07:04,640 --> 00:07:09,110 This spread is high and declining. 122 00:07:09,110 --> 00:07:11,350 So this is the transition part. 123 00:07:11,350 --> 00:07:15,710 This is Taiwan experiencing technological progress that's 124 00:07:15,710 --> 00:07:21,820 lowering, or even I guess it was in the steady state, 125 00:07:21,820 --> 00:07:24,280 at different levels of technological progress 126 00:07:24,280 --> 00:07:25,570 in monitoring output. 127 00:07:29,280 --> 00:07:32,660 And as that happens, as the intermediation system becomes 128 00:07:32,660 --> 00:07:38,420 more efficient, capital is increasing. 129 00:07:38,420 --> 00:07:41,770 You can make better use of capital 130 00:07:41,770 --> 00:07:45,080 because banks have better information, 131 00:07:45,080 --> 00:07:46,880 and they can lend more appropriately 132 00:07:46,880 --> 00:07:51,050 and adjust for risk more appropriately. 133 00:07:51,050 --> 00:07:53,780 Now, even then, this diagram is kind of striking. 134 00:07:53,780 --> 00:07:59,930 Because you can see that this is steadily rising as the spread 135 00:07:59,930 --> 00:08:00,700 is coming down. 136 00:08:00,700 --> 00:08:07,740 But relative to US, the capital to GDP ratio is quite high. 137 00:08:07,740 --> 00:08:09,195 So you know, just be mindful. 138 00:08:11,810 --> 00:08:16,400 There's two variables here, and one of them is GDP. 139 00:08:16,400 --> 00:08:21,080 So low levels of GDP kind of make that ratio high, 140 00:08:21,080 --> 00:08:25,130 and that's going on and also moving these variables 141 00:08:25,130 --> 00:08:27,669 as you compare over time and over countries. 142 00:08:32,669 --> 00:08:40,080 So the idea is that firms, banks, really 143 00:08:40,080 --> 00:08:42,120 monitor cash flows. 144 00:08:42,120 --> 00:08:45,330 The efficiency depends on the amount of resources 145 00:08:45,330 --> 00:08:48,870 you devote to monitoring and to the productivity 146 00:08:48,870 --> 00:08:52,710 of that monitoring technology. 147 00:08:52,710 --> 00:08:56,460 And as well, firms have differences ex ante 148 00:08:56,460 --> 00:09:00,180 in the structure of returns that they face. 149 00:09:03,040 --> 00:09:06,840 So they get the usual sort of why 150 00:09:06,840 --> 00:09:11,100 money doesn't flow from low productive firms 151 00:09:11,100 --> 00:09:12,270 to high productive firms. 152 00:09:12,270 --> 00:09:14,430 And the answer is this monitoring. 153 00:09:14,430 --> 00:09:17,190 At any moment in time, there are firms 154 00:09:17,190 --> 00:09:18,690 with high expected returns. 155 00:09:18,690 --> 00:09:20,520 They tend to be underfunded. 156 00:09:20,520 --> 00:09:22,920 But it's underfunded relative to this world 157 00:09:22,920 --> 00:09:27,290 with complete information. 158 00:09:27,290 --> 00:09:29,930 Whereas, we're in the constrained second best world. 159 00:09:32,480 --> 00:09:34,850 Others have low returns and they're overfunded. 160 00:09:34,850 --> 00:09:40,550 So these are the sort of not so talented wealthier firms. 161 00:09:40,550 --> 00:09:43,040 That theme keeps coming up in all the lectures. 162 00:09:43,040 --> 00:09:45,110 In China, it was the state-owned sector. 163 00:09:45,110 --> 00:09:48,920 And last time it was firms drawing talent, 164 00:09:48,920 --> 00:09:50,870 and some people having low talent which 165 00:09:50,870 --> 00:09:57,560 is persisting and so on, which is sort of what's true here. 166 00:09:57,560 --> 00:09:59,760 And of course, as efficiency increases, 167 00:09:59,760 --> 00:10:03,090 funds are reallocated from less productive 168 00:10:03,090 --> 00:10:05,640 to more productive firms, and the interest rate spread 169 00:10:05,640 --> 00:10:07,070 goes down. 170 00:10:07,070 --> 00:10:09,510 There's going to be perfect competition in the banking 171 00:10:09,510 --> 00:10:10,870 sector. 172 00:10:10,870 --> 00:10:12,540 So the cost of borrowing is going 173 00:10:12,540 --> 00:10:21,780 to fall because these firms are going to be kind of less 174 00:10:21,780 --> 00:10:25,560 able to walk away with the money in terms of the monitoring 175 00:10:25,560 --> 00:10:26,950 technology. 176 00:10:26,950 --> 00:10:32,670 So here is a picture of the interest rate spread. 177 00:10:32,670 --> 00:10:38,190 And it kind of goes the way you would think. 178 00:10:38,190 --> 00:10:40,590 The low interest rate spread countries 179 00:10:40,590 --> 00:10:45,300 are these highly industrialized countries, 180 00:10:45,300 --> 00:10:50,130 and they tend to have low capital to GDP ratios, 181 00:10:50,130 --> 00:10:55,230 in part, because they have high GDP. 182 00:10:55,230 --> 00:10:59,160 And here's a cross-section of countries 183 00:10:59,160 --> 00:11:03,510 that vary with respect to the level of income and the capital 184 00:11:03,510 --> 00:11:04,155 output ratio. 185 00:11:08,290 --> 00:11:12,410 And this is, again, very similar, 186 00:11:12,410 --> 00:11:16,280 except this is TFP over here. 187 00:11:16,280 --> 00:11:22,195 So you can see that interest rate spread is a bad thing, 188 00:11:22,195 --> 00:11:24,790 and the higher the spread, the worse the system. 189 00:11:24,790 --> 00:11:29,290 So as the interest rate spread goes down, TFP is going up. 190 00:11:29,290 --> 00:11:33,130 So TFP is a function of this sort of financial efficiency. 191 00:11:33,130 --> 00:11:33,930 Yeah, Matt. 192 00:11:33,930 --> 00:11:36,845 AUDIENCE: What data do they use for the interest rate spreads? 193 00:11:36,845 --> 00:11:37,470 PROFESSOR: Oh-- 194 00:11:37,470 --> 00:11:39,760 AUDIENCE: [INAUDIBLE] the interest rate [INAUDIBLE].. 195 00:11:39,760 --> 00:11:42,880 PROFESSOR: I don't know if they grabbed some IMF statistics. 196 00:11:42,880 --> 00:11:46,120 I think IMF financial statistics [INAUDIBLE] have that. 197 00:11:46,120 --> 00:11:49,390 I'm not sure if that's what they used. 198 00:11:49,390 --> 00:11:52,450 But basically, yeah, it's some measure 199 00:11:52,450 --> 00:11:54,980 of the difference between the lending rate and the borrowing 200 00:11:54,980 --> 00:11:55,480 rate. 201 00:11:55,480 --> 00:11:57,790 And then that sounds good at first, 202 00:11:57,790 --> 00:12:00,850 until you start thinking about all the many ways you could 203 00:12:00,850 --> 00:12:04,030 get at the borrowing rate. 204 00:12:04,030 --> 00:12:08,650 Often, these things ignore the informal sector entirely, 205 00:12:08,650 --> 00:12:10,430 for example. 206 00:12:10,430 --> 00:12:16,850 This is-- which this paper is doing as well in the theory. 207 00:12:16,850 --> 00:12:19,370 And this is the usual growth accounting stuff. 208 00:12:19,370 --> 00:12:19,870 OK? 209 00:12:19,870 --> 00:12:23,290 I don't know that we should have had this on the earlier slide. 210 00:12:23,290 --> 00:12:28,990 You want to predict a country's per capita income. 211 00:12:28,990 --> 00:12:32,650 It's efficient countries with higher TFP. 212 00:12:32,650 --> 00:12:36,040 So this is sort of behind the Lucas thing, which 213 00:12:36,040 --> 00:12:38,950 is why isn't money flowing very quickly 214 00:12:38,950 --> 00:12:41,470 from low-income countries to high-income countries, 215 00:12:41,470 --> 00:12:45,130 since high TFP tends to mean high marginal product 216 00:12:45,130 --> 00:12:47,470 of capital and so on. 217 00:12:47,470 --> 00:12:48,270 But it's amazing. 218 00:12:48,270 --> 00:12:49,915 They almost all lie on the line. 219 00:12:52,515 --> 00:12:54,873 AUDIENCE: Is there a TFP solo residual? 220 00:12:54,873 --> 00:12:55,540 PROFESSOR: Yeah. 221 00:12:55,540 --> 00:12:57,850 It's just a solo residual. 222 00:12:57,850 --> 00:13:01,480 And that's a good comment, because this paper 223 00:13:01,480 --> 00:13:07,420 is about the different ways to decompose at the firm level 224 00:13:07,420 --> 00:13:10,300 the sort of coefficient that's hanging 225 00:13:10,300 --> 00:13:13,000 outside the front of the production function. 226 00:13:13,000 --> 00:13:21,580 And then add it up the way [INAUDIBLE] were doing. 227 00:13:21,580 --> 00:13:27,700 So they also do this sort of Rajan/Zingales thing, which 228 00:13:27,700 --> 00:13:31,720 is to try to calibrate against the US economy, 229 00:13:31,720 --> 00:13:33,760 they look at the firm size distributions 230 00:13:33,760 --> 00:13:38,530 and the interest rate spreads in the US 231 00:13:38,530 --> 00:13:43,000 to get some of their parameters, and then take it 232 00:13:43,000 --> 00:13:45,070 to these cross-country data. 233 00:13:51,388 --> 00:13:53,151 AUDIENCE: Sir. 234 00:13:53,151 --> 00:13:53,818 PROFESSOR: Yeah. 235 00:13:53,818 --> 00:13:55,762 AUDIENCE: Actually, are [INAUDIBLE] 236 00:13:55,762 --> 00:13:56,734 distributions similar? 237 00:13:56,734 --> 00:13:59,180 Or is it developed [INAUDIBLE] country? 238 00:13:59,180 --> 00:14:00,520 PROFESSOR: No. 239 00:14:00,520 --> 00:14:03,670 So what they're saying is the US would be a more or less 240 00:14:03,670 --> 00:14:06,350 efficient financial system. 241 00:14:06,350 --> 00:14:11,440 So you could sort of think about grabbing 242 00:14:11,440 --> 00:14:15,160 the relative technology parameters from that, 243 00:14:15,160 --> 00:14:19,540 and then take those sort of what are 244 00:14:19,540 --> 00:14:21,250 counterfactual-- those parameters are 245 00:14:21,250 --> 00:14:23,800 put in the context of a developed country 246 00:14:23,800 --> 00:14:27,760 with lots of other restrictions. 247 00:14:27,760 --> 00:14:30,370 So the idea is to get the technology right, 248 00:14:30,370 --> 00:14:33,700 and then somehow figure out what difference the financial system 249 00:14:33,700 --> 00:14:35,090 makes. 250 00:14:35,090 --> 00:14:35,590 Yeah? 251 00:14:35,590 --> 00:14:37,270 AUDIENCE: What are the main assumptions 252 00:14:37,270 --> 00:14:39,166 you have to make in your own model 253 00:14:39,166 --> 00:14:41,626 to get that [INAUDIBLE] distribution [INAUDIBLE].. 254 00:14:41,626 --> 00:14:43,557 Because I mean, [INAUDIBLE] then it 255 00:14:43,557 --> 00:14:45,390 seems to be about the distribution of talent 256 00:14:45,390 --> 00:14:46,880 or distribution of [INAUDIBLE]. 257 00:14:46,880 --> 00:14:47,547 PROFESSOR: Yeah. 258 00:14:47,547 --> 00:14:52,540 Each one of these models tries to be really careful about how 259 00:14:52,540 --> 00:14:56,440 to map the data into the notation and assumptions 260 00:14:56,440 --> 00:14:57,070 they're making. 261 00:15:01,210 --> 00:15:05,470 And that's-- here, I can give you a hint of that, 262 00:15:05,470 --> 00:15:09,820 although I can't get to it in as much detail as I would like. 263 00:15:15,230 --> 00:15:17,830 So here's this production function. 264 00:15:17,830 --> 00:15:23,320 There is some aggregate shock that hits all the firms. 265 00:15:23,320 --> 00:15:27,700 There is some theta, which is an idiosyncratic-specific 266 00:15:27,700 --> 00:15:28,870 firm-level shock. 267 00:15:28,870 --> 00:15:31,060 And then we have the usual-- 268 00:15:31,060 --> 00:15:34,060 Cobb-Douglas actually just noting constant returns 269 00:15:34,060 --> 00:15:34,900 to scale. 270 00:15:34,900 --> 00:15:38,690 Different papers differ on that. 271 00:15:38,690 --> 00:15:45,340 And where do these idiosyncratic shocks come from? 272 00:15:45,340 --> 00:15:49,840 They're drawn at random from-- 273 00:15:49,840 --> 00:15:52,300 in this case, some simple two-point distribution. 274 00:15:52,300 --> 00:15:55,720 It could be high or low. 275 00:15:55,720 --> 00:15:59,050 But here is over and above that, another source 276 00:15:59,050 --> 00:16:02,320 of the heterogeneity that different firms 277 00:16:02,320 --> 00:16:05,090 have different sets of thetas. 278 00:16:05,090 --> 00:16:07,300 So the whole sort of range domain 279 00:16:07,300 --> 00:16:11,620 can be different, depending on, quote, "the tau type" 280 00:16:11,620 --> 00:16:14,140 that these firms are. 281 00:16:14,140 --> 00:16:17,200 Intermediaries are competitive. 282 00:16:17,200 --> 00:16:21,010 So they're going to sort of not get any surplus out of this. 283 00:16:21,010 --> 00:16:23,920 They raise funds from consumers and lend them to firms. 284 00:16:23,920 --> 00:16:26,740 That's typical intermediation structure. 285 00:16:26,740 --> 00:16:29,740 And the intermediary knows the firm's type tau. 286 00:16:29,740 --> 00:16:36,250 However, it doesn't know output. 287 00:16:36,250 --> 00:16:39,610 It doesn't know the idiosyncratic shock theta. 288 00:16:39,610 --> 00:16:43,550 And it doesn't know the labor supply. 