YT YouTube2Text

YouTube2Text → Transcripts → Conducting Empirical Research on Corporate Finance

Conducting Empirical Research on Corporate Finance — Transcript

by Financial Management Association International · 11,478 words · 1,671 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:04like to welcome you to this this session
0:07that we've titled conducting empirical
0:09research and in corporate finance and
0:11before we get going I want to first at
0:14the risk of embarrassing am I see Bob
0:15Marino in the back there and I want to
0:17thank him first of all for inviting me
0:19to do this particular session but also
0:23perhaps more importantly thank him for
0:25putting together this entire program
0:27it's a big job and we're grateful for
0:29all the work you've done on this Bob
0:31thank thanks very much it's also I have
0:34to say we're very I'm very grateful or
0:37so many people here in the room I'm
0:38flattered that that's the case but it's
0:41also somewhat reassuring to me it's a at
0:45lunch today I was having lunch with a
0:47couple of colleagues from different
0:48schools who work in the in the area of
0:51corporate finance and we were kind of
0:53bemoaning the fact that there seems to
0:55be less and less corporate finance
0:57actually getting done by people people
1:00in our field even by people who say
1:02they're doing corporate finance and you
1:04see what they're doing and it turns out
1:06it's really household finance or
1:08something a little bit different not
1:10that there's anything wrong with those
1:11fields of us just as corporate finance
1:13scholars you know we have a particular
1:15interest in the questions that that are
1:18eyes and in corporate finance and so I'm
1:21presuming the fie the fact that you're
1:22here that you have a similar interest
1:24and that to me is very reassuring to see
1:26so many young people in the profession I
1:28mean having an interest in the questions
1:30that that we tend to study and in this
1:33particular field so you know I always
1:36tell my PhD students you know when
1:38you're writing papers you know you need
1:40to learn how to write a good
1:42introduction a good introduction kind of
1:44spells things out for the readers right
1:46away so they they know where you're
1:48going with this so let me sort of get
1:50you to the punchline of what I'm talking
1:52about today and tell you what what what
1:55this talk is and what this talk is not
1:57what what I'm really trying to get and
2:00convince you of ultimately in this talk
2:02I think is that our field of corporate
2:05finances is full of very interesting
2:07questions and these questions can be
2:10posed in a variety of ways right there's
2:14not just one right way to pose the
2:16question but there
2:17trying to get out even if they're
2:19getting at the same fundamental issue
2:20there's different ways you can ask the
2:22question and these different ways of
2:25asking the question naturally lead
2:27themselves to the different sort of
2:29empirical approaches to answering those
2:31questions I I would take the approach
2:35that there's no right way to do research
2:37there are many different ways that you
2:40can do conduct research in a way that's
2:44going to be informative to the
2:45profession so part of what I'm going to
2:47talk about are what I perceive as or the
2:50main ways or the main empirical
2:52approaches that people take in our
2:54profession that have been successful and
2:56by successful I mean both yeah yes they
2:59got their paper published sugared
3:01journal right but also these papers have
3:03made an impact on the field I'm not here
3:06to single out specific papers but rather
3:08to to talk more about the approaches to
3:11getting at questions and again with the
3:14idea in mind that there's many different
3:16ways to get to the types of answers that
3:19we hope to get in this field of of
3:21corporate finance all right so with that
3:24in mind I think the outline of my talk
3:26is really pretty straightforward I'm
3:28going to talk for a few minutes about
3:30you know what is sort of the nature of
3:31the questions that we tend to address in
3:34corporate finance and then I'm just
3:36going to run through a list of what I
3:38perceive as or the primary empirical
3:40approach is that people do take to
3:43address these questions and none of them
3:45are perfect they all have their
3:47limitations they all have their virtues
3:49as well I mean it's sort of the matching
3:51process I think when we do when we
3:53conduct research matching the question
3:55to the approach that we take that
3:57sometimes goes a long way to to
4:00ultimately determining the success of
4:02the project so I'm going to talk about
4:04six primary approaches one sort of the
4:07old school events study that we don't
