YouTube2Text

YouTube transcript (DKWVDvk1QYo) — Transcript

10,446 words · 1,734 segments · language en · Watch on YouTube

Full transcript

  1. 0:00Welcome to the meetings.
  2. 0:02Uh, into this particular session, I'm
  3. 0:04impressed to draw off this early to come
  4. 0:06out for this session and
  5. 0:09glad glad to have all of you here. We
  6. 0:11can talk for a little while on the
  7. 0:13subject, but you know, my plan is to
  8. 0:15leave plenty of time
  9. 0:17at the end for
  10. 0:19for Q&A. So,
  11. 0:20if you have any any thoughts along the
  12. 0:22way, throw them out and
  13. 0:24I'm sure it won't be
  14. 0:26unique.
  15. 0:27Um,
  16. 0:29So, this this topic is one I think that
  17. 0:32those those of you, presumably all of
  18. 0:34you since you're here, do work in in the
  19. 0:36area of corporate finance encounter
  20. 0:40problems associated with endogeneity and
  21. 0:42identification in in your empirical
  22. 0:45work. And so, I thought I would start
  23. 0:47like I do sometimes in a in a classroom
  24. 0:49setting. Start with a question
  25. 0:51and say, "Of among all of you who have
  26. 0:55submitted papers in in the field of
  27. 0:57corporate finance, how many of you
  28. 0:59have
  29. 1:00a referee or an editor say,
  30. 1:03"I don't buy your identification
  31. 1:05strategy." or something to that effect?
  32. 1:10That's all, really? Does that mean the
  33. 1:13rest of you haven't submitted papers yet
  34. 1:14to
  35. 1:15to this conference?
  36. 1:17I think you're hard-pressed
  37. 1:20in corporate finance to
  38. 1:23to do work without that
  39. 1:26endogeneity problem coming up without
  40. 1:29your identification strategy being
  41. 1:31criticized in some way. And so, it begs
  42. 1:34a lot of questions, I think, in terms of
  43. 1:37you know, how how how can we do this
  44. 1:39research in corporate finance in a
  45. 1:41meaningful way while still
  46. 1:45dealing with this endogeneity
  47. 1:47endogeneity problem. Now, a lot of
  48. 1:48people complain about how they're
  49. 1:50treated by referees and and editors on
  50. 1:53this issue and and and refer to
  51. 1:56so-called identification police. So, an
  52. 1:58alternative title for this talk might be
  53. 2:01eluding the long arm of identification
  54. 2:03police.
  55. 2:05But, the my idea today is not really to
  56. 2:08to talk about econometrics. There's been
  57. 2:12several very good tutorials at the FMA
  58. 2:14over years on that subject. I'm more
  59. 2:17interested in talking about sort of
  60. 2:20research design issues in empirical
  61. 2:22corporate finance with a particular eye
  62. 2:25towards expanding the types of studies
  63. 2:28that we do cuz I one of the concerns
  64. 2:31that I have is that this identification
  65. 2:34problem has has become so pervasive in
  66. 2:37our work that it's sort of narrowed down
  67. 2:40the the styles of papers that we think
  68. 2:42are acceptable to to journals now. And
  69. 2:45as a result, we're potentially limiting
  70. 2:47the scope of questions that we're asking
  71. 2:50in the field. And I I don't think that
  72. 2:51would be a good development if that's
  73. 2:53the case. So,
  74. 2:55uh what I'm going to do is is is the
  75. 2:57following. I'm just going to talk a
  76. 2:59little bit at at the start about the
  77. 3:01basics of of what the endogeneity
  78. 3:03problem is. I think everyone's pretty
  79. 3:05familiar with that as well as some of
  80. 3:08the econometric responses that that
  81. 3:10exist in the literature. And then I'm
  82. 3:12going to talk about those. Again, this
  83. 3:14isn't a talk about econometrics, but but
  84. 3:17more just to point out that none of
  85. 3:19these are are perfect solutions by any
  86. 3:22stretch. They all have uh
  87. 3:24some limitations to them. We're never
  88. 3:26completely solving the problem.
  89. 3:29Uh
  90. 3:30but what we have done is sort of
  91. 3:31narrowed down the the
  92. 3:34the set of papers I think that that we
  93. 3:36tend to do in empirical corporate
  94. 3:38finance. And that that raised these
  95. 3:40these concerns that I alluded to
  96. 3:43a couple minutes ago.
  97. 3:45And then what I'm going to do is is sort
  98. 3:46of apply some of these ideas in in the
  99. 3:49field of research that I've been doing
  100. 3:51more work on lately, which is capital
  101. 3:53structure research. So, I'm going to to
  102. 3:56talk about alternative different types
  103. 3:58of of of research designs, all of which
  104. 4:02I think are useful in terms of what
  105. 4:04we're really trying to do in the
  106. 4:06profession, which is expand the the set
  107. 4:08of knowledge that we have on a
  108. 4:10particular topic. None of them are
  109. 4:12perfect. Many of them will suffer from
  110. 4:15identification problems, but that
  111. 4:17doesn't mean they they they aren't
  112. 4:19useful. And so, I I I think uh this is
  113. 4:22this is where I'm going to try to
  114. 4:24convince you that we need to to
  115. 4:25re-expand the types of studies that that
  116. 4:28we're doing if we really want to get at
  117. 4:30some of the big questions that that we
  118. 4:32encounter in in corporate finance. And
  119. 4:34And I'll run through several examples
  120. 4:37from from the recent capital structure
  121. 4:39literature that I think
  122. 4:41uh illustrate the points that I'm trying
  123. 4:43to make, both in terms of what we can
  124. 4:45learn from those types of studies as
  125. 4:47well as what we what we can't learn, and
  126. 4:49how
  127. 4:51think about doing studies like that
  128. 4:54um in a way that uh will be viewed as
  129. 4:58useful by the journals and and therefore
  130. 5:00uh publishable in in uh in the
  131. 5:02literature. And then
  132. 5:05draw some conclusions from that.
  133. 5:07Okay, so the basic endogeneity problem I
  134. 5:10think everyone is is pretty familiar
  135. 5:11with. A lot of the research
  136. 5:14that we do in in corporate finance is
  137. 5:16really geared towards trying to
  138. 5:18establish some causal connection between
  139. 5:21some variable X1 in this case and and
  140. 5:24some
  141. 5:25dependent variable um
  142. 5:27that let's list as Y in in this
  143. 5:30particular case. So, you think of it in
  144. 5:32a regression framework, doesn't have to
  145. 5:34be a linear regression like this, but
  146. 5:36it's easier to represent this way.
  147. 5:39Um we have the we have the association
  148. 5:41between Y and X1 while controlling for
  149. 5:44some factor
  150. 5:46uh of of X variables that we think are
  151. 5:49are other possible determinants of Y.
  152. 5:52We're trying to establish whether access
  153. 5:55effects
  154. 5:56on on Y. So,
  155. 5:59common examples in the corporate finance
  156. 6:01literature, you know, does The question
  157. 6:03might be, does debt constrain
  158. 6:06investments on Y? So, we can we can test
  159. 6:09the association between debt levels and
  160. 6:11levels of investment.
  161. 6:13Uh do taxes affect capital structure?
  162. 6:16So, it's the association between some
  163. 6:18measure of the of the of the tax
  164. 6:21obligation that the firm is facing and
  165. 6:23the leverage choice that they make.
  166. 6:25Uh or does does governance affect value
  167. 6:31notable example that has a connection
  168. 6:33between ownership and some measure of
  169. 6:36value like like Tobin's Q.
  170. 6:38All right. So, that's These are typical
  171. 6:40tests, all of which suffer from this
  172. 6:43this identification problem because we
  173. 6:45don't know
  174. 6:47whether there's a causal connection
  175. 6:48between this variable X X sub one and
  176. 6:52the dependent variable or whether we've
  177. 6:53got some correlated variables. So,
  178. 6:56ultimately, the the issue is whether
  179. 6:58there is some some correlation between
  180. 7:00this X1 variable and the error term of
  181. 7:03this model. Now, obviously, if we had
  182. 7:06some idea what these omitted variables
  183. 7:09were, we'd put them in
  184. 7:11and and then we could take care of the
  185. 7:12problem. But, the issue is we don't know
  186. 7:15what they are.
  187. 7:16I mean, and it's very easy for a referee
  188. 7:20or an editor to say, "Well, there's
  189. 7:22something out there
  190. 7:23that's that's correlated both with your
  191. 7:26X1 variable and this dependent variable
  192. 7:29that you're trying to test for."
  193. 7:31And so, again, I just don't buy
  194. 7:33identification strategy. I don't buy
  195. 7:35that there is a a causal connection
  196. 7:37between the two. So, for example,
  197. 7:40you know, if we run if we run
  198. 7:41regressions on investment, on leverage,
  199. 7:44we will find that the higher is the
  200. 7:46leverage, the lower is the level of
  201. 7:48investment.
  202. 7:49Right? And that could be
  203. 7:51because debt constrains investment, or
  204. 7:54it could or it could just be that firms
  205. 7:56that have poor growth opportunities
  206. 7:59uh end up higher leverage. So, we would
  207. 8:02predict they shouldn't be in capital
  208. 8:03structure models.
  209. 8:05Right? And you simply can't perfectly
  210. 8:07control for growth opportunities. So,
  211. 8:09even if we have some proxies for growth
  212. 8:11opportunities on the right-hand side, if
  213. 8:13we can't perfectly control for them, the
  214. 8:16omitted part is is embedded in the error
  215. 8:18term, and we have this positive
  216. 8:19correlation between
  217. 8:21uh debt and and the error term,
  218. 8:24and we're stuck.