289 00:16:43,550 --> 00:16:47,290 So all he's doing here is being careful to throw 290 00:16:47,290 --> 00:16:51,280 in enough sources of unknown that you can't basically 291 00:16:51,280 --> 00:16:54,100 back out from the output the unknown things 292 00:16:54,100 --> 00:16:56,830 on the right-hand side. 293 00:16:56,830 --> 00:17:00,460 And he probably could have gone with different assumptions, 294 00:17:00,460 --> 00:17:03,220 but that's certainly the spirit of it. 295 00:17:03,220 --> 00:17:07,074 Or at least, doesn't observe these things costlessly. 296 00:17:10,630 --> 00:17:14,260 But there is this monitoring technology. 297 00:17:20,030 --> 00:17:26,030 The firms experience their true productivity, say, high or low, 298 00:17:26,030 --> 00:17:30,710 call it for the theta i for a given firm. 299 00:17:30,710 --> 00:17:35,050 And suppose the firm lies about it and says theta j. 300 00:17:35,050 --> 00:17:36,560 Theta's not observed nor inferred 301 00:17:36,560 --> 00:17:40,340 from any of the other ex ante or ex post. 302 00:17:40,340 --> 00:17:45,680 Then the intermediary devotes labor to monitoring the claim. 303 00:17:48,620 --> 00:17:51,770 So this technology of verification 304 00:17:51,770 --> 00:17:53,210 is actually resource using. 305 00:17:53,210 --> 00:17:54,720 It's another production function. 306 00:17:58,540 --> 00:18:03,070 And then the size of the loan is determined. 307 00:18:03,070 --> 00:18:04,810 That's the capitalization of the firm. 308 00:18:04,810 --> 00:18:06,970 They don't have any wealth on their own. 309 00:18:06,970 --> 00:18:10,180 That's kind of the scale of the project. 310 00:18:10,180 --> 00:18:15,550 The level of productivity and monitoring is z. 311 00:18:15,550 --> 00:18:18,880 So that's kind of like whether it's a completely noisy signal, 312 00:18:18,880 --> 00:18:24,250 or an extremely accurate signal of the actual true state theta 313 00:18:24,250 --> 00:18:25,870 i. 314 00:18:25,870 --> 00:18:29,710 So this object, Pij is-- 315 00:18:29,710 --> 00:18:33,130 that's the probability that if the firm 316 00:18:33,130 --> 00:18:35,950 made a counterfactual report of theta j 317 00:18:35,950 --> 00:18:39,700 and really were a theta i, with this amount of labor 318 00:18:39,700 --> 00:18:43,480 devoted to monitoring, given this loan size, 319 00:18:43,480 --> 00:18:46,480 and given this index of the productivity of the technology, 320 00:18:46,480 --> 00:18:49,000 this is the probability of finding out 321 00:18:49,000 --> 00:18:54,640 what the truth of it really is. 322 00:18:54,640 --> 00:19:00,230 And so-- and then this mapping here 323 00:19:00,230 --> 00:19:03,770 is to provide between productivity 324 00:19:03,770 --> 00:19:10,300 and financial sector z, which, say, we don't see. 325 00:19:10,300 --> 00:19:16,160 And the things that we do see, at least 326 00:19:16,160 --> 00:19:19,470 we, as econometricians, see output and interest rate 327 00:19:19,470 --> 00:19:19,970 spreads. 328 00:19:19,970 --> 00:19:24,090 Now, let me pause for a second because you might well 329 00:19:24,090 --> 00:19:26,160 think of the obvious thing is, wait a minute. 330 00:19:26,160 --> 00:19:28,410 Didn't you just say output was unobserved? 331 00:19:28,410 --> 00:19:31,290 Well, this model and others make a distinction 332 00:19:31,290 --> 00:19:34,080 between the exposed data that we have 333 00:19:34,080 --> 00:19:37,320 as analysts about firm output, as 334 00:19:37,320 --> 00:19:41,860 opposed to what the intermediary is seeing along the way. 335 00:19:41,860 --> 00:19:44,247 So we're imagining we do see output 336 00:19:44,247 --> 00:19:45,580 and these interest rate spreads. 337 00:19:45,580 --> 00:19:48,490 Those were the slides I just showed you. 338 00:19:48,490 --> 00:19:51,370 And then this is where I'm going to-- 339 00:19:51,370 --> 00:19:53,260 Matt-- going to be a bit vague. 340 00:19:53,260 --> 00:19:57,700 If you looked at the equations in the notation, 341 00:19:57,700 --> 00:20:04,950 they can basically invert those equations to get x and z. 342 00:20:04,950 --> 00:20:07,510 And you can do it at a point in time for a given country, 343 00:20:07,510 --> 00:20:10,240 or you can do it at two different points in time. 344 00:20:10,240 --> 00:20:12,220 If you're getting z, you're getting 345 00:20:12,220 --> 00:20:16,000 a measure of this productivity of the monitoring technology. 346 00:20:16,000 --> 00:20:17,650 And that's why, for example, they 347 00:20:17,650 --> 00:20:22,410 are able to make statements about Taiwan's productivity 348 00:20:22,410 --> 00:20:26,260 increased over whatever it was, 10, 15 years, 349 00:20:26,260 --> 00:20:30,010 along with the drop in the intermediation spreads. 350 00:20:30,010 --> 00:20:34,750 They infer that from the drop in the intermediation spreads 351 00:20:34,750 --> 00:20:38,108 and the capital output ratio. 352 00:20:38,108 --> 00:20:38,900 AUDIENCE: Question. 353 00:20:38,900 --> 00:20:40,058 PROFESSOR: Yep. 354 00:20:40,058 --> 00:20:44,998 AUDIENCE: So the assumption that the probability of getting 355 00:20:44,998 --> 00:20:48,460 [INAUDIBLE] is decreasing. 356 00:20:48,460 --> 00:20:49,250 PROFESSOR: Yeah. 357 00:20:49,250 --> 00:20:52,940 So this is a scale effect. 358 00:20:52,940 --> 00:20:54,980 This came up in Alp's lecture the other day 359 00:20:54,980 --> 00:20:59,510 as well, which is if you have a fixed cost of monitoring 360 00:20:59,510 --> 00:21:02,780 or something, and then the country gets richer and richer, 361 00:21:02,780 --> 00:21:04,850 and credit is going up and up, then 362 00:21:04,850 --> 00:21:07,280 basically, the cost of verification 363 00:21:07,280 --> 00:21:08,970 is going down and down. 364 00:21:08,970 --> 00:21:13,580 But here, a country's getting richer and richer, 365 00:21:13,580 --> 00:21:16,970 and you don't want this sort of information imperfection 366 00:21:16,970 --> 00:21:20,030 to go away other than through investment 367 00:21:20,030 --> 00:21:21,320 in improved technology. 368 00:21:21,320 --> 00:21:25,460 So it allows that-- 369 00:21:25,460 --> 00:21:28,880 I guess that conceptually, it's like bunch of little firms 370 00:21:28,880 --> 00:21:29,690 all spread out. 371 00:21:29,690 --> 00:21:32,090 And the bigger the loan size, the more plants 372 00:21:32,090 --> 00:21:36,052 you got to go to see- something like that. 373 00:21:36,052 --> 00:21:39,850 AUDIENCE: And is the result partially 374 00:21:39,850 --> 00:21:41,380 depending on this assumption? 375 00:21:41,380 --> 00:21:44,133 Because it might be the direction is-- 376 00:21:44,133 --> 00:21:44,800 PROFESSOR: Yeah. 377 00:21:44,800 --> 00:21:45,300 Yeah. 378 00:21:45,300 --> 00:21:47,110 It matters quite a bit. 379 00:21:47,110 --> 00:21:48,415 Now, this is kind of linear. 380 00:21:51,280 --> 00:21:53,970 If it's diminishing, the effect eventually goes away. 381 00:21:53,970 --> 00:21:57,345 If it's convex, then I don't know what happens. 382 00:21:57,345 --> 00:21:57,970 But it matters. 383 00:22:03,510 --> 00:22:04,010 All right. 384 00:22:04,010 --> 00:22:06,320 So that's the spirit of that paper. 385 00:22:06,320 --> 00:22:10,330 Again, hopefully, it whets your appetite and-- 386 00:22:10,330 --> 00:22:13,100 or at least you can get some sense of what people 387 00:22:13,100 --> 00:22:16,915 are doing in the literature. 388 00:22:16,915 --> 00:22:18,290 And there's one other that I want 389 00:22:18,290 --> 00:22:22,340 to talk about before we get to the main paper of this lecture. 390 00:22:22,340 --> 00:22:25,910 And this I discovered recently, actually-- 391 00:22:25,910 --> 00:22:29,270 "International Capital Flows and Credit Markets-- 392 00:22:29,270 --> 00:22:31,520 A Tale of Two Frictions." 393 00:22:31,520 --> 00:22:37,000 Somehow everyone's got taken with Dickens. 394 00:22:37,000 --> 00:22:40,397 The one we looked at last time was "A Tale of Two Sectors." 395 00:22:40,397 --> 00:22:42,730 I don't know whether you thought that was clever or not. 396 00:22:42,730 --> 00:22:44,950 This paper was written after that. 397 00:22:49,010 --> 00:22:53,590 So it's kind of intriguing, and I'm not 398 00:22:53,590 --> 00:22:55,270 quite sure I totally believe it. 399 00:22:55,270 --> 00:23:00,710 But let me try to walk you through it. 400 00:23:00,710 --> 00:23:04,570 So the idea is we see a lot of capital 401 00:23:04,570 --> 00:23:07,675 moving around the world, and we see boom/bust cycles. 402 00:23:10,053 --> 00:23:12,220 And they're going to try to get both of those things 403 00:23:12,220 --> 00:23:14,140 in into one model. 404 00:23:14,140 --> 00:23:16,780 So it's ambitious. 405 00:23:16,780 --> 00:23:19,390 Global imbalances, large and persistent capital 406 00:23:19,390 --> 00:23:21,940 flows coming out of Asia to the US-- 407 00:23:21,940 --> 00:23:25,840 we talked about that last time in the-- 408 00:23:25,840 --> 00:23:27,370 last time in the China paper. 409 00:23:30,900 --> 00:23:35,130 And there the idea is it's hard to make good use of it 410 00:23:35,130 --> 00:23:39,690 within China for the reasons that that model and others 411 00:23:39,690 --> 00:23:41,860 postulate, namely, financial friction. 412 00:23:41,860 --> 00:23:43,920 So then it should go somewhere else. 413 00:23:51,750 --> 00:23:58,190 But you could also think about it on the other side of it. 414 00:23:58,190 --> 00:23:59,780 This is the tricky part. 415 00:23:59,780 --> 00:24:13,000 Namely, that that money is going to unproductive investment 416 00:24:13,000 --> 00:24:14,320 in the US. 417 00:24:14,320 --> 00:24:18,680 So at least that's ex post the judgment. 418 00:24:18,680 --> 00:24:20,590 So then you have to model sort of what's 419 00:24:20,590 --> 00:24:24,490 going on with the US interest rate that would allow 420 00:24:24,490 --> 00:24:28,390 the existence of bad projects and the coexistence 421 00:24:28,390 --> 00:24:29,965 of bad projects with good projects. 422 00:24:39,160 --> 00:24:40,380 So they try to do both. 423 00:24:46,600 --> 00:24:51,930 So again, there's credit market which 424 00:24:51,930 --> 00:24:54,480 intermediates the money that comes from savers 425 00:24:54,480 --> 00:24:56,760 and goes to investors. 426 00:24:56,760 --> 00:25:00,560 Individuals are endowed with some resources 427 00:25:00,560 --> 00:25:03,780 and with an investment project. 428 00:25:03,780 --> 00:25:06,240 And they have to decide whether to do the project 429 00:25:06,240 --> 00:25:09,255 and become an entrepreneur, in which case 430 00:25:09,255 --> 00:25:11,130 they're going to want some credit to fund it, 431 00:25:11,130 --> 00:25:14,940 or to forego their project and become savers, in which case 432 00:25:14,940 --> 00:25:17,340 they supply resources to the credit market. 433 00:25:17,340 --> 00:25:20,610 Now, this building block is the same on almost all the papers 434 00:25:20,610 --> 00:25:22,880 that we've looked at. 435 00:25:22,880 --> 00:25:26,120 It's an occupation choice model of-- 436 00:25:29,410 --> 00:25:34,120 that's used to address financial issues. 437 00:25:34,120 --> 00:25:36,760 To give adverse selection a crucial role, 438 00:25:36,760 --> 00:25:38,170 we assume individual productivity 439 00:25:38,170 --> 00:25:42,610 is private information and unobservable by the lenders. 440 00:25:42,610 --> 00:25:47,810 That kind of sounds like the previous Greenwood paper. 441 00:25:47,810 --> 00:25:49,840 So there is some cross-subsidization 442 00:25:49,840 --> 00:25:55,670 between high and low productivity entrepreneurs. 443 00:25:55,670 --> 00:25:59,290 A key assumption-- there's only one interest rate. 444 00:25:59,290 --> 00:26:01,360 So there's no menu of contracts. 445 00:26:01,360 --> 00:26:04,825 There's no other ways to separate good from bad. 446 00:26:07,750 --> 00:26:11,610 And the thing, to say it intuitively, 447 00:26:11,610 --> 00:26:15,840 about adverse selection is, who wants to borrow 448 00:26:15,840 --> 00:26:18,120 at a given interest rate? 449 00:26:18,120 --> 00:26:20,430 Well, that's really attractive for people who 450 00:26:20,430 --> 00:26:24,040 expect never to repay loans. 451 00:26:24,040 --> 00:26:26,410 So the lender is going to have to somehow break even 452 00:26:26,410 --> 00:26:29,590 on those bad types that are indistinguishable 453 00:26:29,590 --> 00:26:30,490 from the good types. 