4:08see as much of anymore Jeff might
4:11remember it but the rest of us really
4:13don't I the the next next five are
4:20really what we see more in the
4:21literature today but I don't mean to say
4:24the event studies are not useful or
4:26passe because I do think they they can
4:29play a very important role in in
4:30research
4:31but I think we do want to understand
4:33what the limitations to them are as well
4:35so I'll talk about the united studies
4:37well I what I call exploratory data
4:40analysis it is almost purely a
4:42descriptive exercise sort of your
4:45classic OLS regression approach that may
4:48be involved panel data techniques if you
4:50have that sort of data
4:51I fourth what I think people consider as
4:55the Holy Grail in terms of identifying
4:58causal relationships in our fields or
5:01the randomized controlled experiment and
5:03I'll say up front that you know we're in
5:05corporate finance we're not in a hard
5:08science we don't have perfect randomized
5:10controlled experiments but we have ways
5:12to try to approximate that and that's
5:14what we tend to do with natural
5:16experiments in our field fifth clinical
5:19studies which is another type of study
5:21that you don't see as much of as you did
5:24at some earlier points in time but
5:26that's one that I'm going to push a
5:27little bit on as a way to supplement
5:30your research that it can be very
5:32informative I can add a lot of value to
5:34to your studies and then finally
5:37structural estimation as an alternative
5:39way of getting up some of the same
5:41questions and I'll just try to tie
5:43together things in the end I'm going to
5:46try not to drone on too too long in the
5:49talk so that there's plenty of time for
5:51Q&A in the end but I guess I would say
5:54if along the way there's there's some
5:56clarifying questions that you've got
5:58don't don't hesitate to ask but but I am
6:01intending to leave time for for Q&A in
6:04the end all right so so what what kinds
6:09of things do we ask in corporate finance
6:11right for years I was at Purdue
6:13University I taught the core corporate
6:16finance class with my colleague John
6:18McConnell and we always start off by
6:20telling the students that you know
6:21there's really only two questions in
6:23corporate finance and we try to make it
6:25as simple a field as we could possibly
6:27get for them didn't work for a lot of
6:28for a lot of them but but in reality
6:31we've it's true that there really are
6:34only two fundamental questions in
6:36corporate finance right from where and
6:39how do companies get their capital right
6:42and secondly how they allocate capital
6:45I pretty much everything that we asked
6:48in corporate finance can really be put
6:51into those there's two primary bins you
6:54know you think of the classic capital
6:56structure questions of that versus
6:58equity structured finance financial
7:02distress entrepreneurial finance venture
7:05capital private equity all that will fit
7:08into how companies get their capital as
7:11would cash and liquidity policies and
7:13payout policies as well we're thinking
7:15in terms of sources and uses of funds
7:17constraints in a decision that you're
7:19making about payout and and liquidity
7:22policy has something to say about where
7:24and how companies are obtaining the
7:26funds that they need in order to conduct
7:28their their operations that's one
7:32fundamental question that we're asking
7:34the other how they allocate those funds
7:36they are we see whole literature's like
7:38the M&A literature or literature that's
7:41grown a lot in the last decade or so on
7:43innovation but even the field of
7:45governance you can really think of
7:47primarily is ultimately about how is the
7:50company allocating its capital okay so I
7:54mean I don't I don't think it's
7:56necessary that we that we think it has
7:57to fit into these two bins but I think
8:00it's it it's one way to simplify the
8:02fact that there really are only a couple
8:04fundamental questions that we ask in
8:07this field of corporate finance it
8:09doesn't mean it's a narrow set of
8:10questions ultimately right but the
8:13fundamentals are actually reasonably
8:16narrow but how we ask the questions that
8:18get at those those fundamental issues
8:21can be very different I mean so I've
8:24listed here sort of five categories this
8:27is not meant to be an exclusive list and
8:30exhaustive list right but these are
8:32different ways that we're gonna approach
8:34those fundamental issues that we study
8:37in corporate finance my one is to what
8:41extent do certain actions or certain
8:44policies affect the value of the
8:47enterprise or the value of the share
8:48that shareholders claim on an enterprise