  219. 8:26So, it's What do we do about that? So,
  220. 8:28where I think the profession has has
  221. 8:30really made some some big strides in in
  222. 8:33the last 10-15 years is in understanding
  223. 8:36econometrics and in the application of
  224. 8:39these uh econometric methods to problems
  225. 8:42associated with with identification.
  226. 8:46Right? So,
  227. 8:47it's you know, it's fairly standard now
  228. 8:49that you see panel data techniques used,
  229. 8:52firm fixed effects, industry fixed
  230. 8:54effects, year fixed effects.
  231. 8:57Those are useful to a point. I mean,
  232. 9:00you've got some omitted variable that is
  233. 9:03firm-specific
  234. 9:05and doesn't vary through time, firm
  235. 9:07fixed effects are taken care of.
  236. 9:09All right, well, good for you if that's
  237. 9:11the case. But, the problem is we don't
  238. 9:13know that that's that's the nature of
  239. 9:15the omitted variable. So, if there's
  240. 9:17some firm-specific
  241. 9:19but time-varying
  242. 9:21variable that is that is the source of
  243. 9:23this endogeneity,
  244. 9:25the panel data techniques aren't aren't
  245. 9:27really going to do it for us. And And
  246. 9:29again, unless we really know what the
  247. 9:31omitted variable is, which we can't by
  248. 9:33definition, uh it it's it we're still
  249. 9:37going to subject at times to this this
  250. 9:39criticism. Now, what panel data
  251. 9:42techniques are
  252. 9:43useful for is that they're at least
  253. 9:45limiting the scope
  254. 9:47of the endogeneity problem.
  255. 9:49This is you can't be criticized
  256. 9:52for omitting something that is just some
  257. 9:54factor that isn't going to vary through
  258. 9:56time within a firm. You have to control
  259. 9:58for that. So, it's a bit of a solution
  260. 10:00in that sense, but it can never
  261. 10:03completely take care of the problem
  262. 10:05unless you really know precisely the the
  263. 10:07nature of the problem.
  264. 10:09All right. Another possibility is some
  265. 10:11sort of regression discontinuity
  266. 10:13technique.
  267. 10:14So, here here the idea would be that
  268. 10:18we have some sort of threshold event
  269. 10:21and when we cross this threshold, we can
  270. 10:23test whether there's a difference in
  271. 10:25behavior or something. So, an example
  272. 10:26would be uh if you're if you're trying
  273. 10:29to test
  274. 10:30uh whether creditor control has some
  275. 10:32impact on real investment decisions.
  276. 10:36Let's say something like Chava and
  277. 10:37Roberts in
  278. 10:38in their paper. So, you could you can
  279. 10:41observe a covenant violation.
  280. 10:43All right. And the question is when you
  281. 10:45cross this threshold and violate the
  282. 10:47covenant,
  283. 10:48all right, do you observe a difference
  284. 10:50in behavior? Now, the the nice advantage
  285. 10:52of the regression discontinuity
  286. 10:54technique is that right around that
  287. 10:56threshold,
  288. 10:57uh you can plausibly say that firms that
  289. 11:00just on either side of that threshold
  290. 11:03are roughly the same.
  291. 11:05And their their characteristics are are
  292. 11:07roughly the same. And if you can make
  293. 11:08that claim, all right, then this is a
  294. 11:10pretty solid way to try to deal with
  295. 11:13with the endogeneity criticism.
  296. 11:15Limitation here is that
  297. 11:17you don't come across that many clean
  298. 11:20threshold types of events that you can
  299. 11:22really use in in the context that that
  300. 11:25we study quite a lot.
  301. 11:26If you can find them, it's great.
  302. 11:29By all means, use them.
  303. 11:31Um but
  304. 11:34my experience is are difficult to find
  305. 11:37and therefore
  306. 11:38not often do you come across
  307. 11:41situations where that's really going to
  308. 11:42help you out with your identification
  309. 11:45problem.
  310. 11:46Third possibility is to surround to some
  311. 11:48sort of matching models and propensity
  312. 11:50score type of approach. And again,
  313. 11:53that's great if you can if you can
  314. 11:55identify what you think are are the most
  315. 11:58plausible set of factors that you can
  316. 12:00match up firms and do a difference in
  317. 12:02difference sort of approach. But again,
  318. 12:05almost by definition, these can't be
  319. 12:07perfect because we're talking about
  320. 12:09correlated omitted variables that we
  321. 12:11don't really know what they are.
  322. 12:13If we knew what they are, sure we could
  323. 12:14perfectly match on them and and and run
  324. 12:17this type of test, but not knowing what
  325. 12:20they are leaves us a little bit in
  326. 12:22limbo.
  327. 12:23Uh instrumental variables, not another
  328. 12:26possibility. Again, has some
  329. 12:28limitations. We need some some variable
  330. 12:30that's going to satisfy both the
  331. 12:32relevance criteria and the exclusion
  332. 12:36criteria.
  333. 12:37Relevance is pretty easy to demonstrate
  334. 12:39most of the time with with an IV, but
  335. 12:42it's also it's really easy to criticize
  336. 12:44instrumental variables on on the
  337. 12:47on the exclusion part. And so,
  338. 12:50uh they're difficult to find. Good
  339. 12:51instruments are really difficult to come
  340. 12:53across. They've got to be based on on
  341. 12:56underlying economics, but you know,
  342. 12:58there's no great test for
  343. 13:01for that instrument, and so you're
  344. 13:03always open to this criticism that
  345. 13:05you just got that instrument.
  346. 13:07And so,
  347. 13:09all of these techniques are useful.
  348. 13:11All of them should be done
  349. 13:13in different contexts, but we we can
  350. 13:16never use these techniques and think,
  351. 13:18"Okay, we solved
  352. 13:20the endogeneity problem." We don't We
  353. 13:22don't We don't solve it using
  354. 13:24econometrics.
  355. 13:25So So, where the profession is of headed
  356. 13:28more towards is is the fifth one, which
  357. 13:31is is natural experiments.
  358. 13:34All right. What's great about natural
  359. 13:35experiments is that you can identify
  360. 13:38some plausibly exogenous event that's
  361. 13:41truly exogenous, and you can do sort of
  362. 13:45a different diff sort of a different
  363. 13:47diff analysis and and claim causality in
  364. 13:51a pretty credible fashion.
  365. 13:53All right. And because you can do that
  366. 13:56and defend it much more easily than some
  367. 13:58of the others, I this has become sort of
  368. 14:01the holy grail of science in terms of
  369. 14:03doing empirical corporate finance. Uh
  370. 14:07but it it's not without problems, I
  371. 14:09think. And it's the the these are the
  372. 14:10concerns
  373. 14:12that I was alluding to before. I think
  374. 14:14it it I think the way which we started
  375. 14:16to head
  376. 14:18is is to have this general belief that
  377. 14:20every study's got to have some clean
  378. 14:22natural experiment or some some perfect
  379. 14:25instrument. Otherwise, the study's doing
  380. 14:28you know, just a quick check on SSRN
  381. 14:30that I did, you know, just
  382. 14:32putting natural experiment in the
  383. 14:34abstract of papers, I get over 1,600
  384. 14:37entries of of recent papers
  385. 14:40uh in the field, mostly in corporate
  386. 14:42finance where where you see that.
  387. 14:44Now, I want to
  388. 14:46make sure I'm clear that I I'm not at
  389. 14:48all criticizing natural experiments. I
  390. 14:51mean, they're they're great if you've
  391. 14:53got them and and they and they fit the
  392. 14:55question you're asking, but my concern
  393. 14:57is sort of twofold is that what you tend
  394. 15:00to see a little bit more often now than
  395. 15:03you than than you saw before is that
  396. 15:05authors are kind of starting with the
  397. 15:07experiment
  398. 15:08rather than with the question.
  399. 15:11That's not
  400. 15:12quite the way you want to you want to go
  401. 15:14about doing research. I'll say more
  402. 15:15about that in in a minute. In a way, it
  403. 15:18it it's kind of similar to
  404. 15:21uh concerns that have come up over the
  405. 15:23years as sources of data have changed.
  406. 15:27You know, back in the dark ages when I
  407. 15:29was getting my degree,
  408. 15:31there wasn't a whole lot of machine
  409. 15:34readable data.
  410. 15:35And now,
  411. 15:36most data is is machine readable.
  412. 15:39Um, and as a result, you saw over time,
  413. 15:43at least my perception is that you saw
  414. 15:45authors start to shy away from data
  415. 15:48collection data. They would say, "Well,
  416. 15:49they would do a study unless they could
  417. 15:51identify a readily readily available
  418. 15:54data source." And then you even to the
  419. 15:55point where
  420. 15:57uh, they're just be looking for data all
  421. 15:59the time, see the data in in good
  422. 16:01machine readable form, and say, "Well,
  423. 16:03let me figure out something to do
  424. 16:05with this data."
  425. 16:06Again, that's that's not really the way
  426. 16:08you want to conduct research. You want
  427. 16:09to you want to start um,
  428. 16:12with the question.
  429. 16:14Um, so it it alluded to sometimes as,
  430. 16:17you know, are
  431. 16:18are we reaching a point where we
  432. 16:21unnatural obsession with natural
  433. 16:23experiments. I think we've got to have
  434. 16:24these things all the time. And so that's
  435. 16:27one concern
  436. 16:29that is sort of changing the way that
  437. 16:30people go about doing research. A
  438. 16:32related concern though, which I think is
  439. 16:34the bigger one in terms of of the field
  440. 16:37or the profession in general, is that
  441. 16:40it can be very limiting in terms of the
  442. 16:42scope for the questions that you ask.
  443. 16:45And there's some big questions
  444. 16:47in our field like that should be
  445. 16:50addressed.