454 00:26:30,490 --> 00:26:34,600 And that is a force determining the interest rate. 455 00:26:38,540 --> 00:26:44,800 Now, the question is whether that kind of perverse dynamics 456 00:26:44,800 --> 00:26:49,210 would move around with wealth and so on. 457 00:26:49,210 --> 00:26:50,006 Yes. 458 00:26:50,006 --> 00:26:52,440 AUDIENCE: Are the lenders endowed with mechanisms 459 00:26:52,440 --> 00:26:53,691 to prevent that? 460 00:26:53,691 --> 00:26:55,780 Is there physical collateral, or is there 461 00:26:55,780 --> 00:26:56,822 some sort of reputation-- 462 00:26:56,822 --> 00:26:57,920 PROFESSOR: Not here. 463 00:26:57,920 --> 00:26:58,420 Not here. 464 00:26:58,420 --> 00:27:01,738 But there is a big literature that argues-- 465 00:27:01,738 --> 00:27:02,280 AUDIENCE: OK. 466 00:27:02,280 --> 00:27:03,812 So in this-- this environment has 467 00:27:03,812 --> 00:27:06,270 the potential for the sort of adverse selection [INAUDIBLE] 468 00:27:06,270 --> 00:27:08,213 type things. 469 00:27:08,213 --> 00:27:08,880 PROFESSOR: Yeah. 470 00:27:08,880 --> 00:27:12,790 There is always an issue in papers and as researchers 471 00:27:12,790 --> 00:27:16,540 at what-- how much-- how many substantive things you 472 00:27:16,540 --> 00:27:18,760 want going on at the same time. 473 00:27:18,760 --> 00:27:22,240 I would say this is among the simpler adverse selection 474 00:27:22,240 --> 00:27:22,740 thing. 475 00:27:22,740 --> 00:27:26,770 Personally, I would always think menus of contracts 476 00:27:26,770 --> 00:27:28,720 could select-- 477 00:27:28,720 --> 00:27:31,930 allow selection or sort of truth-telling. 478 00:27:31,930 --> 00:27:35,080 And that's just shut down. 479 00:27:35,080 --> 00:27:37,420 But Rothschild/Stiglitz shut that down. 480 00:27:40,120 --> 00:27:43,960 So maybe that's off to a tradition of sorts. 481 00:27:48,070 --> 00:27:51,640 So what are the implications of this adverse selection? 482 00:27:54,110 --> 00:27:55,860 The interest rate has to be higher than it 483 00:27:55,860 --> 00:27:58,890 would have been otherwise. 484 00:27:58,890 --> 00:28:00,390 But on the other hand, there's a lot 485 00:28:00,390 --> 00:28:01,680 of borrowing and investment. 486 00:28:01,680 --> 00:28:04,980 Now, be careful here, because these low productivity 487 00:28:04,980 --> 00:28:07,750 guys are happy to be borrowers. 488 00:28:07,750 --> 00:28:09,420 So not all borrowing is a good thing. 489 00:28:09,420 --> 00:28:12,750 And that's kind of the seeds of the bust getting created. 490 00:28:16,640 --> 00:28:18,350 Because of adverse selection we have 491 00:28:18,350 --> 00:28:21,410 this wedge between the equilibrium interest rate 492 00:28:21,410 --> 00:28:24,740 and the marginal return to investment. 493 00:28:24,740 --> 00:28:26,630 Again, it's a common wedge. 494 00:28:26,630 --> 00:28:28,880 We just did an interest rate spread. 495 00:28:28,880 --> 00:28:31,610 Here's another way to think about that spread. 496 00:28:31,610 --> 00:28:33,740 That doesn't mean it doesn't matter. 497 00:28:33,740 --> 00:28:37,770 The underlying assumptions do matter. 498 00:28:37,770 --> 00:28:42,540 But this phenomenon is pervasive across countries in the data, 499 00:28:42,540 --> 00:28:45,280 and is a big part of almost all these models. 500 00:28:45,280 --> 00:28:50,510 So adverse selection means that the economy 501 00:28:50,510 --> 00:28:53,770 can attract more capital, and boost in 502 00:28:53,770 --> 00:28:58,230 that capital flows seems odd just to say it there. 503 00:28:58,230 --> 00:28:59,310 You can see it written. 504 00:29:02,370 --> 00:29:07,410 But again, the idea is that somehow the interest rate 505 00:29:07,410 --> 00:29:09,840 is higher than it would have been in the full information 506 00:29:09,840 --> 00:29:15,822 economy, and outside investors seek the higher interest. 507 00:29:15,822 --> 00:29:17,530 So that's kind of what they mean by that. 508 00:29:20,300 --> 00:29:27,310 But the true marginal return lies below the interest rate. 509 00:29:27,310 --> 00:29:29,710 So something bad is happening. 510 00:29:29,710 --> 00:29:34,570 And the way it comes out in the bath here is the capital-- 511 00:29:34,570 --> 00:29:39,900 there's lower aggregate consumption because you've 512 00:29:39,900 --> 00:29:42,857 used resources inefficiently. 513 00:29:45,840 --> 00:29:48,720 And then they claim this adverse selection actually 514 00:29:48,720 --> 00:29:50,400 is a force for volatility. 515 00:29:54,810 --> 00:29:56,295 This is kind of key. 516 00:29:56,295 --> 00:29:58,950 The incentives of less productive individuals 517 00:29:58,950 --> 00:30:02,490 to become entrepreneurs are strongest-- 518 00:30:02,490 --> 00:30:06,000 there's bad types borrow and don't expect to repay much. 519 00:30:06,000 --> 00:30:09,180 That force is strongest, ironically, when the capital 520 00:30:09,180 --> 00:30:12,360 stock is low and income is low. 521 00:30:18,400 --> 00:30:19,930 And what's that all about? 522 00:30:19,930 --> 00:30:28,730 Well, basically this debt overhang problem 523 00:30:28,730 --> 00:30:31,130 is more severe when the capital stock is low, 524 00:30:31,130 --> 00:30:35,560 because you have to borrow more than you would if the capital 525 00:30:35,560 --> 00:30:37,990 stock were high, and you could almost self-finance. 526 00:30:41,260 --> 00:30:45,040 And there's a lot of untalented firms borrowing 527 00:30:45,040 --> 00:30:49,180 when the total income and capital stock of the economy 528 00:30:49,180 --> 00:30:49,800 is low. 529 00:30:53,920 --> 00:30:56,480 Or conversely, as capital and income increases, 530 00:30:56,480 --> 00:31:00,120 that cross-subsidization is decreasing. 531 00:31:00,120 --> 00:31:05,560 There's more self-financing, and entrepreneurship 532 00:31:05,560 --> 00:31:10,400 loses its appeal for less productive individuals. 533 00:31:10,400 --> 00:31:12,400 Well, I mean, it's actually kind of interesting. 534 00:31:12,400 --> 00:31:15,790 Because you can see the dynamics of occupation choice 535 00:31:15,790 --> 00:31:18,260 that we've had in the other papers. 536 00:31:18,260 --> 00:31:20,830 Who's going to be a firm as opposed to a wage earner 537 00:31:20,830 --> 00:31:24,760 depends on what happens if you were a firm 538 00:31:24,760 --> 00:31:25,735 and decided to borrow. 539 00:31:29,670 --> 00:31:31,770 And then they actually get a boom/bust cycle. 540 00:31:34,280 --> 00:31:41,520 They claim to show that capital inflows are going to fuel 541 00:31:41,520 --> 00:31:43,080 this accumulation period. 542 00:31:43,080 --> 00:31:47,280 And then you'll get a contraction. 543 00:31:51,900 --> 00:31:55,410 Actually, toward the end, the tale of two obstacles or two 544 00:31:55,410 --> 00:32:05,450 frictions, is they throw pledgeability on top of it. 545 00:32:05,450 --> 00:32:09,580 So the more traditional sort of collateral constraint 546 00:32:09,580 --> 00:32:12,370 reduces investment and lowers the interest rate. 547 00:32:12,370 --> 00:32:13,390 We've seen that, right? 548 00:32:13,390 --> 00:32:18,720 You can't borrow as much, so there's less pressure 549 00:32:18,720 --> 00:32:19,660 on the interest rate. 550 00:32:19,660 --> 00:32:23,310 So the interest rate would be lower. 551 00:32:23,310 --> 00:32:27,030 Ironically, that low interest rate, though, 552 00:32:27,030 --> 00:32:30,300 decreases the returns to savings. 553 00:32:30,300 --> 00:32:35,550 And why be a wage earner and get a lower return on your savings 554 00:32:35,550 --> 00:32:38,400 when you can be a not so good entrepreneur 555 00:32:38,400 --> 00:32:40,040 and borrow money at low interest? 556 00:32:40,040 --> 00:32:45,310 So you get this sort of, they claim, 557 00:32:45,310 --> 00:32:48,720 a kind of accelerator or exacerbating feature, 558 00:32:48,720 --> 00:32:51,990 that the adverse selection is interacting 559 00:32:51,990 --> 00:32:54,331 with the limited collateral constraint. 560 00:32:58,660 --> 00:33:05,060 And that should whet your appetite for that paper. 561 00:33:05,060 --> 00:33:10,220 Now again, I apologize in a way, although it wasn't my intention 562 00:33:10,220 --> 00:33:16,220 to go through all the details of that paper. 563 00:33:16,220 --> 00:33:20,180 And it is on the Stellar website. 564 00:33:20,180 --> 00:33:24,350 And we can study it in more detail. 565 00:33:24,350 --> 00:33:25,970 But I offered it here just to show 566 00:33:25,970 --> 00:33:28,370 you the kinds of phenomenon that people 567 00:33:28,370 --> 00:33:32,030 are trying to address with the interplay 568 00:33:32,030 --> 00:33:33,880 of financial frictions. 569 00:33:36,810 --> 00:33:46,690 And then we come to the paper I wish to focus on, 570 00:33:46,690 --> 00:33:48,930 which is, again, a tale of two frictions. 571 00:33:48,930 --> 00:33:50,370 But we can't use that title. 572 00:33:54,850 --> 00:33:59,090 But we're doing it in a different way. 573 00:33:59,090 --> 00:34:04,330 We allow moral hazard and also limited commitment. 574 00:34:04,330 --> 00:34:06,460 So on the one hand, we're going to take 575 00:34:06,460 --> 00:34:10,360 the traditional limited commitment constraint 576 00:34:10,360 --> 00:34:13,719 and remove it, and insert a moral hazard 577 00:34:13,719 --> 00:34:15,290 constraint explicitly. 578 00:34:15,290 --> 00:34:17,469 And this time, I can show you the equations. 579 00:34:21,300 --> 00:34:23,850 But we also allow-- 580 00:34:23,850 --> 00:34:28,830 we don't get rid of the capital constraint entirely. 581 00:34:31,980 --> 00:34:35,159 And then you say, well, what are the rules here 582 00:34:35,159 --> 00:34:36,480 for making this stuff up? 583 00:34:39,110 --> 00:34:41,270 And my answer is the data. 584 00:34:43,989 --> 00:34:48,780 And in particular-- this is the reverse order coming here-- 585 00:34:48,780 --> 00:34:51,179 you will see a paper that I've written 586 00:34:51,179 --> 00:34:55,770 with Alex Karaivanov, where we have consumption, output, 587 00:34:55,770 --> 00:34:59,780 investment, and capital stock data, 588 00:34:59,780 --> 00:35:08,340 and we actually estimate a variety of financial regimes. 589 00:35:08,340 --> 00:35:11,100 And it turns out that the so-called limited commitment 590 00:35:11,100 --> 00:35:16,890 regime fits the data best in the rural data that I have 591 00:35:16,890 --> 00:35:21,180 and in the Northeast, and the moral hazard regime 592 00:35:21,180 --> 00:35:25,633 fits better in the urban areas and in the areas 593 00:35:25,633 --> 00:35:26,550 in and around Bangkok. 594 00:35:29,470 --> 00:35:32,230 So that's the way to think about these two frictions. 595 00:35:32,230 --> 00:35:35,050 They're, for us, going to be microfounded. 596 00:35:35,050 --> 00:35:37,450 Although here, we're just borrowing that and putting it 597 00:35:37,450 --> 00:35:37,950 in. 598 00:35:41,035 --> 00:35:41,535 So-- 599 00:35:46,248 --> 00:35:47,240 AUDIENCE: Sir. 600 00:35:47,240 --> 00:35:48,065 PROFESSOR: Yep. 601 00:35:48,065 --> 00:35:49,440 AUDIENCE: So basically, are you-- 602 00:35:49,440 --> 00:35:52,880 you're going to sort of let these two things interplay, 603 00:35:52,880 --> 00:36:00,770 and you're going to have them be differently 604 00:36:00,770 --> 00:36:03,180 important in different places. 605 00:36:03,180 --> 00:36:03,990 PROFESSOR: Yeah. 606 00:36:03,990 --> 00:36:06,790 So that's where we end up. 607 00:36:06,790 --> 00:36:09,570 And where we start is we have an economy, 608 00:36:09,570 --> 00:36:12,060 and I'll tell you all about the environment. 609 00:36:12,060 --> 00:36:14,910 And then I'll describe what goes on with the limited commitment 610 00:36:14,910 --> 00:36:19,200 constraint as if it were the whole economy. 611 00:36:19,200 --> 00:36:22,590 I'll do the same for moral hazard, the whole economy. 612 00:36:22,590 --> 00:36:28,130 Then I'll have like 50-50, and use the general equilibrium. 613 00:36:28,130 --> 00:36:32,810 And the punch line is going to be the general equilibrium 614 00:36:32,810 --> 00:36:34,370 with the two frictions is not just 615 00:36:34,370 --> 00:36:39,110 a simple convex combination of the two extremes. 616 00:36:39,110 --> 00:36:40,820 And the reason that that's going on 617 00:36:40,820 --> 00:36:43,930 has to do with the general equilibrium. 618 00:36:43,930 --> 00:36:44,930 So you're going to get-- 619 00:36:47,953 --> 00:36:50,370 roughly, you're going to get interest rates and wages that 620 00:36:50,370 --> 00:36:53,520 are somewhat between the two extremes. 