8:50or perhaps the accounting performance of
8:53the enterprise and so typically we're
8:55asking are these policies or are these
8:59actions value increasing for the firm
9:01alright if so can we identify the
9:04channel or the mechanism that drives
9:07that value increase and if not that's
9:10not value increasing how come my is it
9:12is this an example of an agency problem
9:14and can we connect that somehow with
9:17governance issues we want to know the
9:19mechanisms are the channels that drive
9:21these these value effects ultimately
9:24right that's one way that's one approach
9:26one category of question that we might
9:29ask alternatively we might flip this
9:32around right in think in terms of on
9:35average companies making decisions that
9:38are that are optimizing decisions that
9:41maximize the value of the enterprise
9:43right so that if we study the
9:45cross-sectional determinants of how
9:47these policies are set we can learn
9:51something about what are the primary
9:53drivers of value from those particular
9:57policies I'll make this little bit more
9:59concrete shortly a third category of
10:03tests that we might conduct is to try to
10:06provide tests of specific existing
10:09theories right this is something that
10:11from my perception anywhere it was much
10:14more common 30 years ago then then we
10:17see now and I'm not entirely sure why
10:19that's true it might be that the models
10:21themselves sometimes don't lend
10:22themselves to direct empirical tests and
10:25sometimes it's that they do make
10:27specific predictions but they're really
10:29not testable with the types of data that
10:31we have but for whatever reason you
10:33don't see as many studies of that type
10:35these days but that's not to say they
10:37aren't useful right if we've got models
10:40that make specific predictions about how
10:43companies will behave and from a
10:45corporate finance standpoint right we'd
10:48like to be able to test those models in
10:50a very direct way if we can a fourth
10:53type of study that we see more and more
10:55of these days as data becomes more
10:57available in the international context
11:00and across different institutional
11:01frameworks is can we study the
11:03consequences of various legal changes or
11:07regulatory changes or differences in
11:09institutional frameworks on how
11:11companies make
11:13certain decisions my in fifth
11:16I can't the type of study I'm thinking
11:19about in category number five is this
11:21simply starts with direct observation
11:23all right we we are observers observers
11:26of the world and we see things have
11:28changed somehow right for I use an
11:31example here that you know we can see
11:33that companies hold much higher cash
11:35bail so my point here in this slide is
11:38simply to say it if you if we go back to
11:40thinking of there being these two
11:41fundamental issues or fundamental
11:43questions that we address and in
11:45corporate finance there's very different
11:47ways in which we can ask the questions
11:50and as I said before what I'm gonna try
11:53to convince you of is that how you
11:54ultimately attempt to address these
11:57questions should guide the way you
12:00conduct your empirical research so let
12:03me try to make it just a little bit more
12:05concrete by thinking in terms of
12:07governance research right so here the
12:09way I would think of as the the
12:11fundamental question that that
12:13governance research is is after is what
12:17mechanism or set of mechanisms is there
12:19that ensures that managers will take
12:22actions that are ultimately in the
12:23interests of the shareholders and that's
12:27ultimately we're trying to get at with
12:29with corporate governance so how do we
12:31do it well one way is to say well are
12:33there sets of governance mechanisms that
12:36are optimal in the sense that they
12:38maximize value so can we connect the
12:40value the firm with the governance
12:43mechanisms and try to try to deduce
12:45something about the optimal governance
12:47structure that way and of course you
12:49know that there's lots of identification
12:51problems and the way I've written the
12:53equation everybody that's basically the
12:55idea of what we're trying to get it
12:56there alternatively we can turn the
12:59equation around and say well let's try
13:01to explain why some companies have
13:04different governance structures and
13:05other companies do again with the idea
13:08being that on average firms are going to
13:10get it right right there's not a
13:12one-size-fits-all governance policy so
13:15let's try to understand the factors that
13:17determine why one firm has a board
13:21structure it looks like yes and another
13:23firm has a board structure it looks
13:25completely different why