  446. 16:51And when these big questions come up,
  447. 16:54maybe there isn't a perfect natural
  448. 16:57experiment. Maybe there isn't a perfect
  449. 16:59instrumental variable that you can use.
  450. 17:01And if people think, "Well, unless I
  451. 17:03have that, I can't do this study," then
  452. 17:05those studies don't get done. But we're
  453. 17:08giving up a lot if that's the way we we
  454. 17:10approach research, I think, in the sense
  455. 17:12of
  456. 17:13really important questions that we could
  457. 17:15shed some light on, and just
  458. 17:17incrementally move towards a richer
  459. 17:20knowledge of somebody's questions. I'm
  460. 17:23going to talk about the capital
  461. 17:24structure research, and I'm going to try
  462. 17:27to
  463. 17:28highlight a little bit more what I what
  464. 17:29I mean by that.
  465. 17:31All right. So, all right. So, let me
  466. 17:33compare those concerns
  467. 17:35with what I think most of us in the room
  468. 17:38would would say is this how we would
  469. 17:40advise PhD students as to how to go
  470. 17:43about doing research. Right? First and
  471. 17:46foremost, say, "Well, you want to
  472. 17:47identify an interesting question."
  473. 17:50When students come and talk to me, it's
  474. 17:53frequently the first thing I'm saying,
  475. 17:54"Well, what's the question that you're
  476. 17:56asking people to follow?" And And I've
  477. 17:58noticed this even in terms of uh
  478. 18:02interviews uh at the meetings FMA or FA
  479. 18:05meetings or interviewing candidates, and
  480. 18:08you ask the candidate to
  481. 18:10launch into their spiel about about
  482. 18:12their research, and the first thing they
  483. 18:15do is they go straight to their
  484. 18:16experiment. Like, this is the experiment
  485. 18:18I'm I'm running.
  486. 18:20That's not really what most people want
  487. 18:22to hear. They don't want to They don't
  488. 18:23want to know what's what's your
  489. 18:24question. What's the question you're
  490. 18:26really trying to get at? And And this is
  491. 18:29the way I think, as I said, most of us
  492. 18:31would would probably advise students to
  493. 18:33do research, but first identifying an
  494. 18:37interesting question.
  495. 18:38All right. Once you've got that, all
  496. 18:41right, then you should be thinking about
  497. 18:43developing hypotheses from first
  498. 18:45principles, from the ground up.
  499. 18:48And develop a plausible, testable set of
  500. 18:51hypotheses.
  501. 18:53And then, and only then, I think, when
  502. 18:55I've gotten to reach that point, then
  503. 18:57you start thinking about the
  504. 18:59experimental design. Well, what's the
  505. 19:00best way
  506. 19:02that I could go about trying to shed
  507. 19:04some light on on this particular set of
  508. 19:07hypotheses?
  509. 19:09All right. And this This is the part,
  510. 19:10you know, just to reiterate what I said
  511. 19:12a couple minutes ago, like this is the
  512. 19:14part that that at times looks like it's
  513. 19:16getting split
  514. 19:17in in the profession. It's like too many
  515. 19:19studies
  516. 19:21are starting with the experiment
  517. 19:24uh and then uh and then moving on
  518. 19:27uh
  519. 19:28from there. All right, so once we
  520. 19:29identify this experimental design that
  521. 19:32we think is the best for say ISM and
  522. 19:34line on, you know, then we then we
  523. 19:35eventually get some results.
  524. 19:38We got to let the data speak, but we got
  525. 19:39to let the data speak in in in a way
  526. 19:42that we're well aware of what the
  527. 19:44limitations of what we're doing
  528. 19:46and this is
  529. 19:48I guess this is more of a criticism of
  530. 19:51uh referees than it is of authors at
  531. 19:53this point is that if you you tend to
  532. 19:55see a lot of referees
  533. 19:57uh
  534. 19:58who are unwilling to accept limitations.
  535. 20:01Uh say, "Well, if you don't have a
  536. 20:03perfect identification strategy, I'm
  537. 20:05going to reject your paper." That's
  538. 20:07already leading people to think, "Well,
  539. 20:09I got to have a perfect identification
  540. 20:11strategy, otherwise this this paper is
  541. 20:12not worth doing." I'm not sure that's
  542. 20:15the best
  543. 20:16uh development for our profession
  544. 20:18because there are a lot of things that
  545. 20:19we can learn from studies that are
  546. 20:22imperfect.
  547. 20:23All right, because let's be honest,
  548. 20:25every study is imperfect
  549. 20:28in some way.
  550. 20:29Uh and if the particular way is you
  551. 20:32can't necessarily identify causation in
  552. 20:35your study, it doesn't necessarily mean
  553. 20:38this that the study isn't worth doing,
  554. 20:40but it does mean that if you do the
  555. 20:42study and you fail to note those
  556. 20:44limitations, you probably deserve to be
  557. 20:46rejected at this point, right?
  558. 20:48You can you can provide evidence, you
  559. 20:51can interpret that evidence, you can
  560. 20:53understand that maybe you can't
  561. 20:56uh imply causation from that set of
  562. 20:58evidence. If all that is kind of laid
  563. 21:00out for the reader, then the reader can
  564. 21:02decide in the end, well, given all these
  565. 21:05limitations, do I feel like I'm learned
  566. 21:07something from from reading the study?
  567. 21:09And
  568. 21:09the answer is yes. And you learned
  569. 21:12enough,
  570. 21:13then the journal should be willing to
  571. 21:15accept that paper, and they do.
  572. 21:17And they do, and that's
  573. 21:19that's important.
  574. 21:21Um
  575. 21:23Let me let me just back up a second.
  576. 21:24There's one one more
  577. 21:27issue that I was I wanted to note on
  578. 21:28this particular slide.
  579. 21:31I'm using the example of business
  580. 21:33combination laws. Now, so so this is a
  581. 21:36natural experiment I think that has had
  582. 21:37a lot of success
  583. 21:39in the profession. And so let me be
  584. 21:41clear up front, I'm not criticizing this
  585. 21:44this natural experiment. It's a good
  586. 21:45one.
  587. 21:46The way it's useful
  588. 21:48is is in the sense that uh
  589. 21:51if you think about types of studies in
  590. 21:53corporate finance, one big class of
  591. 21:55studies is asking whether governance has
  592. 21:59some causal impact on value or or
  593. 22:03managerial action. You know, these are
  594. 22:06interesting and big questions, I think,
  595. 22:09but they're tough to get at
  596. 22:10from an identification standpoint. So,
  597. 22:14a clever natural experiment
  598. 22:17is the introduction of these of these
  599. 22:19business combination laws. And they they
  600. 22:21these laws are
  601. 22:23are sort of staged over time in in
  602. 22:26different states. So, it makes for a
  603. 22:27really nice exogenous shock to
  604. 22:31governance that we can then connect with
  605. 22:33some sort of outcome variable.
  606. 22:35And so, it's got a lot of the attributes
  607. 22:38that you're looking for in natural
  608. 22:40experiments. Now, it's also got
  609. 22:42it also raises some of the concerns that
  610. 22:44you always see in natural experiments as
  611. 22:47well, because none of these what we call
  612. 22:49natural experiments in the in the social
  613. 22:51science are never really pure natural
  614. 22:54experiments, because somebody had to
  615. 22:56decide
  616. 22:57that we're going to change these laws.
  617. 22:59And the laws are not purely exogenous,
  618. 23:01they're endogenous in some way. And
  619. 23:04maybe what's driving the change in the
  620. 23:06laws is correlated with your outcome
  621. 23:08variable. But these are authors who have
  622. 23:10done these these these studies of
  623. 23:11business combination laws have been well
  624. 23:13aware of those limitations and made a
  625. 23:16plausible case that uh they're they're
  626. 23:19not interfering with the
  627. 23:21interpretation of causation. Right? So
  628. 23:24it's So it's a pretty nice
  629. 23:26clean natural experiment. Um
  630. 23:29but it's probably one that I've seen
  631. 23:31repeated the most times of any sort of
  632. 23:34natural experiments. It's been repeated
  633. 23:36in some good ways.
  634. 23:38Right? But you start It starts to run
  635. 23:40out of juice after a while because, you
  636. 23:43know, you see some some big questions
  637. 23:44like does governance affect value? Does
  638. 23:46governance affect investment? Does
  639. 23:48governance affect innovation? These are
  640. 23:51all
  641. 23:52very useful applications of this natural
  642. 23:54experiment. But But what tends to happen
  643. 23:56in our profession
  644. 23:58is that everyone says, "Well, you need a
  645. 24:01clean natural experiment.
  646. 24:03I can see from these publications that
  647. 24:05the profession seems to accept this one
  648. 24:08as as a useful natural experiment. So
  649. 24:11look and I did with this one." Again,
  650. 24:13you're starting with the experiment then
  651. 24:16rather than the question and we start to
  652. 24:18get into some really marginal questions
  653. 24:20at this point, you know? You just keep
  654. 24:22following this path and you eventually
  655. 24:24start seeing studies that are like, you
  656. 24:26know, "What's What's the impact of
  657. 24:29of columns of two business combination
  658. 24:31laws and what the CEO had for lunch on
  659. 24:33Tuesday?" Well, yeah, I guess you can
  660. 24:36establish causation with this natural
  661. 24:38experiment, but now you're asking a
  662. 24:39question nobody cares about.
  663. 24:42So you have to worry about that a lot
  664. 24:44and you run into this problem precisely
  665. 24:46because you're starting with the
  666. 24:49experiment that you think will pass
  667. 24:51muster with journals instead of
  668. 24:53starting with the with what is the
  669. 24:56interesting and and and big question.