621 00:36:53,520 --> 00:36:56,040 But then given that, the question 622 00:36:56,040 --> 00:36:59,880 about who is to be a firm or a worker depends on the obstacles 623 00:36:59,880 --> 00:37:01,860 that you face. 624 00:37:01,860 --> 00:37:03,780 And actually, what turns out to be 625 00:37:03,780 --> 00:37:07,710 cool at the parameters we have, which we think are reasonable 626 00:37:07,710 --> 00:37:16,260 and are largely similar to what Alex and I had, 627 00:37:16,260 --> 00:37:20,520 that in the places with moral hazard, 628 00:37:20,520 --> 00:37:23,310 output is substantially higher. 629 00:37:27,990 --> 00:37:30,140 There are slightly more firms there. 630 00:37:30,140 --> 00:37:33,120 They're larger. 631 00:37:33,120 --> 00:37:38,040 Those firms are renting more capital or borrowing more. 632 00:37:38,040 --> 00:37:44,240 The ratio of credit to GDP, the financing ratio, is higher. 633 00:37:44,240 --> 00:37:47,460 And they employ more labor. 634 00:37:47,460 --> 00:37:52,580 Without any geography somehow, we've created these-- 635 00:37:52,580 --> 00:37:55,250 replicated these stylized facts of development 636 00:37:55,250 --> 00:37:59,150 when you think about rural areas versus urban areas. 637 00:37:59,150 --> 00:38:03,080 Money is flowing from savings in rural areas 638 00:38:03,080 --> 00:38:07,325 to the urbanizing cities. 639 00:38:07,325 --> 00:38:11,500 Now, I'm not going to claim that we've captured everything going 640 00:38:11,500 --> 00:38:12,700 on in all these countries. 641 00:38:12,700 --> 00:38:17,230 But this simple sort of combination 642 00:38:17,230 --> 00:38:20,470 of financial frictions allows that. 643 00:38:20,470 --> 00:38:23,150 In that sense, this is very different from the other paper. 644 00:38:23,150 --> 00:38:24,005 Yes. 645 00:38:24,005 --> 00:38:27,310 AUDIENCE: And so do you have something that endogenously 646 00:38:27,310 --> 00:38:32,570 makes these two regimes being relevant in those areas? 647 00:38:32,570 --> 00:38:36,790 PROFESSOR: That's a good question, and it's-- 648 00:38:36,790 --> 00:38:40,240 at this point, it's hard to answer. 649 00:38:40,240 --> 00:38:43,960 I mean, it looks as if either there 650 00:38:43,960 --> 00:38:46,990 is a legal restriction on collateral, 651 00:38:46,990 --> 00:38:50,940 or the set of financial institutions that are operating 652 00:38:50,940 --> 00:38:54,600 in the rural areas is different from what's 653 00:38:54,600 --> 00:38:58,260 going on in the urban areas. 654 00:38:58,260 --> 00:39:01,590 And if we were going to follow Jeremy's path, 655 00:39:01,590 --> 00:39:03,590 we would ask your question. 656 00:39:03,590 --> 00:39:07,260 We are asking, and we just don't have answers yet, 657 00:39:07,260 --> 00:39:11,370 as to what is the configuration of the financial service 658 00:39:11,370 --> 00:39:12,480 providers? 659 00:39:12,480 --> 00:39:14,760 You'll see there is a heavy distinction-- we've 660 00:39:14,760 --> 00:39:17,820 already seen, actually, where the commercial banks are 661 00:39:17,820 --> 00:39:20,850 operating and where this government agricultural bank is 662 00:39:20,850 --> 00:39:21,420 operating. 663 00:39:21,420 --> 00:39:24,840 And I think that's probably part of the story. 664 00:39:24,840 --> 00:39:30,540 But in this paper, we do not take a stand on it. 665 00:39:30,540 --> 00:39:33,690 We take as given just this sort of difference 666 00:39:33,690 --> 00:39:36,970 in the financial information regime. 667 00:39:36,970 --> 00:39:38,580 OK. 668 00:39:38,580 --> 00:39:48,130 So this, we've already said. 669 00:39:48,130 --> 00:39:51,720 This, we've already talked about. 670 00:39:51,720 --> 00:39:54,360 So we get to the model. 671 00:39:54,360 --> 00:39:59,980 So a household that could decide to be either a wage 672 00:39:59,980 --> 00:40:05,370 earner or a firm will maximize discounted expected utility. 673 00:40:05,370 --> 00:40:08,100 Beta is the discount rate. u is the common utility 674 00:40:08,100 --> 00:40:10,830 function, the arguments of which are 675 00:40:10,830 --> 00:40:17,420 consumption of this household i, and effort of this household i. 676 00:40:17,420 --> 00:40:22,110 We'll parameterize this when we need to compute something. 677 00:40:22,110 --> 00:40:24,720 At this level, it's quite general. 678 00:40:24,720 --> 00:40:28,440 You can-- if x is equal-- this is like a binary choice. 679 00:40:28,440 --> 00:40:30,900 You can be a worker or a firm. 680 00:40:30,900 --> 00:40:33,460 By convention, when x is 1, you're a firm, 681 00:40:33,460 --> 00:40:36,430 and when x is zero, you're a worker. 682 00:40:36,430 --> 00:40:39,200 And that's going to be endogenous. 683 00:40:39,200 --> 00:40:45,950 If you do become a firm, you have this productivity draw z, 684 00:40:45,950 --> 00:40:48,710 and it evolves over time according to sort 685 00:40:48,710 --> 00:40:51,590 of a standard Markov process. 686 00:40:51,590 --> 00:40:54,110 I was rushing toward the end of the lecture last time, 687 00:40:54,110 --> 00:40:59,960 but this was the thing that was in the [INAUDIBLE] 688 00:40:59,960 --> 00:41:03,023 paper toward the end, and I guess 689 00:41:03,023 --> 00:41:04,190 in some of the other papers. 690 00:41:04,190 --> 00:41:06,350 And finally, we have wealth. 691 00:41:06,350 --> 00:41:09,180 So at the beginning of the period, 692 00:41:09,180 --> 00:41:13,680 everyone has some predetermined level of wealth, 693 00:41:13,680 --> 00:41:16,290 and it's in the bank essentially. 694 00:41:16,290 --> 00:41:19,710 And so an individual is characterized 695 00:41:19,710 --> 00:41:23,160 by their initial wealth and their current talent. 696 00:41:26,490 --> 00:41:29,410 Here is the production function. 697 00:41:29,410 --> 00:41:33,520 It's diminishing returns to scale in capital 698 00:41:33,520 --> 00:41:34,540 and hired labor. 699 00:41:37,520 --> 00:41:41,330 And then we have these sort of things out front here. 700 00:41:41,330 --> 00:41:44,700 The notation varies from one paper to the next. 701 00:41:44,700 --> 00:41:49,610 But epsilon is here, an idiosyncratic shock. 702 00:41:49,610 --> 00:41:55,960 iid over households, and z is that level of productivity 703 00:41:55,960 --> 00:41:59,830 we already talked about that's evolving over time. 704 00:41:59,830 --> 00:42:05,730 Now, there's actually-- the effort 705 00:42:05,730 --> 00:42:09,030 of the entrepreneur is another factor of production. 706 00:42:09,030 --> 00:42:13,440 And that's basically little e. 707 00:42:13,440 --> 00:42:18,180 So the way that the effort of the entrepreneur is modeled 708 00:42:18,180 --> 00:42:20,380 is through this. 709 00:42:20,380 --> 00:42:25,080 The epsilon, the idiosyncratic productivity, 710 00:42:25,080 --> 00:42:29,160 depends on how hard the entrepreneur is, quote, 711 00:42:29,160 --> 00:42:31,710 "working," or thinking, you know, 712 00:42:31,710 --> 00:42:35,020 staying up at night worrying-- basically, due diligence. 713 00:42:35,020 --> 00:42:36,900 This is not measured. 714 00:42:36,900 --> 00:42:40,140 You do see the epsilon, but you don't see the e. 715 00:42:40,140 --> 00:42:44,670 And that's if-- well, at least in the moral hazard model, 716 00:42:44,670 --> 00:42:45,720 you don't see the e. 717 00:42:53,340 --> 00:42:56,510 But I should qualify that we're going 718 00:42:56,510 --> 00:42:58,100 through the standard environment, 719 00:42:58,100 --> 00:43:00,890 and there's going to be a full information 720 00:43:00,890 --> 00:43:03,200 environment with limited commitment in which everything 721 00:43:03,200 --> 00:43:06,460 is seen and full insurance is possible, 722 00:43:06,460 --> 00:43:08,510 versus the moral hazard environment which 723 00:43:08,510 --> 00:43:12,500 is going to be the source of limited insurance, 724 00:43:12,500 --> 00:43:14,750 and it's going to interact-- 725 00:43:14,750 --> 00:43:19,460 both frictions interact with productivity. 726 00:43:19,460 --> 00:43:26,270 So there are banks or financial intermediaries, 727 00:43:26,270 --> 00:43:32,180 and you should think of them as basically risk syndicates. 728 00:43:32,180 --> 00:43:38,210 Because not only do they sort of lend out of the wealth 729 00:43:38,210 --> 00:43:41,310 that households have been depositing with them, 730 00:43:41,310 --> 00:43:46,250 they can also smooth over idiosyncratic shocks. 731 00:43:46,250 --> 00:43:49,730 So there's a state contingent payment 732 00:43:49,730 --> 00:43:51,850 that the banks can make-- 733 00:43:51,850 --> 00:43:54,200 it could be negative-- 734 00:43:54,200 --> 00:43:56,600 say, to the households as a function 735 00:43:56,600 --> 00:44:00,575 of their realized and observed epsilon. 736 00:44:03,980 --> 00:44:08,440 Now, in some respects, it's still-- 737 00:44:11,430 --> 00:44:15,420 we're not solving a massive programming problem 738 00:44:15,420 --> 00:44:17,760 and then capturing the interest rates 739 00:44:17,760 --> 00:44:21,690 and wages off some first order conditions. 740 00:44:21,690 --> 00:44:24,150 Instead, we're going to imagine that these risk 741 00:44:24,150 --> 00:44:28,740 syndicates, these banks, and the firms and the households 742 00:44:28,740 --> 00:44:33,840 all take as given economy-wide interest rates and wages. 743 00:44:33,840 --> 00:44:35,280 And like many of the other papers, 744 00:44:35,280 --> 00:44:39,210 we're going to have to determine those by supply and demand 745 00:44:39,210 --> 00:44:40,680 mechanically. 746 00:44:40,680 --> 00:44:45,630 So that's a computational burden, 747 00:44:45,630 --> 00:44:49,690 but it is one that other papers have faced as well. 748 00:44:49,690 --> 00:44:52,290 Now, what to do about talent-- 749 00:44:52,290 --> 00:44:53,880 is it insurable or not? 750 00:44:53,880 --> 00:44:57,120 Well, we kind of decided to try to be realistic 751 00:44:57,120 --> 00:44:58,770 and to not let it be insured. 752 00:45:02,080 --> 00:45:03,170 We could go the other way. 753 00:45:03,170 --> 00:45:06,640 In some sense, we know exactly how to do it. 754 00:45:06,640 --> 00:45:09,610 And there's some duals which would be satisfied 755 00:45:09,610 --> 00:45:11,470 if we had allowed it. 756 00:45:11,470 --> 00:45:14,530 But we thought it was just on the side of realism 757 00:45:14,530 --> 00:45:20,510 to say don't worry about your job market. 758 00:45:23,990 --> 00:45:25,640 We're going to cover that. 759 00:45:25,640 --> 00:45:28,820 NYT is going to fully insure you. 760 00:45:28,820 --> 00:45:31,650 Well, of course, we could put moral hazard on the z thing. 761 00:45:31,650 --> 00:45:34,785 But anyway, we didn't get into that. 762 00:45:39,510 --> 00:45:44,082 So you're going assign the occupation, induce effort-- 763 00:45:44,082 --> 00:45:44,790 where did I say-- 764 00:45:44,790 --> 00:45:47,190 I guess I didn't say it. 765 00:45:47,190 --> 00:45:49,110 It'll come up in a second, hopefully. 766 00:45:53,540 --> 00:45:54,040 Yeah. 767 00:45:54,040 --> 00:45:54,860 It was right here. 768 00:45:59,400 --> 00:46:01,890 We actually also allow there to be 769 00:46:01,890 --> 00:46:07,060 some insurance and moral hazard issues on the worker side. 770 00:46:07,060 --> 00:46:12,430 Now, for simplicity, we give them this same p of e function. 771 00:46:12,430 --> 00:46:14,960 We could have allowed it to be different. 772 00:46:14,960 --> 00:46:16,100 What is this? 773 00:46:16,100 --> 00:46:20,020 Well, the idea is you can sort of put a lot of effort 774 00:46:20,020 --> 00:46:22,930 into working, and the firm is going 775 00:46:22,930 --> 00:46:26,300 to observe your total sort of productivity. 776 00:46:26,300 --> 00:46:30,850 But the firm doesn't see your effort. 777 00:46:30,850 --> 00:46:33,720 It's like piece rate in some sense. 778 00:46:33,720 --> 00:46:37,320 So we just got tired of being asymmetric, 779 00:46:37,320 --> 00:46:41,130 as so many papers are about-- you know, there's an insurance 780 00:46:41,130 --> 00:46:43,800 problem for intermediaries and wage earners don't 781 00:46:43,800 --> 00:46:46,410 suffer from things like that. 782 00:46:50,770 --> 00:46:52,840 So here's the timeline. 783 00:46:52,840 --> 00:46:55,740 The household comes in with-- 784 00:46:55,740 --> 00:46:58,680 a certain household i comes in with a certain amount 785 00:46:58,680 --> 00:47:01,980 of wealth and talent. 