13:26those things might be optimal for for
13:29both of those forms a third type of
13:32category of tests that you might see in
13:35the governance literature is to think
13:37about the actions that firms or boards
13:39are supposed to take and act and ask
13:42whether they are a function of the
13:44governance structure so think of the
13:46action for example as the board
13:48dismissal of a CEO all right we would
13:53think that that the likelihood of that
13:55dismissal is a function of performance
13:57so think of the X variable as
13:58performance and then we might be asking
14:01about whether the sensitivity to
14:03performance is different for firms with
14:05different governance structures I if we
14:09think the answer is yes and it tells us
14:11something about one governance structure
14:13maybe being more effective than the
14:15other governance structure all right so
14:17again three different ways of posing the
14:20question all of which trying are trying
14:22to get back to the fundamental issue
14:25that this particular literature is after
14:28okay so so how can we how can we
14:31approach these these different
14:33categories of questions in our research
14:36in corporate finance well as I mentioned
14:39before sort of events so if I go back to
14:41the time that I was getting my PhD back
14:44in the dark ages very hard to read a
14:47paper in corporate finance that didn't
14:50have some sort of events study in it I
14:52guess I'll talk about momentarily I mean
14:55part of the reason for that is I think
14:57there's some real virtues to the event
15:00study and there's some simplicity to it
15:01I did is attractive to all of us I
15:04thought there are limitations as well
15:06and that may explain some of those
15:08limitations may explain why we don't see
15:10as many event studies now as we did back
15:13then but the basic idea is pretty
15:14straightforward we can I we can identify
15:18announcements of events of some sort you
15:21know if you a merger or a change in
15:23financing policy all right we see
15:25companies announce what they're doing we
15:27can observe how the stock market or the
15:30bond market reacts to that all right we
15:32could do some a short run fashion all
15:35right what's a one-day return we can do
15:37so over longer haul if
15:40you know we're not sure about the speed
15:42of adjustment for the market makes that
15:44information we can even do with
15:46operating performance or some measure of
15:47accounting performance as well if we
15:49think that's a better indicator of value
15:51ultimately so examples would be you know
15:55what are the wealth effects of takeovers
15:57for the parties that are involved how
16:00does the market react when companies
16:02change their financing policy or how is
16:05it different if they choose different
16:06securities to raise the same amount of
16:09money what happens to the wealth of the
16:12shareholders if the company alters its
16:15payout policy tons of studies of that
16:18sort
16:19back in the 1970s in the 1980s and you
16:23know I think most people are familiar
16:24with with basically what the evidence
16:26says and this is a gross
16:28oversimplification of that evidence but
16:31if you look at the M&A literature
16:32basically says targets game bidders
16:35around zero at least for or acquisitions
16:39of public companies combined wealth
16:41effects are positive and if we look at a
16:44payout policy if you increase the payout
16:46the shareholders a form of dividends or
16:49stock buyback stock price goes up if you
16:52reduce the payout stock price goes down
16:55finance events are a little more
16:57complicated in a sense that if you look
17:00at equity financing events stock prices
17:02go down debt financing events maybe down
17:06a little bit close to close to zero
17:09private placements stock price actually
17:11goes up all right so if you stop and
17:15think about okay well what do we
17:17conclude from that type of events you
17:19kind of start to see what might be some
17:21of the limitations of the event study
17:24approach because you can think well on
17:26the one hand there will virtue of this
17:29approach is you get pretty clean
17:31identification of a causal impact here
17:34of an event announcement on shareholder
17:38wealth and by the way I think sometimes
17:42because we become very attuned to the
17:45identification problem in our profession
17:48in the last let's say ten years or so
17:51there's sometimes there's this
17:52misperception
17:54we didn't understand this identification
17:56problem before but I think everybody
17:59understood I back on these event studies
18:01are being done that was one of the
18:02virtues of them they say well I can get
18:04pretty clean identification if I do the