  670. 24:58All right, so now let me
  671. 25:00talk a bit about capital structure
  672. 25:03research as as an example of some of
  673. 25:05these these themes that that I'm talking
  674. 25:07about. So,
  675. 25:09um there's nothing particularly unique
  676. 25:12about capital structure research with
  677. 25:14respect to these identification
  678. 25:16problems. I'm choosing simply because
  679. 25:18I've been doing papers in this area, so
  680. 25:20I know this area a little bit more and
  681. 25:22you'll see in some of the examples I'm
  682. 25:24using in my own papers.
  683. 25:26It's not because I think these are the
  684. 25:28best papers in in that field, but I know
  685. 25:30these papers better and then they're
  686. 25:32representative of some of the points
  687. 25:34that that that I'm trying to make. But,
  688. 25:37there's some really fundamental
  689. 25:38questions in in capital structure
  690. 25:41research that amazingly enough after all
  691. 25:44these years we haven't done great job of
  692. 25:46answering them quite yet.
  693. 25:48We we don't really have
  694. 25:51a great idea of what are the primary
  695. 25:53determinants of
  696. 25:54of capital structure decisions. We have
  697. 25:56some pretty good theory
  698. 25:58about what sort of market frictions
  699. 26:00often matters,
  700. 26:02taxes, bankruptcy costs, agency costs,
  701. 26:06asymmetric information,
  702. 26:08uh but it's it's difficult. It's it's
  703. 26:11proven difficult to establish what
  704. 26:13factors are really first-order
  705. 26:14determinants of of capital structure
  706. 26:17decisions and part of the issue has
  707. 26:18always been
  708. 26:20uh proper identification. Now, another
  709. 26:22whole segment of this literature is this
  710. 26:24sort of hearing on static models
  711. 26:27versus dynamic models. I'm not really
  712. 26:29going to talk that much about that cuz
  713. 26:30it's not
  714. 26:31as directly relevant to to to topic
  715. 26:34today, but that's also a challenging
  716. 26:36aspect in this literature. Um
  717. 26:40but what you see and what what I want to
  718. 26:42talk about in in the rest of the time
  719. 26:45are a variety of different sort of
  720. 26:47research approaches in this field that I
  721. 26:50would argue are all very useful. They
  722. 26:53they all
  723. 26:54give us some information different
  724. 26:56aspects of the of the set of of the set
  725. 26:59of information that we need to
  726. 27:00ultimately draw some conclusions about
  727. 27:03about capital structure.
  728. 27:05Uh and only one of them
  729. 27:07falls in this category of of natural
  730. 27:10experiments. But
  731. 27:11what's important, I think, to to note is
  732. 27:14that even though all these are
  733. 27:16potentially useful,
  734. 27:18they all have different limitations,
  735. 27:20too.
  736. 27:21So, you're you're getting different
  737. 27:22pieces of information from them. So, you
  738. 27:26you can't view them as as
  739. 27:28being able to draw the same sort of
  740. 27:30inferences in terms of causation.
  741. 27:33Uh but you can view them sometimes as as
  742. 27:35telling you a little bit more about
  743. 27:37what's first order and what's not first
  744. 27:39order. And I'll I'll I'll be a little
  745. 27:40bit more precise about that in in a
  746. 27:42couple of minutes. So,
  747. 27:44So, the classic approach, of course, is
  748. 27:46just doing OLS type of regression.
  749. 27:48You've got leverage on off the hand
  750. 27:50side. You've got a set of your testable
  751. 27:53determinants on the right hand side and
  752. 27:55other control variables. Uh you might
  753. 27:57also
  754. 27:58estimate this using using panel
  755. 28:00regression techniques uh to try to
  756. 28:03control for fixed effects of some sort.
  757. 28:06And the literature has got tons of
  758. 28:08studies that that fit into this
  759. 28:11particular category. But there's proven
  760. 28:12to be two main problems. The main one is
  761. 28:15is the identification problem that we're
  762. 28:17we're primarily focused on. But as a
  763. 28:20practical matter, too, the second
  764. 28:21problem has been that it hasn't proven
  765. 28:24that fruitful. You know, we we we end up
  766. 28:26identifying some relatively vague
  767. 28:29factors that that seem to matter. You
  768. 28:32can say that leverage is associated with
  769. 28:34profitability,
  770. 28:36growth opportunities as measured by
  771. 28:38market-to-book ratios, usually. Uh firm
  772. 28:40size.
  773. 28:42That's great in a way, but when we think
  774. 28:44about relating that to capital structure
  775. 28:47theories, we find ourselves in this trap
  776. 28:50where it's consistent with multiple
  777. 28:52theories,
  778. 28:53uh even multiple classes of theories
  779. 28:56between static model, trade-off models,
  780. 28:58and pecking order models. They can be
  781. 29:00reconciled with this this set of
  782. 29:04observed determinants. And so, in the
  783. 29:06end,
  784. 29:07you feel like you don't learn that much
  785. 29:09at times with with with this literature.
  786. 29:12All right. So,
  787. 29:13the way the literature is progressed, I
  788. 29:15think, and this is this is a useful
  789. 29:17progression, is is to think about
  790. 29:19different types of empirical approaches
  791. 29:22that can shed some additional light.
  792. 29:24Now, one one type would be a natural
  793. 29:27experiment.
  794. 29:28All right. Can we Can we observe some
  795. 29:30exogenous shock to one of these
  796. 29:32hypothesized determinants while the
  797. 29:35others are held constant
  798. 29:37and see whether that shock seems to
  799. 29:39elicit a reaction from firms in in
  800. 29:42choice of capital structure.
  801. 29:44All right. That's one way to shed some
  802. 29:46light
  803. 29:47on
  804. 29:48a specific fact.
  805. 29:50All right. Another possibility
  806. 29:52is to conduct what's really more of
  807. 29:54descriptive data analysis type of of
  808. 29:57studies. Now,
  809. 29:59these next two, descriptive data and
  810. 30:01longitudinal types of studies, are are
  811. 30:03kind of similar in the sense that you
  812. 30:06would probably characterize them both,
  813. 30:08well, one certainly by the name, but
  814. 30:10both of them are are sort of
  815. 30:11descriptive.
  816. 30:14All right. And by that, I mean they're
  817. 30:16not generally set up as being specific
  818. 30:20hypotheses that are tested in sort of a
  819. 30:22traditional scientific way. All right.
  820. 30:25And and sometimes these kind of studies
  821. 30:27get a bad name. You see people use
  822. 30:29descriptive in a as a pejorative term.
  823. 30:33Oh, that study is just descriptive. Um
  824. 30:35but I would argue that this has a place
  825. 30:40in in the set of research approaches,
  826. 30:43uh and in particular in capital
  827. 30:45structure, I think it's been useful. I'm
  828. 30:46going to
  829. 30:47I'll give you a couple of examples of of
  830. 30:50how it they have been useful.
  831. 30:51Longitudinal studies are are similar in
  832. 30:54the sense of they they look over a long
  833. 30:56period of time whether capital structure
  834. 30:59has changed over that period of time and
  835. 31:01whether we can somehow connect that with
  836. 31:04changes in either firm or macroeconomic
  837. 31:07characteristics in a way that is
  838. 31:09consistent, you know, inconsistent with
  839. 31:11with what we think of as as the theory
  840. 31:14underlying it. So,
  841. 31:15the longitudinal models are are a little
  842. 31:18bit like natural experiments in that you
  843. 31:21you can have over a long period of time
  844. 31:23shocks to multiple
  845. 31:25possible determinants of capital
  846. 31:27structure. You can sort of think of it
  847. 31:29as sort of a horse race among these
  848. 31:31different factors as to what seems to be
  849. 31:33first-order and and and what's not.
  850. 31:36Another approach might be structural
  851. 31:38models,
  852. 31:39right, where we really try to get into
  853. 31:41the dynamics of of of
  854. 31:44capital structure policy over time. Can
  855. 31:47we specify some sort of objective
  856. 31:50function that the manager is maximizing,
  857. 31:53specify a set of exogenous determinants
  858. 31:56as well as a set of endogenous variables
  859. 31:58that be connected with uh with leverage.
  860. 32:01And so, in a sense, specify the nature
  861. 32:04of the endogeneity problem and then take
  862. 32:07this to the data through some structural
  863. 32:09estimation to see whether the data seem
  864. 32:11consistent with that uh with the
  865. 32:13predictions from that structural model.
  866. 32:16Lots of good examples of that in the
  867. 32:18recent literature as well.
  868. 32:20All right. And last and and certainly
  869. 32:22not least is that I think there's room
  870. 32:25in this literature for clinical or case
  871. 32:28study sort of approaches. Now, it's
  872. 32:30difficult, I think, to do a standalone
  873. 32:34clinical study.
  874. 32:35Uh I mean, years ago the JFE published
  875. 32:38lots of those. It's sort of backed away
  876. 32:40from that over time, but
  877. 32:43but I think clinical studies are very
  878. 32:45useful in conjunction with with some of
  879. 32:48the others. Now, I'll give you some
  880. 32:49examples of that how you might combine
  881. 32:52survey descriptive data analysis or
  882. 32:54longitudinal study with with uh clinical
  883. 32:57analysis of the subset of your data in a
  884. 32:59way that sheds a lot of light on what's
  885. 33:02on what's really going on. It really
  886. 33:04helps you tease out the identification
  887. 33:07issues as well.
  888. 33:09And so so let me give you
  889. 33:12a few different examples
  890. 33:14from the literature that that highlight
  891. 33:16the use of these different approaches.
  892. 33:18And good, so I want I want to ultimately
  893. 33:20make two points. One is that each of
  894. 33:21them are useful in shedding light on an
  895. 33:25aspect of of the capital structure
  896. 33:27problem.