786 00:47:01,980 --> 00:47:06,550 There's going to be an assignment, if you will, 787 00:47:06,550 --> 00:47:09,150 of who's a worker or a firm. 788 00:47:09,150 --> 00:47:11,490 Then if a firm-- 789 00:47:11,490 --> 00:47:15,400 well, if a worker or firm ever gets determined, if a firm-- 790 00:47:15,400 --> 00:47:18,300 the capitalization and labor hiring of the firm 791 00:47:18,300 --> 00:47:23,560 gets determined, then the epsilon hits. 792 00:47:23,560 --> 00:47:24,670 Effort comes first. 793 00:47:24,670 --> 00:47:32,210 Epsilon comes after that and gets in the way, so to speak, 794 00:47:32,210 --> 00:47:34,970 especially if effort is not observed. 795 00:47:34,970 --> 00:47:40,430 So this is like idiosyncratic risk subject to moral hazard 796 00:47:40,430 --> 00:47:41,510 potentially. 797 00:47:41,510 --> 00:47:45,770 Then consumption and the level of your savings 798 00:47:45,770 --> 00:47:47,720 that you're carrying over putting 799 00:47:47,720 --> 00:47:50,150 into the bank for tomorrow, those things 800 00:47:50,150 --> 00:47:52,910 are functions potentially of those epsilons. 801 00:47:52,910 --> 00:47:56,870 So the point is, you do bear the consequences 802 00:47:56,870 --> 00:48:00,200 in the moral hazard model of shirking. 803 00:48:00,200 --> 00:48:04,640 If you shirked and your productivity 804 00:48:04,640 --> 00:48:06,470 were low as a consequence of that, 805 00:48:06,470 --> 00:48:09,560 you're going to have to take a hit in terms 806 00:48:09,560 --> 00:48:14,060 of lower consumption and lower savings from tomorrow on. 807 00:48:23,110 --> 00:48:30,600 So here's the optimization problem really of the bank, 808 00:48:30,600 --> 00:48:33,360 but it's more like an equilibrium outcome. 809 00:48:33,360 --> 00:48:38,430 The banks are basically competitive. 810 00:48:38,430 --> 00:48:40,290 There's free entry. 811 00:48:40,290 --> 00:48:44,770 And for any sort of cohort fully observed, by the way, 812 00:48:44,770 --> 00:48:48,450 az sort of people out there-- 813 00:48:48,450 --> 00:48:52,830 there's a lot of them in every little cubbyhole-- 814 00:48:52,830 --> 00:48:55,170 the banks compete to offer them contracts. 815 00:48:55,170 --> 00:48:59,890 In effect, they would bid down things 816 00:48:59,890 --> 00:49:03,380 until utility is maximized. 817 00:49:03,380 --> 00:49:07,660 So it's as if they're maximizing the utility of this cohort 818 00:49:07,660 --> 00:49:10,180 by choice of all these things I just went through. 819 00:49:14,020 --> 00:49:17,770 And this is the resource constraint 820 00:49:17,770 --> 00:49:21,650 that the risk syndicate faces. 821 00:49:21,650 --> 00:49:23,600 Partly it's familiar and partly not. 822 00:49:23,600 --> 00:49:27,800 Namely, you've got uses and sources. 823 00:49:27,800 --> 00:49:33,170 The uses of money or resources is in consumption and how much 824 00:49:33,170 --> 00:49:35,860 you save for next time. 825 00:49:35,860 --> 00:49:40,740 And the sources comes from if you're 826 00:49:40,740 --> 00:49:45,680 a firm in this syndicate, your net profits, 827 00:49:45,680 --> 00:49:48,660 after subtracting off the cost of labor and capital 828 00:49:48,660 --> 00:49:53,690 and depreciating the capital, and from the other guys 829 00:49:53,690 --> 00:49:54,920 who are wage earners. 830 00:49:58,700 --> 00:50:00,920 And as is standard in many models, 831 00:50:00,920 --> 00:50:07,970 you begin the period with basically the savings 832 00:50:07,970 --> 00:50:10,250 that you had as a bank. 833 00:50:10,250 --> 00:50:13,260 But now they've sort of accrued interest. 834 00:50:13,260 --> 00:50:14,730 So you have principal and interest 835 00:50:14,730 --> 00:50:16,022 at the beginning of the period. 836 00:50:16,022 --> 00:50:20,040 Now, it looks like a standard sort of incomplete markets 837 00:50:20,040 --> 00:50:23,700 model, where you have the choices between saving 838 00:50:23,700 --> 00:50:26,670 and, quote, "borrowing/lending." 839 00:50:26,670 --> 00:50:30,550 The difference here and-- well, and some profit maximization 840 00:50:30,550 --> 00:50:31,050 embedded. 841 00:50:31,050 --> 00:50:36,220 But the difference really is the summing over epsilon. 842 00:50:36,220 --> 00:50:42,860 So this is the total sort of assigned consumption 843 00:50:42,860 --> 00:50:45,890 and savings throughout the whole syndicate, 844 00:50:45,890 --> 00:50:48,140 the average per capita number. 845 00:50:48,140 --> 00:50:51,670 And this is the per capita sources. 846 00:50:51,670 --> 00:50:55,250 The intermediaries in trying to break even on every epsilon 847 00:50:55,250 --> 00:50:57,170 type in the population-- 848 00:50:57,170 --> 00:51:00,590 that's the risk-sharing part of it, 849 00:51:00,590 --> 00:51:02,570 that some people could have higher output 850 00:51:02,570 --> 00:51:04,410 than other people. 851 00:51:04,410 --> 00:51:08,100 but subject to incentives or insurance. 852 00:51:08,100 --> 00:51:11,660 The intermediary will smooth that by assigning, say, 853 00:51:11,660 --> 00:51:14,840 higher consumption than output for the low epsilon guys, 854 00:51:14,840 --> 00:51:18,360 and conversely for the high epsilon guys. 855 00:51:18,360 --> 00:51:18,860 Yes. 856 00:51:18,860 --> 00:51:21,170 AUDIENCE: So the financial intermediary, they 857 00:51:21,170 --> 00:51:24,620 can control the distribution between consumption and asset? 858 00:51:24,620 --> 00:51:25,850 PROFESSOR: Yeah. 859 00:51:25,850 --> 00:51:26,840 That's all observed. 860 00:51:26,840 --> 00:51:29,480 AUDIENCE: So household can't bypass that 861 00:51:29,480 --> 00:51:32,493 by saving themselves. 862 00:51:32,493 --> 00:51:33,410 PROFESSOR: They don't. 863 00:51:33,410 --> 00:51:33,910 Yeah. 864 00:51:33,910 --> 00:51:36,320 We assume not. 865 00:51:36,320 --> 00:51:45,830 But you could assume that they get 866 00:51:45,830 --> 00:51:48,170 their sort of assignment of assets, 867 00:51:48,170 --> 00:51:49,510 and then they act on their own. 868 00:51:49,510 --> 00:51:50,885 But they would put it in the bank 869 00:51:50,885 --> 00:51:52,170 because they accrue interest. 870 00:51:52,170 --> 00:51:55,570 So there's really no loss in letting it sit in the bank. 871 00:51:55,570 --> 00:52:00,093 Now, we can have banks competing with one another. 872 00:52:00,093 --> 00:52:01,135 That's a bit more subtle. 873 00:52:01,135 --> 00:52:01,660 AUDIENCE: [INAUDIBLE] the other way around though, right? 874 00:52:01,660 --> 00:52:03,100 PROFESSOR: Hmm? 875 00:52:03,100 --> 00:52:06,110 AUDIENCE: You can't like really-- 876 00:52:06,110 --> 00:52:06,610 OK. 877 00:52:06,610 --> 00:52:08,235 Like, the household can choose to save, 878 00:52:08,235 --> 00:52:12,828 but household can choose to consume too? 879 00:52:12,828 --> 00:52:13,370 I don't know. 880 00:52:13,370 --> 00:52:17,450 I feel that here, bank has more control over household, 881 00:52:17,450 --> 00:52:19,360 and they can say exactly how much you can-- 882 00:52:19,360 --> 00:52:21,640 PROFESSOR: Well, that's probably a better way 883 00:52:21,640 --> 00:52:22,720 to think about it anyway. 884 00:52:22,720 --> 00:52:28,060 Because as you'll see, these information frictions 885 00:52:28,060 --> 00:52:29,710 and collateral restrictions create 886 00:52:29,710 --> 00:52:33,947 pressures that make savings more or less than what households 887 00:52:33,947 --> 00:52:36,280 might want to do if they were just following their Euler 888 00:52:36,280 --> 00:52:36,790 equation. 889 00:52:36,790 --> 00:52:39,460 So in the end, I agree with you. 890 00:52:39,460 --> 00:52:41,200 What's going to be actually interesting 891 00:52:41,200 --> 00:52:42,670 is that it's going-- 892 00:52:42,670 --> 00:52:43,920 and you'll see this-- 893 00:52:43,920 --> 00:52:45,310 it's going to go a different way. 894 00:52:45,310 --> 00:52:49,060 In the collateral constraint model, 895 00:52:49,060 --> 00:52:53,050 you would like to borrow more, actually, 896 00:52:53,050 --> 00:52:55,340 in order to have more assets in the future. 897 00:52:55,340 --> 00:52:57,190 That's not allowed to happen. 898 00:52:57,190 --> 00:52:59,890 So they're going to be borrowing-constrained 899 00:52:59,890 --> 00:53:03,500 in terms of looking at their intertemporal consumption path. 900 00:53:03,500 --> 00:53:08,467 The guy subject to this sort of incentive moral hazard 901 00:53:08,467 --> 00:53:10,300 constraint, they're actually going to end up 902 00:53:10,300 --> 00:53:11,740 being savings-constrained. 903 00:53:11,740 --> 00:53:13,520 They would like to save more. 904 00:53:13,520 --> 00:53:18,760 So it's already a hint that the dynamics of the household 905 00:53:18,760 --> 00:53:21,160 decision problems are very different, 906 00:53:21,160 --> 00:53:23,710 not just of the firm, but the households. 907 00:53:27,380 --> 00:53:31,550 So what do we mean by the incentive constraint? 908 00:53:31,550 --> 00:53:35,570 Well, when effort is unobserved, you 909 00:53:35,570 --> 00:53:37,850 want to induce the assigned effort. 910 00:53:37,850 --> 00:53:40,010 So these are the good boys. 911 00:53:40,010 --> 00:53:45,200 They were recommended to do e, and they actually do it. 912 00:53:45,200 --> 00:53:48,920 So this is the distribution of output, epsilon, 913 00:53:48,920 --> 00:53:50,640 that results from it. 914 00:53:50,640 --> 00:53:54,200 And they go into tomorrow with the assigned levels 915 00:53:54,200 --> 00:53:57,590 of savings and draw talent. 916 00:53:57,590 --> 00:54:00,050 But effort is not seen. 917 00:54:00,050 --> 00:54:03,680 So they actually contemplate some 918 00:54:03,680 --> 00:54:05,660 out of equilibrium behavior, like doing 919 00:54:05,660 --> 00:54:10,140 e hat, which is other than e. 920 00:54:10,140 --> 00:54:12,430 Now, what are the consequences of that? 921 00:54:12,430 --> 00:54:14,430 Well, first of all, there's a direct consequence 922 00:54:14,430 --> 00:54:16,830 in the utility function, because shirking 923 00:54:16,830 --> 00:54:21,780 might be good in terms of higher utility. 924 00:54:21,780 --> 00:54:24,090 But also, there is a direct consequence 925 00:54:24,090 --> 00:54:27,220 in terms of the distribution of output. 926 00:54:27,220 --> 00:54:30,720 So this is sort of the moral hazard induced 927 00:54:30,720 --> 00:54:33,870 productivity consequence. 928 00:54:33,870 --> 00:54:38,340 And if you do some typical monotonicity or something, 929 00:54:38,340 --> 00:54:41,580 then higher effort is associated with likely 930 00:54:41,580 --> 00:54:44,490 getting higher epsilons. 931 00:54:44,490 --> 00:54:50,910 So that's a standard-looking incentive constraint. 932 00:54:50,910 --> 00:54:52,770 The-- yes. 933 00:54:52,770 --> 00:54:55,540 AUDIENCE: So probably we could introduce 934 00:54:55,540 --> 00:54:58,530 some costs associated with changing jobs 935 00:54:58,530 --> 00:55:01,682 between [INAUDIBLE]. 936 00:55:01,682 --> 00:55:04,399 Does it change the analysis at all? 937 00:55:08,172 --> 00:55:09,630 PROFESSOR: Well, you get to choose. 938 00:55:09,630 --> 00:55:10,210 But yeah. 939 00:55:10,210 --> 00:55:11,400 There's no extra cost. 940 00:55:14,040 --> 00:55:16,900 It would increase the dimensionality of the state 941 00:55:16,900 --> 00:55:19,980 space rather enormously, because then you'd not only 942 00:55:19,980 --> 00:55:23,940 have to worry about the current distribution of wealth 943 00:55:23,940 --> 00:55:26,370 and talent, you'd have to worry about what these guys were 944 00:55:26,370 --> 00:55:28,200 doing last period. 945 00:55:28,200 --> 00:55:31,620 And it's actually hard enough to get any kind of solutions, 946 00:55:31,620 --> 00:55:33,790 analytic or numeric. 947 00:55:33,790 --> 00:55:37,860 So I think that's the constraint here is keeping the state. 948 00:55:37,860 --> 00:55:41,070 I mean, there is-- you'll see that there's already 949 00:55:41,070 --> 00:55:42,960 a huge distribution of wealth, and we're 950 00:55:42,960 --> 00:55:46,410 going to have to keep track of how that histogram is evolving 951 00:55:46,410 --> 00:55:52,250 the population jointly with the talent distribution. 952 00:55:52,250 --> 00:55:56,960 So hopefully, good enough for starters. 953 00:55:56,960 --> 00:55:57,460 Oh. 954 00:55:57,460 --> 00:56:00,090 I'm going to skip this slide. 955 00:56:00,090 --> 00:56:02,590 I'll tell you what was on it, and then we'll come back to it 956 00:56:02,590 --> 00:56:04,060 when we get to the microdata. 957 00:56:07,950 --> 00:56:10,980 Basically, instead of solving directly 958 00:56:10,980 --> 00:56:15,600 for consumption as a function of epsilon and all of that stuff, 959 00:56:15,600 --> 00:56:20,950 we trick it into a linear programming problem. 960 00:56:20,950 --> 00:56:25,680 And we do that by keeping track of histograms of things, 961 00:56:25,680 --> 00:56:29,220 basically, the joint distribution of consumption, 962 00:56:29,220 --> 00:56:36,770 epsilon, recommended effort and so on, respecting the timing. 963 00:56:36,770 --> 00:56:39,860 So in words, hopefully, you would think 964 00:56:39,860 --> 00:56:41,650 that's kind of equivalent. 965 00:56:41,650 --> 00:56:43,320 Anyway. 966 00:56:43,320 --> 00:56:45,290 Now, what's the gain? 967 00:56:45,290 --> 00:56:48,710 The gain is it literally is a linear program. 968 00:56:48,710 --> 00:56:52,520 So that little module can be solved as a linear programming 969 00:56:52,520 --> 00:56:54,350 problem, and that's how we're computing 970 00:56:54,350 --> 00:56:59,020 the solutions to that part. 971 00:56:59,020 --> 00:57:03,060 But this isn't the right context to go into the notation. 972 00:57:03,060 --> 00:57:08,350 We'll do it in a very simple convex cost or a fixed cost 973 00:57:08,350 --> 00:57:10,510 problem first, and then build up to it later. 974 00:57:13,350 --> 00:57:16,600 So I'm missing something. 975 00:57:16,600 --> 00:57:17,100 Oh, yeah. 976 00:57:17,100 --> 00:57:17,640 This thing. 977 00:57:20,340 --> 00:57:23,100 So this is the other constraint, that the capitalization 978 00:57:23,100 --> 00:57:26,200 of the firm is just some proportion of your wealth. 979 00:57:26,200 --> 00:57:29,620 Lambda could be greater than 1, but it's not infinity. 980 00:57:29,620 --> 00:57:31,710 So this is the constraint that we had 981 00:57:31,710 --> 00:57:34,940 on so many of the other papers. 982 00:57:34,940 --> 00:57:36,540 And you can tell stories and model 983 00:57:36,540 --> 00:57:39,190 this about running away with a capital and so on. 984 00:57:39,190 --> 00:57:41,010 But this is the essence of it. 985 00:57:44,010 --> 00:57:45,110 We could do both. 986 00:57:45,110 --> 00:57:47,810 We're going to imagine that it's one or the other. 987 00:57:47,810 --> 00:57:52,260 I mean, both simultaneously for a given firm. 988 00:57:52,260 --> 00:57:56,040 But we're going to put people in one sector or the other. 989 00:57:56,040 --> 00:57:59,790 So then we're going to solve for the factors occupation choice, 990 00:57:59,790 --> 00:58:02,385 labor, hiring, capitalization. 991 00:58:02,385 --> 00:58:05,610 And you want to think about labor supply-- 992 00:58:05,610 --> 00:58:07,560 it's just only slightly tricky. 993 00:58:07,560 --> 00:58:09,470 X means entrepreneur. 994 00:58:09,470 --> 00:58:11,880 1 minus x means wage earner. 995 00:58:11,880 --> 00:58:15,780 What's the total labor supply coming from wage earners? 996 00:58:15,780 --> 00:58:18,750 Well, epsilon is units of labor supply. 997 00:58:18,750 --> 00:58:20,190 It's induced by the effort. 998 00:58:23,230 --> 00:58:25,300 There's a histogram. 999 00:58:25,300 --> 00:58:29,760 So we basically add up overall the epsilons 1000 00:58:29,760 --> 00:58:34,490 at the fractions with which they exist in the population. 1001 00:58:34,490 --> 00:58:36,940 And this is total labor supply. 1002 00:58:36,940 --> 00:58:47,920 And so then the rest is pretty less daunting 1003 00:58:47,920 --> 00:58:49,810 in terms of notation. 1004 00:58:49,810 --> 00:58:54,820 We have-- we're going to equate labor supply to labor demand. 1005 00:58:54,820 --> 00:58:58,780 This is the l being employed within the firms. 1006 00:58:58,780 --> 00:59:01,960 The firms vary in terms of their a and z. 1007 00:59:01,960 --> 00:59:05,380 And everyone's facing these economy-wide wage 1008 00:59:05,380 --> 00:59:07,960 and interest rates. 1009 00:59:07,960 --> 00:59:10,720 But this does depend on the az argument, 1010 00:59:10,720 --> 00:59:12,910 so we have to add up over-- 1011 00:59:12,910 --> 00:59:16,370 or integrate up over-- all the az people in the population. 1012 00:59:16,370 --> 00:59:18,580 So again, in practice, we're going 1013 00:59:18,580 --> 00:59:22,180 to have a large finite number of wealths, 1014 00:59:22,180 --> 00:59:24,120 a finite number of talents, and this is going 1015 00:59:24,120 --> 00:59:28,070 to be some kind of summation. 1016 00:59:28,070 --> 00:59:30,710 This is also the capital clearing constraint. 1017 00:59:30,710 --> 00:59:32,930 All that wealth sitting there-- forget the d. 1018 00:59:32,930 --> 00:59:35,720 Don't know why that's there. 1019 00:59:35,720 --> 00:59:38,360 You have all that wealth sitting in the bank, 1020 00:59:38,360 --> 00:59:45,420 and it's going to be lent out at interest to the firms that 1021 00:59:45,420 --> 00:59:46,560 are renting the capital. 1022 00:59:49,930 --> 00:59:57,370 So we grab some parameters and some functional forms. 1023 00:59:57,370 --> 01:00:00,232 These are pretty traditional. 1024 01:00:00,232 --> 01:00:02,440 I guess we didn't want to do anything unconventional, 1025 01:00:02,440 --> 01:00:05,860 because we want to show what difference the obstacles make 1026 01:00:05,860 --> 01:00:08,650 rather than force you into a utility function 1027 01:00:08,650 --> 01:00:10,390 you don't believe. 1028 01:00:10,390 --> 01:00:15,550 Also, we can grab from the literature values for these 1029 01:00:15,550 --> 01:00:16,900 that people believe. 1030 01:00:16,900 --> 01:00:24,130 I don't mean to belittling calibration all the time. 1031 01:00:24,130 --> 01:00:25,550 Beta is realistic here. 1032 01:00:25,550 --> 01:00:31,760 It's about basically 5%, or 95% of the future is valued. 1033 01:00:31,760 --> 01:00:33,560 Sigma's the degree of risk aversion. 1034 01:00:33,560 --> 01:00:34,700 It's 1.5. 1035 01:00:34,700 --> 01:00:39,080 People have values well within that range-- 1036 01:00:39,080 --> 01:00:44,440 just utility is a bit tricky to parameterize of effort, 1037 01:00:44,440 --> 01:00:46,960 but the power function is the Frisch elasticity. 1038 01:00:46,960 --> 01:00:48,760 People have estimated that. 1039 01:00:48,760 --> 01:00:54,940 This is a simple transition which we made up. 1040 01:00:54,940 --> 01:00:58,750 But in other papers, people talk about entry and exit of firms, 1041 01:00:58,750 --> 01:01:03,380 and we could imagine doing that. 1042 01:01:03,380 --> 01:01:05,300 And they are also similar to the parameters 1043 01:01:05,300 --> 01:01:08,490 that Alex and I estimate would likely [INAUDIBLE] functions. 1044 01:01:08,490 --> 01:01:09,620 So here is the-- 1045 01:01:09,620 --> 01:01:15,070 with the Euler equation, it's at the household side. 1046 01:01:15,070 --> 01:01:18,862 Borrowing-constrained households have a Lagrange multiplier 1047 01:01:18,862 --> 01:01:20,445 on that limited commitment constraint. 1048 01:01:20,445 --> 01:01:27,660 It turns out to be tomorrow, but in expectation, it 1049 01:01:27,660 --> 01:01:31,260 leads to this same phenomenon that the marginal utility 1050 01:01:31,260 --> 01:01:35,580 of consumption is high today and lower in expectation tomorrow. 1051 01:01:35,580 --> 01:01:38,940 So these guys are borrowing-constrained. 1052 01:01:38,940 --> 01:01:39,670 Yes. 1053 01:01:39,670 --> 01:01:42,400 AUDIENCE: Question, sir-- going back to [INAUDIBLE].. 1054 01:01:42,400 --> 01:01:44,825 So when you described it earlier, 1055 01:01:44,825 --> 01:01:48,670 so something that entrepreneur can do to [INAUDIBLE] 1056 01:01:48,670 --> 01:01:50,230 to make productivity high. 1057 01:01:50,230 --> 01:01:52,960 If it was mainly labor monitoring, 1058 01:01:52,960 --> 01:01:55,765 I guess it would maybe sort of really only scale with labor. 1059 01:01:55,765 --> 01:01:57,390 Would that-- how much would that change 1060 01:01:57,390 --> 01:01:59,080 what goes on in the model. 1061 01:01:59,080 --> 01:02:01,360 PROFESSOR: Oh, that's a bit like in the spirit 1062 01:02:01,360 --> 01:02:03,990 of Jeremy's model, and we're not doing that. 1063 01:02:03,990 --> 01:02:06,370 But yes, you could put-- 1064 01:02:06,370 --> 01:02:08,320 it's very much like what Jeremy is doing. 1065 01:02:08,320 --> 01:02:10,900 You could put the technology for monitoring effort 1066 01:02:10,900 --> 01:02:13,420 to create a signal at least of effort, 1067 01:02:13,420 --> 01:02:15,340 and that would mitigate-- that would move us 1068 01:02:15,340 --> 01:02:17,155 more toward the full insurance solution. 1069 01:02:21,870 --> 01:02:28,780 And this is the weird one. 1070 01:02:28,780 --> 01:02:32,370 And I'm not sure if you've seen this yet or not 1071 01:02:32,370 --> 01:02:34,110 in public finance. 1072 01:02:34,110 --> 01:02:38,400 But-- so I'm just going to basically assert 1073 01:02:38,400 --> 01:02:41,460 that the first order condition, when 1074 01:02:41,460 --> 01:02:44,370 you're subject to moral hazard, looks like this. 1075 01:02:44,370 --> 01:02:46,740 And for obvious reasons, this is sometimes 1076 01:02:46,740 --> 01:02:51,380 called the inverse Euler equation, 1077 01:02:51,380 --> 01:02:53,045 because this is 1 over u prime. 1078 01:02:56,210 --> 01:02:59,150 The point here is Jensen's inequality. 1079 01:02:59,150 --> 01:03:02,700 You know, if you replace-- 1080 01:03:02,700 --> 01:03:06,260 if you pull the expectation operator outside of it, 1081 01:03:06,260 --> 01:03:08,540 you'd have 1 over u prime, but the inverse 1082 01:03:08,540 --> 01:03:10,880 would take it back to u prime. 1083 01:03:10,880 --> 01:03:14,540 But when you move that expectation operator outside, 1084 01:03:14,540 --> 01:03:18,706 you're taking an average over the interior object. 1085 01:03:18,706 --> 01:03:19,550 I'm sorry. 1086 01:03:19,550 --> 01:03:22,100 This is an average over the interior object as opposed 1087 01:03:22,100 --> 01:03:25,160 to an integration from the outside. 1088 01:03:25,160 --> 01:03:29,940 So it is basically a classical example of Jensen's inequality, 1089 01:03:29,940 --> 01:03:33,790 and it makes the right-hand side higher. 1090 01:03:33,790 --> 01:03:36,777 That's mechanical, but hopefully-- 1091 01:03:36,777 --> 01:03:38,110 we've already talked about this. 1092 01:03:38,110 --> 01:03:44,720 You can see that in the limited commitment, 1093 01:03:44,720 --> 01:03:48,910 people cannot borrow as much as they want. 1094 01:03:48,910 --> 01:03:52,690 That's going to make the demand for funds less, 1095 01:03:52,690 --> 01:03:54,970 and the interest rate is going to be 1096 01:03:54,970 --> 01:03:58,390 less when all the economy is subject from that limited 1097 01:03:58,390 --> 01:04:00,010 commitment constraint, as opposed 1098 01:04:00,010 --> 01:04:04,060 to the moral hazard economy where households and firms are 1099 01:04:04,060 --> 01:04:05,800 savings-constrained. 1100 01:04:05,800 --> 01:04:07,240 So they would like to save more. 1101 01:04:07,240 --> 01:04:08,390 Savings is less. 1102 01:04:08,390 --> 01:04:09,130 It's scarce. 1103 01:04:09,130 --> 01:04:10,630 You have a higher interest rate. 1104 01:04:10,630 --> 01:04:13,670 So these interest rate numbers kind of make sense. 1105 01:04:13,670 --> 01:04:16,950 Don't be too spooked by the negative interest rate. 1106 01:04:16,950 --> 01:04:20,620 You know, we have a depreciation rate in this economy. 1107 01:04:20,620 --> 01:04:23,230 So it's still sort of on net-- 1108 01:04:23,230 --> 01:04:24,250 it's positive. 1109 01:04:27,180 --> 01:04:32,610 And then you can see, comparing these two different countries, 1110 01:04:32,610 --> 01:04:36,810 GDP, TFP-- 1111 01:04:36,810 --> 01:04:39,450 the TFP dynamics are very different, 1112 01:04:39,450 --> 01:04:42,270 although the numbers, though different, 1113 01:04:42,270 --> 01:04:44,040 are not radically different. 1114 01:04:46,770 --> 01:04:49,740 TFP is dragged down in the limited commitment economy, 1115 01:04:49,740 --> 01:04:53,760 because high productivity firms cannot borrow as much as they 1116 01:04:53,760 --> 01:04:55,030 want. 1117 01:04:55,030 --> 01:04:57,930 They can't exploit their z. 1118 01:04:57,930 --> 01:05:00,690 Whereas, in the moral hazard economy, 1119 01:05:00,690 --> 01:05:04,065 you have to induce effort, and that's kind of a drag. 1120 01:05:10,540 --> 01:05:15,520 I mean, if you were draconian and you had no insurance, 1121 01:05:15,520 --> 01:05:19,400 then yes, people would be working very hard. 1122 01:05:19,400 --> 01:05:21,140 But that's actually not optimal. 1123 01:05:21,140 --> 01:05:24,990 The right thing to do is to have this blend between insurance 1124 01:05:24,990 --> 01:05:27,440 and productivity. 