  897. 33:29And then secondly, I'll make sure we
  898. 33:31understand that each of them also has
  899. 33:33limitations. Okay, so you have to be
  900. 33:35careful when you do these studies just
  901. 33:37to understand what you can and cannot
  902. 33:40conclude from the type of study that
  903. 33:42you've chosen to do.
  904. 33:44All right, so one approach as we
  905. 33:45mentioned is is a natural experiment.
  906. 33:48A good example of that are these studies
  907. 33:51that have that have looked at staggered
  908. 33:54changes in the income tax rates either
  909. 33:57across states within the US like the
  910. 33:59Hadlock paper or
  911. 34:01across different countries throughout
  912. 34:03the world like Lacho and and Shue at al.
  913. 34:07So it's sort of a classic natural
  914. 34:10experiment diff-in-diff kind of approach
  915. 34:12where you've got the change in in data,
  916. 34:14change in leverage on the left-hand side
  917. 34:18as a function of change in the tax rate,
  918. 34:20that's the T variable, set of
  919. 34:23firm-specific variables that denote you
  920. 34:26with with X and I's.
  921. 34:29Uh and and uh industry-specific
  922. 34:32variables as well.
  923. 34:34So, you're trying to control for all
  924. 34:35these other possible determinants of of
  925. 34:38leverage or CEO shock to tax rates at
  926. 34:42point in time,
  927. 34:44do you see firms respond
  928. 34:46uh with changes in leverage? All right.
  929. 34:48Now,
  930. 34:50as with most natural experiments, as I
  931. 34:52mentioned before, they're not pure
  932. 34:54natural experiments. So, there's there's
  933. 34:56still a little bit of endogeneity that
  934. 34:59you worry about in this. So,
  935. 35:01you might worry that somehow there's
  936. 35:03still some systematic differences that
  937. 35:05you can't control for between firms,
  938. 35:08say, within states that had a tax rate
  939. 35:10change versus those that didn't. So,
  940. 35:13you'll see clever clever ways in to try
  941. 35:16to get at that by looking right around
  942. 35:18the borders between states, look at
  943. 35:20counties that are right along the
  944. 35:21border, with the idea being that the
  945. 35:23economic conditions of of companies on
  946. 35:26either side of the border have got to be
  947. 35:28roughly
  948. 35:29identical, or borders of countries in
  949. 35:32the case of of of Argentina, issue.
  950. 35:35And so, you you So, you do your best
  951. 35:38to hold constant all the other possible
  952. 35:41economic determinants of this leverage,
  953. 35:43but one group has the shock tax rates,
  954. 35:46the other group did not have that shock
  955. 35:47tax rates, do we see a difference
  956. 35:50uh in leverage as a result of that of
  957. 35:53that shock to leverage. So, if done
  958. 35:57in in a
  959. 35:59in a correct manner, this type of study
  960. 36:01is really useful for getting at this
  961. 36:03causal connection between this factor,
  962. 36:06taxes,
  963. 36:07and capital structure change.
  964. 36:11What it can do,
  965. 36:13right, is tell you something about
  966. 36:14whether taxes are really a first-order
  967. 36:16consideration or not, because
  968. 36:19you by definition, if you've done study
  969. 36:21right, you've held all these other
  970. 36:23factors constant.
  971. 36:25And you've isolated on on tax rate. Now,
  972. 36:28and that's not a criticism of this type
  973. 36:30of study.
  974. 36:31I it's just
  975. 36:33a point that
  976. 36:35in doing this sort of study, you got to
  977. 36:36be clear on what it is your goal is.
  978. 36:39Like if you're going in this study
  979. 36:41saying, "Well, my question is I want to
  980. 36:43identify what determines capital
  981. 36:45structure." This isn't really doing
  982. 36:47that.
  983. 36:48All right.
  984. 36:50If your question though is, "Do taxes
  985. 36:53have an impact on capital structure at
  986. 36:55the margin?" Yeah, this is the way you
  987. 36:57want to do it. You you're able to
  988. 36:59isolate in on the causal impact of taxes
  989. 37:03on leverage decisions.
  990. 37:05It's very useful in that
  991. 37:07But it's a subset of the information
  992. 37:09that we're trying to put together in
  993. 37:11order to get a bigger picture
  994. 37:13of what's really happening with
  995. 37:16corporate capital structure decisions.
  996. 37:18All right, so that that's one
  997. 37:20one type of study.
  998. 37:22All right, so the other one of the other
  999. 37:23types of studies I talked about is is
  1000. 37:25descriptive data analysis. Now,
  1001. 37:28in some respects, this is
  1002. 37:31a tricky way to go about doing empirical
  1003. 37:35research because as I said before,
  1004. 37:37you're not necessarily starting with
  1005. 37:39very firm hypotheses, but rather it's a
  1006. 37:42little bit more exploratory in nature.
  1007. 37:44So, the goal ultimately in a study like
  1008. 37:48this is to provide enough information to
  1009. 37:51the reader that they that they somehow
  1010. 37:53will think differently about capital
  1011. 37:55structure than they did before. All
  1012. 37:57right, if they don't feel that way,
  1013. 38:00then you shouldn't have anything. You
  1014. 38:01just described a bunch of data that
  1015. 38:03everyone shrugs their shoulders and
  1016. 38:05says, "Well, so what?" Okay.
  1017. 38:08Um
  1018. 38:09so, let me I'm going to give you a
  1019. 38:11couple of examples of this type of work,
  1020. 38:13one of which is by Donald Stiefel and
  1021. 38:15Ian Dew-Becker in the RFS a couple of
  1022. 38:16years ago.
  1023. 38:18And so, here the the approach we took in
  1024. 38:20this in this study
  1025. 38:21is is to say, "Oh, let's let's get away
  1026. 38:24from thinking about it in in sort of a
  1027. 38:27standard regression framework where
  1028. 38:28we're looking at a bunch of right-hand
  1029. 38:30side variables determining what's on the
  1030. 38:32left-hand side, which is leverage."
  1031. 38:34Let's Let's instead let's start with the
  1032. 38:37left-hand side.
  1033. 38:39All right. Well, so let's try to observe
  1034. 38:42some major discontinuity in the
  1035. 38:44financing behavior of the firm and then
  1036. 38:46back out what seems to be driving that
  1037. 38:50that change in financing behavior. So,
  1038. 38:52the way we went about doing this is to
  1039. 38:54say, "Okay, well, let's let's first
  1040. 38:55identify firms that have what we call
  1041. 38:58proactive change in leverage." And by
  1042. 39:01proactive, we simply mean that they
  1043. 39:03actually issued debt in order to change
  1044. 39:06their leverage as opposed to they bought
  1045. 39:08back shares or had some other action
  1046. 39:10that just ended up causing their their
  1047. 39:12leverage to be different. We want them
  1048. 39:14to be proactively issuing debt in a way
  1049. 39:17that changes their leverage ratio. And
  1050. 39:19the second aspect of that was that we
  1051. 39:22were requiring that their resulting
  1052. 39:24leverage ratio
  1053. 39:26was at least 10% above some estimate of
  1054. 39:29their target, where the target has been
  1055. 39:31estimated using sort of standard
  1056. 39:34empirical models of of of
  1057. 39:36capital structure.
  1058. 39:38All right. So, we're putting in this
  1059. 39:40second requirement because
  1060. 39:43we thought, "Well, this is this is sort
  1061. 39:44of a a unique and challenging situation
  1062. 39:47for the literature in the sense that
  1063. 39:49we've now got firms that are
  1064. 39:51deliberately taking an action that
  1065. 39:53pushes them well away from what we think
  1066. 39:56of as their target leverage ratio."
  1067. 39:59All right. And so, we're just asking two
  1068. 40:01very simple questions subsequently. Why
  1069. 40:04did they do it?
  1070. 40:05All right. What looks like is the
  1071. 40:07underlying motivation
  1072. 40:09for undertaking this leverage change?
  1073. 40:11And then secondly, how do we see the
  1074. 40:13leverage ratio evolve subsequently?
  1075. 40:17Does that tell us something about
  1076. 40:18whether they they seem to be targeting a
  1077. 40:21particular
  1078. 40:23leverage ratio. All right, so
  1079. 40:25So, what do we find? So, well, first of
  1080. 40:27all, what we tend to observe is that
  1081. 40:29they seem to be doing this
  1082. 40:32uh for reasons of of financing specific
  1083. 40:36investment needs. Sometimes it's capital
  1084. 40:38expenditures or R&D. Sometimes it's it's
  1085. 40:41more working capital considerations,
  1086. 40:43right? But, these don't appear
  1087. 40:45that they're borrowing
  1088. 40:47in order to say, for example, buy back
  1089. 40:49shares.
  1090. 40:50And it also doesn't appear that they're
  1091. 40:53borrowing so that they can move to some
  1092. 40:54new target leverage ratio.
  1093. 40:57All right, but also the second main
  1094. 40:59result is that when you look at what
  1095. 41:01happens after this initial jump in
  1096. 41:03leverage,
  1097. 41:05like it it looks like they're starting
  1098. 41:07to move back towards what we think is
  1099. 41:09the target leverage ratio,
  1100. 41:11but they're in no hurry at all to do it.
  1101. 41:14All right, it's a really slow adjustment
  1102. 41:16back uh
  1103. 41:18uh towards the target, and they even
  1104. 41:21though they have opportunities to be a
  1105. 41:23little bit more proactive about it, and
  1106. 41:25they could have moved back to the target
  1107. 41:27quicker in some cases, they don't.
  1108. 41:30And you you sort of see it drifting back
  1109. 41:33down. And what's kind of interesting too
  1110. 41:34is that when we see subsequent
  1111. 41:36situations where they seem like they
  1112. 41:38have a need for funds, they go out and
  1113. 41:40borrow again. Even though they're still
  1114. 41:42well above their target, they get even
  1115. 41:44further above their target. They don't
  1116. 41:46know.