1125 01:05:27,440 --> 01:05:30,760 So partially, it's-- this language people use about moral 1126 01:05:30,760 --> 01:05:33,820 hazard in the banking system and in the press 1127 01:05:33,820 --> 01:05:36,280 and the policymakers, pick that up like, 1128 01:05:36,280 --> 01:05:38,140 let's get rid of moral hazard. 1129 01:05:38,140 --> 01:05:40,930 That's not the right way to think about it. 1130 01:05:40,930 --> 01:05:42,880 Moral hazard is an information problem. 1131 01:05:42,880 --> 01:05:46,210 You want incentives to do as best you can, 1132 01:05:46,210 --> 01:05:47,830 given the moral hazard problem. 1133 01:05:47,830 --> 01:05:51,010 But it's like-- eliminate moral hazard is like, let's 1134 01:05:51,010 --> 01:05:52,450 just shut down all the insurance, 1135 01:05:52,450 --> 01:05:53,490 then you don't have a problem. 1136 01:05:53,490 --> 01:05:55,115 But yeah, well, you got other problems. 1137 01:06:00,160 --> 01:06:01,507 Wages are a bit different. 1138 01:06:01,507 --> 01:06:02,965 They're higher in the moral-- these 1139 01:06:02,965 --> 01:06:05,620 are economy-wide wages and interest rates, of course. 1140 01:06:10,870 --> 01:06:12,790 But you can also see a bigger difference 1141 01:06:12,790 --> 01:06:14,725 in the financing ratio. 1142 01:06:17,290 --> 01:06:22,540 Actually, labor here tends to compensate for the more 1143 01:06:22,540 --> 01:06:26,410 restricted capital, and the drag makes labor lower 1144 01:06:26,410 --> 01:06:27,820 in the moral hazard economy. 1145 01:06:32,260 --> 01:06:35,800 External finance is more limited in the limited commitment 1146 01:06:35,800 --> 01:06:36,340 economy. 1147 01:06:36,340 --> 01:06:36,910 Why? 1148 01:06:36,910 --> 01:06:37,990 Because you can't borrow. 1149 01:06:37,990 --> 01:06:40,660 You run into that lambda constraint. 1150 01:06:40,660 --> 01:06:44,090 There's nothing like that in the moral hazard economy. 1151 01:06:44,090 --> 01:06:48,314 So the external finance to GDP ratio is higher. 1152 01:06:48,314 --> 01:06:52,512 AUDIENCE: So why is the interest rate negative? 1153 01:06:52,512 --> 01:06:54,220 PROFESSOR: That's, again, because there's 1154 01:06:54,220 --> 01:06:58,150 an interest rate going on here. 1155 01:06:58,150 --> 01:07:00,460 I mean capital gets utilized, and it gets-- 1156 01:07:00,460 --> 01:07:01,390 it's depreciating. 1157 01:07:01,390 --> 01:07:05,030 And you've got to basically pay for that somehow. 1158 01:07:05,030 --> 01:07:07,120 So it lowers the net yield. 1159 01:07:09,790 --> 01:07:12,990 And here's interesting-- you know, 1160 01:07:12,990 --> 01:07:16,800 I didn't get to say much at all about Joaquin Blaum's paper 1161 01:07:16,800 --> 01:07:18,330 on inequality. 1162 01:07:18,330 --> 01:07:21,090 But I get to be reminded to say something now, 1163 01:07:21,090 --> 01:07:26,130 because here the wealth inequality curve's generated 1164 01:07:26,130 --> 01:07:29,370 by the different constraints. 1165 01:07:29,370 --> 01:07:33,240 And you can see there's actually more inequality 1166 01:07:33,240 --> 01:07:35,870 in the limited commitment economy 1167 01:07:35,870 --> 01:07:38,386 than there is in the moral hazard economy. 1168 01:07:42,580 --> 01:07:47,053 But again, that has to do with this dispersion of the capital. 1169 01:07:47,053 --> 01:07:49,510 Capital tends to be more-- 1170 01:07:49,510 --> 01:07:52,000 not identical, but more compressed 1171 01:07:52,000 --> 01:07:55,680 in the moral hazard economy. 1172 01:07:55,680 --> 01:07:56,950 And capital is very limited. 1173 01:07:56,950 --> 01:08:04,840 You can't take advantage of your productivity in the other one. 1174 01:08:04,840 --> 01:08:08,050 So the wealth dispersion is playing a bigger role, 1175 01:08:08,050 --> 01:08:10,990 and feeding back in turn to-- 1176 01:08:10,990 --> 01:08:13,990 it's like you can't get the convergences. 1177 01:08:17,010 --> 01:08:19,200 And then we take 50-50. 1178 01:08:19,200 --> 01:08:23,484 Same economy-- now we have the urban/rural of the same size, 1179 01:08:23,484 --> 01:08:23,984 and-- 1180 01:08:26,880 --> 01:08:27,760 AUDIENCE: Question. 1181 01:08:27,760 --> 01:08:29,220 PROFESSOR: Yep. 1182 01:08:29,220 --> 01:08:32,250 AUDIENCE: So in the slides with the parameters, 1183 01:08:32,250 --> 01:08:35,043 you don't talk about lambda. 1184 01:08:35,043 --> 01:08:39,229 How much do you think that to be? 1185 01:08:39,229 --> 01:08:41,340 The-- so the-- how much is the constraint? 1186 01:08:41,340 --> 01:08:43,939 PROFESSOR: I'm not sure I remember the number. 1187 01:08:43,939 --> 01:08:45,180 AUDIENCE: But-- I mean, do you remember how you configured it? 1188 01:08:45,180 --> 01:08:46,290 PROFESSOR: It should have been on that list of parameters? 1189 01:08:46,290 --> 01:08:47,500 It wasn't on that page? 1190 01:08:47,500 --> 01:08:48,000 No. 1191 01:08:48,000 --> 01:08:49,950 I don't remember. 1192 01:08:49,950 --> 01:08:52,830 We probably-- it was probably something like 1.3, 1193 01:08:52,830 --> 01:08:54,703 because that's-- 1194 01:08:54,703 --> 01:08:56,370 but that's right off the top of my head. 1195 01:08:56,370 --> 01:08:56,939 AUDIENCE: How do you-- 1196 01:08:56,939 --> 01:08:58,265 [INAUDIBLE] in the same way? 1197 01:08:58,265 --> 01:08:59,640 Because that's how-- like, that's 1198 01:08:59,640 --> 01:09:00,960 something you estimated somewhere else? 1199 01:09:00,960 --> 01:09:01,979 PROFESSOR: Hopefully, we just drew it 1200 01:09:01,979 --> 01:09:03,810 from this literature, the [INAUDIBLE] paper 1201 01:09:03,810 --> 01:09:06,750 that you've seen before. 1202 01:09:06,750 --> 01:09:08,399 I don't think we matched it with data. 1203 01:09:10,920 --> 01:09:13,522 That's actually hard to pin down in the data. 1204 01:09:13,522 --> 01:09:14,939 AUDIENCE: Would it be particularly 1205 01:09:14,939 --> 01:09:17,010 sensitive to the choice of lambda? 1206 01:09:17,010 --> 01:09:18,660 PROFESSOR: Oh, yeah. 1207 01:09:18,660 --> 01:09:19,160 Yeah. 1208 01:09:19,160 --> 01:09:23,020 Remember the [INAUDIBLE] thing, where lambda goes from-- 1209 01:09:23,020 --> 01:09:25,990 I think it was 1.3 to infinity. 1210 01:09:25,990 --> 01:09:28,410 And at infinity, there's no constraint at all. 1211 01:09:28,410 --> 01:09:30,870 You're back to the full insurance, full productivity 1212 01:09:30,870 --> 01:09:31,392 solution. 1213 01:09:31,392 --> 01:09:32,934 AUDIENCE: So some of the things about 1214 01:09:32,934 --> 01:09:36,647 like the wealth inequality and [INAUDIBLE] 1215 01:09:36,647 --> 01:09:37,689 expect those [INAUDIBLE]. 1216 01:09:37,689 --> 01:09:38,800 PROFESSOR: Oh, yeah. 1217 01:09:38,800 --> 01:09:41,590 Now what-- so what you're not seeing here, 1218 01:09:41,590 --> 01:09:44,380 although we have recently done this, 1219 01:09:44,380 --> 01:09:49,450 is run a whole suite of computations for all kinds 1220 01:09:49,450 --> 01:09:53,170 of-- we put sort of error bands or confidence intervals 1221 01:09:53,170 --> 01:09:59,100 on these just to make sure that none of the things that we want 1222 01:09:59,100 --> 01:10:01,320 to emphasize are-- 1223 01:10:01,320 --> 01:10:03,110 either they're always true, or we're 1224 01:10:03,110 --> 01:10:04,660 going to say they're not always true, 1225 01:10:04,660 --> 01:10:06,700 and they depend on these parameters. 1226 01:10:06,700 --> 01:10:09,060 But you're absolutely right, that just showing 1227 01:10:09,060 --> 01:10:12,710 some simulations looks a bit arbitrary, especially 1228 01:10:12,710 --> 01:10:14,460 for things like lambda, which I'm not even 1229 01:10:14,460 --> 01:10:15,752 remembering in the moment, so-- 1230 01:10:18,680 --> 01:10:20,930 But different things would matter for the moral hazard 1231 01:10:20,930 --> 01:10:21,710 economy too. 1232 01:10:21,710 --> 01:10:25,190 You know, the curvature of the labor supply utility 1233 01:10:25,190 --> 01:10:26,690 that generates labor supply-- 1234 01:10:26,690 --> 01:10:29,630 that's going to be that Frisch elasticity. 1235 01:10:29,630 --> 01:10:33,500 We'll see that again when we get to the sort of microdata. 1236 01:10:33,500 --> 01:10:36,660 And the chi that sits in front of labor disutility-- 1237 01:10:36,660 --> 01:10:41,450 that's a huge number, important number. 1238 01:10:41,450 --> 01:10:44,130 So what are these slides? 1239 01:10:44,130 --> 01:10:45,960 Well, this is a bit like generating 1240 01:10:45,960 --> 01:10:47,670 many, many, many simulations. 1241 01:10:47,670 --> 01:10:52,290 We have many economies, and we vary this fraction 1242 01:10:52,290 --> 01:10:53,460 from zero to one. 1243 01:10:53,460 --> 01:10:56,580 I just showed you the 50-50 economy. 1244 01:10:56,580 --> 01:11:00,510 But this allows everyone to be in limited commitment, everyone 1245 01:11:00,510 --> 01:11:04,500 to be in moral hazard, or anywhere in between. 1246 01:11:04,500 --> 01:11:08,100 And so you can see how GDP, TFP, and so on, 1247 01:11:08,100 --> 01:11:13,380 how these things move around with the fraction 1248 01:11:13,380 --> 01:11:17,760 of the population that are subject to moral hazard. 1249 01:11:17,760 --> 01:11:24,450 And partly, this mirrors what I said before in labor supply. 1250 01:11:24,450 --> 01:11:27,180 But on other things, you can see things aren't monotone. 1251 01:11:27,180 --> 01:11:29,330 So those are surprises. 1252 01:11:29,330 --> 01:11:32,880 It actually goes up, and then it comes down again. 1253 01:11:32,880 --> 01:11:35,190 GDP climbs pretty fast, and then it kind of 1254 01:11:35,190 --> 01:11:37,390 wobbles around a bit. 1255 01:11:37,390 --> 01:11:39,960 We've tried to get rid of as many wobbles as possible 1256 01:11:39,960 --> 01:11:42,900 for worries that they're just computational. 1257 01:11:42,900 --> 01:11:45,480 But some of them-- some of the wobbles remain. 1258 01:11:45,480 --> 01:11:48,970 So I've kind of learned to not pay too much attention. 1259 01:11:48,970 --> 01:11:52,620 But this is not numerical error. 1260 01:11:52,620 --> 01:11:56,952 This is systematic, this drop. 1261 01:11:56,952 --> 01:11:59,160 And you can see what the wages and the interest rates 1262 01:11:59,160 --> 01:12:04,390 are doing, and the fraction of firms for that matter. 1263 01:12:04,390 --> 01:12:11,880 Now, here is back to 1/2, 1/2. 1264 01:12:11,880 --> 01:12:17,370 But within that economy, we have the limited commitment sector 1265 01:12:17,370 --> 01:12:18,540 and the moral hazard sector. 1266 01:12:22,230 --> 01:12:25,680 Obviously, there is economy-wide wages 1267 01:12:25,680 --> 01:12:28,170 and interest rates are the same. 1268 01:12:28,170 --> 01:12:30,760 It's one economy clearing. 1269 01:12:30,760 --> 01:12:33,910 But other things vary across these two columns. 1270 01:12:33,910 --> 01:12:35,740 Probably the most exciting part is 1271 01:12:35,740 --> 01:12:40,610 this, which is let's look at labor, for example. 1272 01:12:40,610 --> 01:12:44,030 Labor employed in the sector 0.38; 1273 01:12:44,030 --> 01:12:47,650 labor supplied by the sector 0.53. 1274 01:12:47,650 --> 01:12:50,500 So they're exporters of labor, if you allow 1275 01:12:50,500 --> 01:12:53,270 me to use that terminology. 1276 01:12:53,270 --> 01:12:56,090 Or people are costlessly migrating 1277 01:12:56,090 --> 01:13:00,260 from the limited commitment sector to the moral hazard. 1278 01:13:00,260 --> 01:13:03,500 And indeed, you'll pick up the other side of this thing, 1279 01:13:03,500 --> 01:13:09,560 basically, labor supplied, 0.46; labor demanded, 0.61. 1280 01:13:09,560 --> 01:13:12,230 Now, you say, why isn't it all zero and one. 1281 01:13:12,230 --> 01:13:13,190 No, no, no, no, no. 1282 01:13:13,190 --> 01:13:14,420 Remember the z's. 1283 01:13:14,420 --> 01:13:15,680 Remember the productivity. 1284 01:13:15,680 --> 01:13:19,400 Every sector has some really inefficient people 1285 01:13:19,400 --> 01:13:21,380 who shouldn't be running firms. 1286 01:13:21,380 --> 01:13:24,350 They're going to supply labor no matter what, 1287 01:13:24,350 --> 01:13:28,880 or at least at these equilibrium wages. 1288 01:13:28,880 --> 01:13:32,210 And the same thing is true with the flow of funds 1289 01:13:32,210 --> 01:13:35,160 if we had the geographic decomposition-- 1290 01:13:35,160 --> 01:13:37,790 and we got Mexico to do this, by the way. 1291 01:13:37,790 --> 01:13:41,180 It's really cool-- through their CNBV, 1292 01:13:41,180 --> 01:13:42,530 their financial regulator. 1293 01:13:47,003 --> 01:13:48,670 And the question there, is money flowing 1294 01:13:48,670 --> 01:13:51,250 to Mexico City from the outlying areas 1295 01:13:51,250 --> 01:13:53,140 as they improve intermediation? 