  1117. 41:47All right, so
  1118. 41:49what does all that mean? How is how is
  1119. 41:51this then a useful study in capital
  1120. 41:54cycle? Some of you might think it isn't
  1121. 41:56useful study in the capital cycle. Why
  1122. 41:58do we think it's it's it's a useful uh
  1123. 42:01study to do? Well, well, first of all,
  1124. 42:03if you if you think about it,
  1125. 42:05what this seems to imply that
  1126. 42:09this managing towards some sort of
  1127. 42:11stationary or static target leverage
  1128. 42:13ratio doesn't seem to be a first order
  1129. 42:15concern for managers. They don't seem to
  1130. 42:17be paying that much
  1131. 42:20Well, that's kind of interesting in the
  1132. 42:22sense of
  1133. 42:23what we think of in terms of capital
  1134. 42:25structure models. Many of them are sort
  1135. 42:27of stationary static target sorts of
  1136. 42:30models. And this is kind of saying maybe
  1137. 42:33we shouldn't be thinking along those
  1138. 42:35lines in terms of theoretical capital
  1139. 42:37structure. And the second aspect of it
  1140. 42:40that it seems to be pointing
  1141. 42:43towards financing and investment needs
  1142. 42:45and then the subsequent evolution of
  1143. 42:47cash flows
  1144. 42:49as being more first order drivers of the
  1145. 42:51dynamics of leverage through time.
  1146. 42:55All right. Now, let's be clear that, you
  1147. 42:57know, when you do a study like this,
  1148. 43:01we in no way can claim some causation in
  1149. 43:04any of the anything that I just
  1150. 43:06described to you. You know, and we have
  1151. 43:07to be really careful as we're as we're
  1152. 43:09laying out these results, that we're not
  1153. 43:12at all making any such claims
  1154. 43:14about causation. All right. But why
  1155. 43:17might it still be useful? What in a
  1156. 43:19sense what we're trying to do is we're
  1157. 43:20sort of
  1158. 43:22reducing the the scope of the models
  1159. 43:24that are really plausible models for how
  1160. 43:27firms really behave with respect to
  1161. 43:29capital structure. And if we can do
  1162. 43:31that,
  1163. 43:32and if we can do that sufficiently,
  1164. 43:35uh then we feel like, well, that's a
  1165. 43:37contribution that that is useful to the
  1166. 43:40capital structure literature.
  1167. 43:41Fortunately for us, a referee agreed
  1168. 43:44that
  1169. 43:45uh it was it was useful enough.
  1170. 43:47All right. Now, a similar sort of
  1171. 43:50example
  1172. 43:51uh is this recent paper by DeAngelo and
  1173. 43:53Roll that um I don't think it's out yet
  1174. 43:56in the in Journal of Finance. That's
  1175. 43:57that's forthcoming
  1176. 43:59in Journal of Finance. So,
  1177. 44:01what they do at a very basic level is is
  1178. 44:04is fairly simple. They're really just
  1179. 44:06describing
  1180. 44:08the variation in leverage ratios
  1181. 44:11within individual firms through time.
  1182. 44:14Now,
  1183. 44:15why should you do that? I mean, why
  1184. 44:17would you start a study like that?
  1185. 44:21If you think you're going to contribute
  1186. 44:23to the capital structure literature.
  1187. 44:24Well, the the idea
  1188. 44:26is that most of the models and most of
  1189. 44:29our most of the thinking, I think,
  1190. 44:31within the profession through time has
  1191. 44:33been that capital structure within firms
  1192. 44:36is reasonably stable
  1193. 44:38through time.
  1194. 44:40Right? And where you really get a lot of
  1195. 44:41variation is in the cross-section.
  1196. 44:44Right? And that's kind of reflected in
  1197. 44:46the type of work that you see in the
  1198. 44:48capital structure literature. That most
  1199. 44:50of the models and most empirical work
  1200. 44:52are are designed to try to explain that
  1201. 44:54cross-section.
  1202. 44:56Why the cross-sectional variation? Well,
  1203. 44:58that's not necessarily a good approach
  1204. 45:01if what's really going on is that a lot
  1205. 45:04of variation is happening within the
  1206. 45:05firm over time. If that's just as
  1207. 45:07important as the cross-section, then we
  1208. 45:09sort of got to rethink
  1209. 45:11what our models are really saying and
  1210. 45:13how we're approaching the empirical
  1211. 45:14works. That's sort of where this where
  1212. 45:17the study is headed from the outset.
  1213. 45:19Again, the idea is it's descriptive data
  1214. 45:22analysis that's trying to describe the
  1215. 45:24data in a way that's going to cause
  1216. 45:26people to rethink
  1217. 45:28how they're approaching capital
  1218. 45:30structure and if they're finding
  1219. 45:31something that's unique enough
  1220. 45:33that does cause people to do this
  1221. 45:35rethinking of
  1222. 45:37how capital structure is determined,
  1223. 45:39this is a useful way to to approach the
  1224. 45:42problem. Now, what did they find? Well,
  1225. 45:43first of all, they do find
  1226. 45:45uh substantial
  1227. 45:47instability in the leverage ratios of of
  1228. 45:50individual firms
  1229. 45:52through time.
  1230. 45:54Right? A lot of variability. And
  1231. 45:56surprisingly, I think, to a to a lot of
  1232. 45:58people, a ton of variability in
  1233. 45:59individual firms. They do find episodes
  1234. 46:03in which firms look like they have
  1235. 46:05relatively stable leverage ratios, but
  1236. 46:08those are pretty limited
  1237. 46:10to time periods in which leverage level
  1238. 46:13is really quite low, often times zero.
  1239. 46:16So, you see low stability where some
  1240. 46:18firms have no debt and they keep no debt
  1241. 46:20for a while or really low debt, they
  1242. 46:22keep that for a while.
  1243. 46:24Right? And that's when you tend to see
  1244. 46:25stability. When leverage is higher, it
  1245. 46:27tends to be bouncing all over the place.
  1246. 46:29Now, why does it bounce all over the
  1247. 46:30place? Well, they find like like we did
  1248. 46:33in a paper of Steve McKeon and I, they
  1249. 46:35find it seems to be connected with the
  1250. 46:38episodes in which there are strong
  1251. 46:40investment needs. Right? And this is
  1252. 46:42where they sort of will they they do a
  1253. 46:44really nice job
  1254. 46:46of combining some clinical or case-based
  1255. 46:48evidence to complement what they're
  1256. 46:50doing on on a large-scale basis in in
  1257. 46:53terms of documenting these stability.
  1258. 46:55Right? They they dig into the data and
  1259. 46:58identify well, what seems to be the
  1260. 47:01underlying cause for these individual
  1261. 47:03firms for
  1262. 47:04for increasing their leverage when they
  1263. 47:06see an increase in or decrease in when
  1264. 47:08they see it go down. And and they find
  1265. 47:10pretty strong evidence that it seems to
  1266. 47:12be associated first with the company
  1267. 47:15expansions
  1268. 47:16uh during the changing economic
  1269. 47:18circumstances for those firms and then
  1270. 47:20subsequently
  1271. 47:22uh reduction in leverage in those time
  1272. 47:24periods in which it it it's actually
  1273. 47:28uh either excess cash flows or they're
  1274. 47:30producing other assets that they
  1275. 47:31transfer.
  1276. 47:33Right? So, again, I think this is very
  1277. 47:35useful type of evidence in this approach
  1278. 47:37even though it looks like a very simple
  1279. 47:40data description approach at the at the
  1280. 47:42outset, it's carefully done in a way
  1281. 47:45that ultimately when you're done with
  1282. 47:46it, you're thinking, "Well, I got to
  1283. 47:48rethink the way I'm thinking about
  1284. 47:50capital structure as a result of this."
  1285. 47:52So, it's very informative to subsequent
  1286. 47:54theory because now any theory that comes
  1287. 47:57along that purports to be modeling how
  1288. 48:01capital structure is actually chosen has
  1289. 48:03to deal with with this sort of facts
  1290. 48:05that are out there, these stylized facts
  1291. 48:07that that say that the firm's action
  1292. 48:15Okay, that's sort of that's sort of what
  1293. 48:17one is summarizing here. There's
  1294. 48:18incredible theories of capital structure
  1295. 48:21that are going to have to deal with this
  1296. 48:22time series variation and not just focus
  1297. 48:25on some construction anymore.
  1298. 48:27It also is implying that
  1299. 48:30sort of contrary to probably what most
  1300. 48:32of us teach in in the classroom with
  1301. 48:34respect to capital structure that the
  1302. 48:37leverage per se is really kind of a
  1303. 48:39second order importance
  1304. 48:41to valuation.
  1305. 48:42And the firm's behavior is that they
  1306. 48:44don't really care that much what their
  1307. 48:46leverage ratio is at any point in time.
  1308. 48:49You know,
  1309. 48:50if you're like me, you're teaching in
  1310. 48:52the classroom you're oftentimes teaching
  1311. 48:54a sort of a
  1312. 48:55you know, static tradeoff sort of model
  1313. 48:56where you're optimizing at this point
  1314. 48:59that
  1315. 49:00is the capital structure that maximizes
  1316. 49:02the value. This kind of implies that
  1317. 49:04either that that nice
  1318. 49:07inverted U-shaped curve that we draw up
  1319. 49:09on on the whiteboard for capital
  1320. 49:12structure is either really flat
  1321. 49:15or it's total nonsense. One of one one
  1322. 49:17of the three.
  1323. 49:20And third, then obviously then what
  1324. 49:22seems to be true then is that the main
  1325. 49:24determinants of leverage ratios
  1326. 49:27must be factors that are a little
  1327. 49:29different than what we traditionally
  1328. 49:30think of as standard measures of taxes
  1329. 49:33and distress costs that are their
  1330. 49:35tradeoff.