1296 01:13:53,140 --> 01:13:57,040 And what consequences does that have for the outlying areas? 1297 01:13:57,040 --> 01:14:01,540 Well, here, money's flowing from the limited commitments sector 1298 01:14:01,540 --> 01:14:02,840 to the moral hazard sector. 1299 01:14:02,840 --> 01:14:05,650 So this is where the action is, so to speak. 1300 01:14:05,650 --> 01:14:07,390 We don't get a lot of difference in terms 1301 01:14:07,390 --> 01:14:09,700 of the numbers of firms, but you're clearly 1302 01:14:09,700 --> 01:14:12,920 seeing differences in terms of employment size. 1303 01:14:12,920 --> 01:14:18,030 And remember the China paper was like this, where labor-- 1304 01:14:18,030 --> 01:14:22,320 the allocation of labor was playing a huge role. 1305 01:14:39,240 --> 01:14:41,720 So-- sorry. 1306 01:14:41,720 --> 01:14:43,805 We had more results over the weekend. 1307 01:14:48,270 --> 01:14:49,682 We've been trying to get these. 1308 01:14:49,682 --> 01:14:51,390 This, we've had for a long time, but I'll 1309 01:14:51,390 --> 01:14:54,885 show you something else that's quite exciting. 1310 01:14:57,580 --> 01:14:59,300 First of all, speed of transitions-- 1311 01:14:59,300 --> 01:15:05,370 you remember that where a paper was all about the puzzle, which 1312 01:15:05,370 --> 01:15:08,220 is, why don't miracle Asian economies grow 1313 01:15:08,220 --> 01:15:17,220 even faster if you believe they were in a solo world? 1314 01:15:17,220 --> 01:15:18,360 But of course, there is-- 1315 01:15:18,360 --> 01:15:20,290 and then they assume this financial friction, 1316 01:15:20,290 --> 01:15:22,850 which slowed things down. 1317 01:15:22,850 --> 01:15:26,450 The friction you assume is going to make a big difference 1318 01:15:26,450 --> 01:15:28,340 to the speed of transitions. 1319 01:15:28,340 --> 01:15:30,560 Actually, here, even in the steady state, 1320 01:15:30,560 --> 01:15:34,700 you can ask how fast do people go from an az state 1321 01:15:34,700 --> 01:15:37,250 to an a prime z prime state? 1322 01:15:37,250 --> 01:15:40,220 This is some big Markov matrix. 1323 01:15:40,220 --> 01:15:42,230 It's sort of steady state, but it's 1324 01:15:42,230 --> 01:15:45,680 transition from the household level 1325 01:15:45,680 --> 01:15:48,240 from one state to another. 1326 01:15:48,240 --> 01:15:53,540 So you may remember, you can multiply those matrices 1327 01:15:53,540 --> 01:15:57,530 over and over times each other, or you can basically 1328 01:15:57,530 --> 01:16:02,240 do an eigenvalue, eigenvector decomposition. 1329 01:16:02,240 --> 01:16:07,970 And the closer this eigenvalue is to 1, the slower things are. 1330 01:16:07,970 --> 01:16:09,080 Things are moving. 1331 01:16:09,080 --> 01:16:11,120 The next multiplication of the matrix 1332 01:16:11,120 --> 01:16:13,280 is kind of very similar-- 1333 01:16:13,280 --> 01:16:16,190 that times the control [INAUDIBLE] 1334 01:16:16,190 --> 01:16:18,840 very similar to what is in the previous period. 1335 01:16:18,840 --> 01:16:21,650 It's 0.93. 1336 01:16:21,650 --> 01:16:24,260 That's a very-- sorry. 1337 01:16:24,260 --> 01:16:29,060 0.93 for one; 0.98 for the other. 1338 01:16:29,060 --> 01:16:34,070 So there's roughly three more times difference 1339 01:16:34,070 --> 01:16:38,090 in the speed of sort of within steady state transitions going 1340 01:16:38,090 --> 01:16:38,970 on. 1341 01:16:38,970 --> 01:16:43,123 Now, there's a technical reason for this, basically. 1342 01:16:43,123 --> 01:16:44,540 On the limited commitment thing we 1343 01:16:44,540 --> 01:16:46,880 have this forward-looking savings behavior. 1344 01:16:46,880 --> 01:16:48,710 So people are going to save their way out 1345 01:16:48,710 --> 01:16:50,210 of these constraints. 1346 01:16:50,210 --> 01:16:52,370 And they can do it pretty quickly. 1347 01:16:52,370 --> 01:16:55,220 In fact, you're going to see some homework 1348 01:16:55,220 --> 01:17:03,080 and so on motivated by thinking through [INAUDIBLE] papers 1349 01:17:03,080 --> 01:17:05,270 and job market papers and so on, that kind of 1350 01:17:05,270 --> 01:17:12,380 get at this thing, which is-- the twist is, why doesn't money 1351 01:17:12,380 --> 01:17:13,610 flow from the US to India? 1352 01:17:13,610 --> 01:17:17,600 The answer is, why doesn't India save its way out of constraints 1353 01:17:17,600 --> 01:17:19,730 more quickly than they seem to? 1354 01:17:23,320 --> 01:17:26,290 So there's discussion in the literature about that. 1355 01:17:26,290 --> 01:17:27,820 But here we do have constraints. 1356 01:17:27,820 --> 01:17:29,890 It's just that they're different. 1357 01:17:29,890 --> 01:17:32,470 And essentially what's going on here 1358 01:17:32,470 --> 01:17:35,830 is you don't want to-- you don't want that wealth 1359 01:17:35,830 --> 01:17:38,280 to move very fast. 1360 01:17:38,280 --> 01:17:41,880 You're going to balance off the incentives 1361 01:17:41,880 --> 01:17:46,170 of today's consumption versus tomorrow's wealth. 1362 01:17:46,170 --> 01:17:48,570 Wealth moves a little bit, depending on whether epsilon 1363 01:17:48,570 --> 01:17:49,230 is high or low. 1364 01:17:49,230 --> 01:17:50,980 But you don't want it to move a whole lot. 1365 01:17:50,980 --> 01:17:55,350 Because if it did, that's the bulk of your utility. 1366 01:17:55,350 --> 01:17:56,430 That's kind of very-- 1367 01:17:56,430 --> 01:17:59,640 that is actually algebraically in the dual, 1368 01:17:59,640 --> 01:18:02,940 equivalent with discounted expected utility. 1369 01:18:02,940 --> 01:18:05,820 And so if you move that quickly, there's 1370 01:18:05,820 --> 01:18:09,030 a big welfare loss because of the concavity. 1371 01:18:09,030 --> 01:18:11,530 So you don't want it to move fast. 1372 01:18:11,530 --> 01:18:14,700 So that's-- OK. 1373 01:18:14,700 --> 01:18:18,720 And then if you look at growth rates, 1374 01:18:18,720 --> 01:18:22,890 the distribution of growth rates looks like it's all zero. 1375 01:18:22,890 --> 01:18:24,930 Actually, what the point is, the growth rates 1376 01:18:24,930 --> 01:18:26,610 are all very similar to each other, 1377 01:18:26,610 --> 01:18:29,280 because there isn't this big dispersion in the moral hazard 1378 01:18:29,280 --> 01:18:32,663 economy because they don't move that fast. 1379 01:18:32,663 --> 01:18:34,080 But a limited commitment-- there's 1380 01:18:34,080 --> 01:18:36,890 a huge dispersion and growth rate. 1381 01:18:36,890 --> 01:18:40,290 And you can imagine going to data with this kind of stuff. 1382 01:18:40,290 --> 01:18:42,670 We didn't look at this in the data. 1383 01:18:42,670 --> 01:18:44,970 But if you have Panel, you can actually 1384 01:18:44,970 --> 01:18:47,730 look not only at static firm size distributions 1385 01:18:47,730 --> 01:18:52,650 but the growth rates, and get some evidence one way 1386 01:18:52,650 --> 01:18:56,170 or the other for one friction. 1387 01:18:56,170 --> 01:18:57,540 Then finally, the weekend-- 1388 01:19:00,090 --> 01:19:02,880 we've had a devil of a time doing this. 1389 01:19:02,880 --> 01:19:05,250 We start with the whole economy subject 1390 01:19:05,250 --> 01:19:07,590 to a limited commitment constraint. 1391 01:19:07,590 --> 01:19:12,450 And then we move it after 10 periods to a moral hazard. 1392 01:19:12,450 --> 01:19:15,420 Now, this is meant to mimic some kind of financial reform. 1393 01:19:15,420 --> 01:19:17,430 If you want to be generous with me, 1394 01:19:17,430 --> 01:19:19,650 think of there being two constraints, 1395 01:19:19,650 --> 01:19:22,170 and then we managed to get rid of one. 1396 01:19:22,170 --> 01:19:23,880 Well, that's not quite what we're doing. 1397 01:19:23,880 --> 01:19:27,220 We kind of substitute limited commitment for moral hazard. 1398 01:19:27,220 --> 01:19:31,530 Basically, they get rid of a constraining legal system, 1399 01:19:31,530 --> 01:19:34,710 but it turns out that created an information problem. 1400 01:19:34,710 --> 01:19:38,570 Anyway, what's really important is 1401 01:19:38,570 --> 01:19:43,740 that you can actually look at the transition dynamics. 1402 01:19:48,100 --> 01:19:51,130 Well, let me just show you the pictures. 1403 01:19:51,130 --> 01:19:55,770 So here, you're in a steady state, 1404 01:19:55,770 --> 01:19:58,230 and then you do a financial reform. 1405 01:19:58,230 --> 01:19:59,280 But it's not like-- 1406 01:20:01,830 --> 01:20:06,060 Kaboski, they did-- they left the financial structure intact, 1407 01:20:06,060 --> 01:20:09,300 and they did real reforms. 1408 01:20:09,300 --> 01:20:11,970 We're leaving everything in the environment and everything 1409 01:20:11,970 --> 01:20:14,790 intact and doing a financial reform, hopefully 1410 01:20:14,790 --> 01:20:18,150 in the spirit of what Levine might have wanted. 1411 01:20:18,150 --> 01:20:22,260 And then we see the implications of that for deeper levels 1412 01:20:22,260 --> 01:20:23,460 of capitalization. 1413 01:20:23,460 --> 01:20:25,085 And look at the wage rate. 1414 01:20:25,085 --> 01:20:29,310 It's no wonder we had trouble finding it-- 1415 01:20:29,310 --> 01:20:31,710 pops up like that instantaneously. 1416 01:20:34,360 --> 01:20:38,320 Now, let me just say, these things are not easy to compute. 1417 01:20:38,320 --> 01:20:41,380 Because-- and [INAUDIBLE] going to say something 1418 01:20:41,380 --> 01:20:43,480 about that on Friday. 1419 01:20:43,480 --> 01:20:46,030 Essentially, when you're out of the steady state, 1420 01:20:46,030 --> 01:20:49,610 you have to figure out where you're going to go 1421 01:20:49,610 --> 01:20:52,090 and how long it's going to take to get there. 1422 01:20:52,090 --> 01:20:54,280 And then you've got to sort of guess 1423 01:20:54,280 --> 01:20:57,700 about the paths of the interest rates and the wages. 1424 01:20:57,700 --> 01:21:00,730 And of course, any arbitrary guess about how long it takes 1425 01:21:00,730 --> 01:21:03,550 and how they're going to move is quite arbitrary. 1426 01:21:03,550 --> 01:21:05,830 So then you need a systematic way 1427 01:21:05,830 --> 01:21:09,460 to adjust period by period to get in new guesses 1428 01:21:09,460 --> 01:21:12,310 and then to try to iterate. 1429 01:21:12,310 --> 01:21:16,450 [INAUDIBLE] managed to do it. 1430 01:21:16,450 --> 01:21:17,320 We have their code. 1431 01:21:17,320 --> 01:21:19,630 I've shared it with many students 1432 01:21:19,630 --> 01:21:21,850 working on related problems. 1433 01:21:21,850 --> 01:21:25,180 It turned out that their code, which 1434 01:21:25,180 --> 01:21:28,090 is getting better and better because they're playing around 1435 01:21:28,090 --> 01:21:33,030 with it too, works for this limited commitment better-- 1436 01:21:33,030 --> 01:21:34,890 much, much better than it ever worked for us 1437 01:21:34,890 --> 01:21:35,700 with moral hazard. 1438 01:21:35,700 --> 01:21:37,890 We were quite despondent about being 1439 01:21:37,890 --> 01:21:40,620 able to compute the transitions. 1440 01:21:40,620 --> 01:21:43,860 But evidently, now we're in business on that. 1441 01:21:51,220 --> 01:21:55,210 So the summary, back to the big picture, 1442 01:21:55,210 --> 01:21:58,060 is these obstacles matter. 1443 01:21:58,060 --> 01:22:01,330 Ideally, I think, if you can manage, 1444 01:22:01,330 --> 01:22:05,740 use the microdata to try to take a stand on what 1445 01:22:05,740 --> 01:22:07,330 the frictions are. 1446 01:22:07,330 --> 01:22:12,820 And through the dynamics of Euler equations and debt 1447 01:22:12,820 --> 01:22:17,930 overhang and so on, you'll generate these dynamic paths. 1448 01:22:17,930 --> 01:22:21,040 And they're going to be general equilibrium 1449 01:22:21,040 --> 01:22:23,680 consequences of the reforms. 1450 01:22:23,680 --> 01:22:28,360 So it's like doing an experiment on these economies 1451 01:22:28,360 --> 01:22:29,870 to see what would happen. 1452 01:22:29,870 --> 01:22:34,120 And this allows us to begin to answer those questions I went 1453 01:22:34,120 --> 01:22:37,870 over in the introductory lecture about finance causes growth, 1454 01:22:37,870 --> 01:22:38,830 but then what? 1455 01:22:38,830 --> 01:22:40,330 What do we do? 1456 01:22:40,330 --> 01:22:43,450 And how do we engineer it? 1457 01:22:43,450 --> 01:22:45,110 And what would we expect to happen? 1458 01:22:45,110 --> 01:22:48,420 And are there winners and losers, and so on?