  1331. 49:38Okay, so let's move on to another
  1332. 49:40another different type of study. So
  1333. 49:42those last two are kind of descriptive
  1334. 49:45data analyses that
  1335. 49:47you know, hopefully I've convinced you
  1336. 49:48that they they shed some useful light
  1337. 49:51on the capital structure puzzle. Another
  1338. 49:53way to go about this
  1339. 49:55uh is to do a longitudinal type of
  1340. 49:58study. And again, the
  1341. 50:00the potential merits of a longitudinal
  1342. 50:02study is that you've got this long
  1343. 50:04period of time where lots of things
  1344. 50:06might be changing. Lots of the
  1345. 50:08underlying factors that could determine
  1346. 50:10capital structure can change over a long
  1347. 50:13period of time, and we can see whether
  1348. 50:15capital structures tend to change in the
  1349. 50:17direction that we think they should
  1350. 50:19based on those changes and and
  1351. 50:22underlying fundamentals. So, this study
  1352. 50:24by Graham, Leary, and Roberts that's
  1353. 50:26coming out
  1354. 50:27uh in the JFE is a good example of that.
  1355. 50:30I mean, they're really looking at
  1356. 50:31leverage ratios over nearly a century
  1357. 50:35uh within the US. So, you've got a
  1358. 50:37period of time that's spanning
  1359. 50:39uh World War II, uh the
  1360. 50:43uh macroeconomic expansion that takes
  1361. 50:45place after World War II in the US, high
  1362. 50:48inflation period in the '70s in the US,
  1363. 50:51and most recently the financial crisis.
  1364. 50:53All this gets spanned
  1365. 50:56uh in their study, and they can ask then
  1366. 50:58some fair fairly simple questions. You
  1367. 51:00know, first, have capital structures
  1368. 51:02changed at all on average over that
  1369. 51:04period of time?
  1370. 51:05Um secondly, can our existing models
  1371. 51:09account for those changes? Are they
  1372. 51:11consistent with those those changes?
  1373. 51:14And then third, if not, and the answer
  1374. 51:16does seem to be no in this case,
  1375. 51:19what underlying forces seem to be
  1376. 51:21driving this this variation? So, so
  1377. 51:23again, when you when you approach the
  1378. 51:26question in this way, the the result
  1379. 51:29that you're going to get
  1380. 51:31um you know, unlike maybe you find that
  1381. 51:33everything's consistent with what we
  1382. 51:34thought. You usually don't find that
  1383. 51:36kind of thing in reality. So, if you
  1384. 51:38don't, all right, your your end result
  1385. 51:41is to present a set of facts that are a
  1386. 51:44challenge to existing theory. If they're
  1387. 51:46going to inform theory, it's going to be
  1388. 51:48useful in that sense.
  1389. 51:50All right, and it's also going to narrow
  1390. 51:52the set of of theories that are truly
  1391. 51:55credible.
  1392. 51:56Any any theory that purports to explain
  1393. 51:58capital structure is going to have to
  1394. 52:00account for this this time series
  1395. 52:02variation in leverage ratios that
  1396. 52:04doesn't seem to be accounted for by our
  1397. 52:07our standard factors.
  1398. 52:09So, what do they find? All right, they
  1399. 52:11What they find is first of all that
  1400. 52:12there's been
  1401. 52:14a fairly large increase in leverage over
  1402. 52:16time. If you go back to sort of the
  1403. 52:18post-World
  1404. 52:20War I period and look at average
  1405. 52:22leverage leverage ratios back then,
  1406. 52:25uh I think
  1407. 52:26I can't remember the magnitudes, but
  1408. 52:28they're nearly double now what they were
  1409. 52:30back then. All right?
  1410. 52:32Coincident with that, we see a fairly
  1411. 52:35large decline in cash holdings.
  1412. 52:39Pronounced as they suffered an increase
  1413. 52:40in leverage. Now, that may be surprising
  1414. 52:42to some because a lot of what you put
  1415. 52:45you read in the press and even in some
  1416. 52:47our studies in corporate finance in
  1417. 52:49recent years talk a lot about these huge
  1418. 52:51cash balances that that US firms have
  1419. 52:54now.
  1420. 52:55That's a more recent phenomenon. And
  1421. 52:57these cash balances actually were
  1422. 52:59enormous back in the early part of the
  1423. 53:02of the last century, and they declined
  1424. 53:05quite a bit over the period of time as
  1425. 53:07these leverage ratios have
  1426. 53:09have gone up. It's only recently that
  1427. 53:11they their sort of cash balances have
  1428. 53:14turned back up.
  1429. 53:15All right.
  1430. 53:16What's interesting now is that firm
  1431. 53:18characteristics that we normally
  1432. 53:20associate
  1433. 53:21with capital structure on average don't
  1434. 53:24really change significantly in a way
  1435. 53:27that's consistent with these firms
  1436. 53:29having greater debt capacity.
  1437. 53:32So, think of our traditional
  1438. 53:34determinants of leverage. They don't
  1439. 53:36account for the fact that leverage
  1440. 53:37ratios are so much higher now than they
  1441. 53:40were, you know, back in 1920 or so.
  1442. 53:44All right. What's also interesting, I
  1443. 53:46think, is that they do find a negative
  1444. 53:49association between corporate borrowing
  1445. 53:53and government borrowing
  1446. 53:55in the US. And what sort of gives some
  1447. 53:57hints at the possibility that there's
  1448. 54:00there's there's some crowding out that's
  1449. 54:02taking place. Government is borrowing a
  1450. 54:04lot. Uh that's crowding out the
  1451. 54:07corporate sector from from borrowing at
  1452. 54:09at reasonable rates. Um
  1453. 54:11So, again, once once they're done with
  1454. 54:14this, have they established any
  1455. 54:16causation between variables and and
  1456. 54:18leverage? No.
  1457. 54:20They're not claiming any such causation.
  1458. 54:22Even with their last result of the
  1459. 54:24government crowding out, the most
  1460. 54:26they're saying is that, you know,
  1461. 54:28there's some hints of that in the data.
  1462. 54:30All right. Which I think is useful in
  1463. 54:32terms of subsequent studies that that
  1464. 54:34that could potentially examine that
  1465. 54:36question in in more detail. So, you
  1466. 54:39know, these aren't This is again not a
  1467. 54:41study that has
  1468. 54:43great identification of anything.
  1469. 54:45All right. But, I would argue it's very
  1470. 54:47useful to us as as scholars that who are
  1471. 54:50interested in in capital structure cuz
  1472. 54:53it's it's again narrowing the range of
  1473. 54:55plausible explanations for for what's
  1474. 54:57going on. You finish this paper, you
  1475. 55:00learn something that you didn't already
  1476. 55:02know
  1477. 55:03about capital structure. That's the
  1478. 55:05definition of a contribution.
  1479. 55:09All right. Now, the last thing I'll talk
  1480. 55:11about is again one of my own and I
  1481. 55:13apologize in advance for for
  1482. 55:16talking about another study of mine.
  1483. 55:18Again, I'm not representing these as the
  1484. 55:20best studies. I'm just
  1485. 55:22representing numbers as representative
  1486. 55:24studies of a particular approach. So,
  1487. 55:26this is This is another longitudinal
  1488. 55:29type of study that I've been working on
  1489. 55:31with co-authors Lance Barrett and and
  1490. 55:33and Ken Lane.
  1491. 55:35In which we're studying capital
  1492. 55:36structure decisions of US firms over a
  1493. 55:39period of time that actually predates
  1494. 55:41the Graham and Harvey and Roberts paper.
  1495. 55:43Our our period of time is 1905
  1496. 55:46to 1924. And I'll apply why I want to
  1497. 55:49study capital structure back in
  1498. 55:52the dark ages like that, especially if
  1499. 55:55if you've ever looked at data back then.
  1500. 55:56It's it's a little challenging to
  1501. 55:59to look at financial data for a period
  1502. 56:02that predates the SEC. It's It's
  1503. 56:04standard They standardization of data is
  1504. 56:06not particularly good
  1505. 56:09back at that time. But,
  1506. 56:11we thought, "Well, this is a very useful
  1507. 56:13time period to take a look at for
  1508. 56:15capital structure theories for two
  1509. 56:17reasons." And one is that it spans the
  1510. 56:19period of time
  1511. 56:21in which the US first introduced
  1512. 56:23corporate and personal income tax rates.
  1513. 56:25So, prior to 1909, I believe it was,
  1514. 56:29there were no corporate income taxes.
  1515. 56:31There were no personal income taxes,
  1516. 56:33either, in the US. Great days, but
  1517. 56:37not anything any of us have experienced.
  1518. 56:40So, if taxes are really important, then
  1519. 56:42we ought to see it when taxes are
  1520. 56:44introduced. Was sort of the logic of of
  1521. 56:46this. At the same time, this spans this
  1522. 56:50period of time in which the US enters
  1523. 56:53World War I,
  1524. 56:55which we show in the paper is associated
  1525. 56:57with very large but short-term shock to
  1526. 57:00investment opportunities for for the US
  1527. 57:03firms.
  1528. 57:04Right? And what's what's neat, we think,
  1529. 57:06from
  1530. 57:07a
  1531. 57:08theoretical standpoint is that that
  1532. 57:10shock to investment opportunities is
  1533. 57:12plausibly quite exogenous. No one's
  1534. 57:14really anticipating the outbreak of
  1535. 57:17World War I.
  1536. 57:19Right? And so, we can see how firms
  1537. 57:21react
  1538. 57:23to the shock to investment opportunities
  1539. 57:25at the same time as they're we're
  1540. 57:28observing how they react to fairly large
  1541. 57:31changes in the personal and corporate
  1542. 57:34income tax rates because one of the
  1543. 57:35other things I won't get into the
  1544. 57:37details, but one other things that you
  1545. 57:39see when the war breaks out is that the
  1546. 57:42US imposes very large excise taxes on on
  1547. 57:46the firms, particularly the firms that
  1548. 57:48would potentially benefit from from the
  1549. 57:50outbreak of of the war. So, there's this
  1550. 57:52very large change in marginal tax rates
  1551. 57:55that actually gives a strong incentive
  1552. 57:57to equity financing during the war. And
  1553. 58:00so, we're able in some sense to have a
  1554. 58:02bit of a horse race going on between tax
  1555. 58:05effects and
  1556. 58:06dynamic investment effects. All right.
  1557. 58:08So, what we find
  1558. 58:10in the paper is that there's really very
  1559. 58:12little evidence that shocks to tax rates
  1560. 58:15have much of an impact at all
  1561. 58:17on the leverage ratios.
  1562. 58:19You see these corporate
  1563. 58:21income taxes initially imposed,
  1564. 58:24basically nothing happens to the average
  1565. 58:26leverage ratio.
  1566. 58:28All right. But, what we do see, which is
  1567. 58:30quite similar actually to the DeAngelo
  1568. 58:32Roll paper, is that even though average
  1569. 58:35leverage ratios aren't changing very
  1570. 58:37much,
  1571. 58:38uh individual firm leverage ratios are
  1572. 58:41bouncing all over the place during this
  1573. 58:44period of time. And we're able to link
  1574. 58:46it pretty tightly
  1575. 58:47with the evolution of investment
  1576. 58:49opportunities and and cash flows in
  1577. 58:52these firms over this over this period
  1578. 58:54of time.
  1579. 58:55All right. So,
  1580. 58:56so our interpretation is kind of similar
  1581. 58:59to the DeAngelo Roll argument that, you
  1582. 59:02know, it looks like the dynamics and
  1583. 59:04investment opportunities and cash flows
  1584. 59:06are what's really driving the leverage
  1585. 59:08ratios rather than some of the
  1586. 59:11traditional determinants of of of
  1587. 59:14leverage. All right. But,
  1588. 59:16you know, in trying to interpret this,
  1589. 59:19there are challenges, some of which are
  1590. 59:21closely related to the identification
  1591. 59:23challenges that that we've been talking
  1592. 59:25about. One is that when you look at when
  1593. 59:28you're doing a historical longitudinal
  1594. 59:29study, if you're going to focus in on
  1595. 59:32specific shocks that interest you, you
  1596. 59:34have to be aware of the fact that
  1597. 59:36they're not the only shocks that are
  1598. 59:37going on
  1599. 59:38during this period of time. So, in our
  1600. 59:41case, there's sort of a
  1601. 59:43a market panic that takes place in in
  1602. 59:451907.
  1603. 59:47The Federal Reserve is created in 1913.
  1604. 59:52And after the the end of World War I,
  1605. 59:55there's actually a pretty strong
  1606. 59:57recession that worries on on the on the
  1607. 1:00:00depression
  1608. 1:00:02during that period of time. All of which
  1609. 1:00:05could potentially have some impact on
  1610. 1:00:07observed leverage ratios. So, even
  1611. 1:00:09though we can find some evidence that's
  1612. 1:00:11consistent
  1613. 1:00:13uh with this investment shock associated
  1614. 1:00:16with World War I being a causal
  1615. 1:00:18determinant of of the leverage changes,
  1616. 1:00:21we can't necessarily point to that
  1617. 1:00:23unless we do something more. I mean, so
  1618. 1:00:25so where we are in the study now is sort
  1619. 1:00:27of trying to supplement what we've got
  1620. 1:00:30with a little bit more of a clinical
  1621. 1:00:32analysis of individual companies. That's
  1622. 1:00:35involved actually collecting the annual
  1623. 1:00:38reports from around 1917 for these 60 or
  1624. 1:00:42so firms that that comprise our sample.
  1625. 1:00:45So, as you might expect, they're they're
  1626. 1:00:47not in most libraries. So, you know, you
  1627. 1:00:50got to go find in our case I go to
  1628. 1:00:52Columbia University Library, get these
  1629. 1:00:55all copied. You know, it's a huge stack
  1630. 1:00:58of annual reports that that we've got to
  1631. 1:01:00go through. But it's a very useful way
  1632. 1:01:03to try to tease out what's really
  1633. 1:01:05driving these changes in leverage. Cuz
  1634. 1:01:07now when you read these annual reports
  1635. 1:01:09and the narrative that's in the annual
  1636. 1:01:11reports, like you can actually see that
  1637. 1:01:14the firms are saying specifically, you
  1638. 1:01:17know, we're borrowing in order to do
  1639. 1:01:19this.
  1640. 1:01:20All right? And that way, you are getting
  1641. 1:01:22directly at the issue of both causation
  1642. 1:01:26and the data by by using a supplementary
  1643. 1:01:29clinical type of of evidence. You don't
  1644. 1:01:31have a clean natural experiment. You
  1645. 1:01:33don't have instrumental variables. I but
  1646. 1:01:36looking in a clinical sense, you are
  1647. 1:01:38getting at some some of the same issues.
  1648. 1:01:40And in fact, I I would encourage you to
  1649. 1:01:42think about
  1650. 1:01:44sort of clinical sort of analysis in in
  1651. 1:01:47in the same way that you would think of
  1652. 1:01:48as a natural experiment. And both of
  1653. 1:01:51them are identifying sort of a narrow
  1654. 1:01:54situation in which you can fairly
  1655. 1:01:57cleanly identify what's really going on.
  1656. 1:02:00The natural experiment does it one way.
  1657. 1:02:03The clinical study does it it does it
  1658. 1:02:05another way. A structural model, for
  1659. 1:02:06that matter, does it yet another way.
  1660. 1:02:08It's a very specific set of conditions,
  1661. 1:02:10this what this is what happens.
  1662. 1:02:12All these are different ways of
  1663. 1:02:15obtaining the same piece
  1664. 1:02:17in in essence.
  1665. 1:02:19All right, so let me let me just kind of
  1666. 1:02:20conclude at this point and then open up
  1667. 1:02:22to to questions. So,
  1668. 1:02:24the points I want to make are first of
  1669. 1:02:26all that
  1670. 1:02:27these identification problems aren't
  1671. 1:02:29going away.
  1672. 1:02:30They're pervasive in corporate finance
  1673. 1:02:33research. They're
  1674. 1:02:35much more so in corporate finance
  1675. 1:02:37research than in some other aspects of
  1676. 1:02:40of finance research. They're there.
  1677. 1:02:42They're real problems. They're things we
  1678. 1:02:44have to deal with as as scholars in
  1679. 1:02:46order to be able to tease out some sort
  1680. 1:02:49of believable conclusion. Econometrics
  1681. 1:02:52can help and has
  1682. 1:02:53really helped. I think we've got great
  1683. 1:02:55advances
  1684. 1:02:57in our understanding of econometrics,
  1685. 1:02:59like I said, over the last 10 to 15
  1686. 1:03:01years. And they really help us narrow
  1687. 1:03:03down the scope of this problem. They can
  1688. 1:03:05never solve the problem. Econometrics
  1689. 1:03:07doesn't make the identification problem
  1690. 1:03:09go away.
  1691. 1:03:11It just sort of narrows the the scope.
  1692. 1:03:14Natural experiments are in some ways the
  1693. 1:03:16most useful econometric solution.
  1694. 1:03:19Um, but but I worry that an
  1695. 1:03:21over-reliance on these natural
  1696. 1:03:24experiments is it can be harmful
  1697. 1:03:27in our pursuit of knowledge in the sense
  1698. 1:03:29I described before, that it really kind
  1699. 1:03:31of
  1700. 1:03:32it it if you think you have to have a
  1701. 1:03:33natural experiment, you are constraining
  1702. 1:03:36yourself in the types of questions that
  1703. 1:03:38you can really ask. You know, if the
  1704. 1:03:39profession as a whole were to do that,
  1705. 1:03:42then we're constraining the body of
  1706. 1:03:43knowledge that that we're producing
  1707. 1:03:46about questions in in uh in
  1708. 1:03:48in corporate finance. So,
  1709. 1:03:51I'd like to encourage you to think about
  1710. 1:03:52these alternative methodologies that
  1711. 1:03:54that we went through as as equally
  1712. 1:03:56useful useful ways of of expanding that
  1713. 1:04:00uh body of knowledge, even if they're
  1714. 1:04:03not able to really provide complete
  1715. 1:04:05identification. And they frequently
  1716. 1:04:07don't. Uh
  1717. 1:04:09but I would argue that's okay. As long
  1718. 1:04:11as you understand and can explain the
  1719. 1:04:14limitations of of what you're doing, you
  1720. 1:04:16can still produce papers that are are
  1721. 1:04:19very useful in terms of our
  1722. 1:04:21understanding of the problems.
  1723. 1:04:23All right.
  1724. 1:04:23So, let me stop there and
  1725. 1:04:25open up to any questions.
  1726. 1:04:29You correct that he's using the
  1727. 1:04:31simultaneous model, but that doesn't
  1728. 1:04:33mean
  1729. 1:04:34it it there's some other omitted
  1730. 1:04:36variable he was trying to specify.
  1731. 1:04:39Thank you.
  1732. 1:04:45Anything else?
  1733. 1:04:50Okay, then. Well, thanks for listening
  1734. 1:04:52and uh yeah.

About this transcript

This page contains the full transcript of YouTube transcript (DKWVDvk1QYo) , generated from the public captions YouTube serves with the video. The transcript has 10,446 words across 1,734 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.