18:06study this way right the real problem is
18:09identification of what exactly right
18:12well you're getting is the clean
18:14identification of the impact of this and
18:16I'm assuming that you have a clean
18:18announcement you know there's not
18:19contaminated events being announced on
18:23the same day you've got a clean
18:24announcement you've got pretty clean
18:26identification of how the market viewed
18:30the impact of the announcements event on
18:32the value of the enterprise and if it's
18:36a stock price is that the value of the
18:37stock right what does that mean exactly
18:41it is really ultimately the question
18:43what the event studies it's really hard
18:46to separate what you think of as the
18:48real effect from some sort of
18:51information effect right so for example
18:53we look at the M&A literature
18:55it was always couched in terms of well
18:57due to mergers create value right stock
19:02prices went up for the targets and what
19:04happened a combined Sun so one
19:06interpretation is yes they create some
19:08value night but it could be equally
19:11plausible interpretation the fact you
19:14became a target signal something about
19:16what your intrinsic value is it's not
19:19the takeover to cause the value increase
19:21but rather there's information about
19:24what the true value is and that is in
19:26fact the way most of the the payout the
19:32interpretation most of the payout
19:34literature were went is that this is not
19:36a real effect as an information effect
19:38right so it's difficult to separate
19:41those two and a lot of the literature
19:43that began with these events studies was
19:46trying to follow up along those lines
19:47can we can we tease out is this an
19:50information effect or or real effect
19:52sometimes I think the literature has
19:54been successful in doing that sometimes
19:56less successful I'm thinking in in my
19:59opinion but my point is that it's
20:02difficult it's a very difficult thing to
20:04do and you have to understand that
20:05limitation of doing events so it doesn't
20:07mean
20:08you don't want to do it right I think in
20:11fact there's a lot of value in
20:12understanding whether the market seemed
20:14to react to a particular event you know
20:17think of the category of studies I
20:19mentioned before those looking at a
20:21regulatory change or a legal change of
20:23some sort might be you want to say or
20:26you want to argue that the regulatory
20:28changed caused firms to behave
20:32differently sort of investment changed
20:34in some way all right well one thing I
20:37think that anyone would like to know say
20:39well you say this regulatory event is
20:41causing this ultimately can you convince
20:45me that this regulatory event was really
20:47big enough to do anything well one way
20:50to do so is the conducted event study
20:52and say look the market reacted to the
20:55announcement of this regulatory change
20:57so the market believed something was
20:58going on at least it doesn't tell you
21:00what yet but at least sort of confirms
21:03the experiment in a sense that this
21:05isn't economically important enough
21:08event to study so I think the bottom
21:10line with the event studies that they
21:12can be useful right but you're rarely
21:14gonna find them to be useful enough in
21:17isolation you really I think these days
21:20we're good to be able to build an entire
21:22paper around just just the event study
21:24night but it does contain some useful
21:27may so so I I think combining it with
21:30other types of empirical analysis can be
21:32quite useful right but but understand
21:35what you can and cannot conclude from
21:37from that that sort of test all right so
21:40now let me turn to the other empirical
21:42approaches and this is this is where I'm
21:44going to talk about these primarily
21:47through through the lens of capital
21:49structure research and that's not
21:51because there's anything unique about
21:53the capital structure structure research
21:55but I think it's just a way to keep the
21:58question constant in a sense the
22:00underlying fundamental question constant
22:02all right while thinking about these
22:05different empirical approaches to that
22:07fundamental question the other reason is
22:09purely selfish that I know this
22:12literature a little bit better than
22:13other literature so I feel like I can
22:15comment more intelligently about it
22:18right what's the fundamental question
22:20yeah I think it's pretty straightforward
22:22is that how should firms optimally
22:25finance their operations in the way that
22:29theoretical literature is gone is that
22:30you know well we can start with the
22:33perfect Markets case of course and then
22:35really doesn't matter but to the extent
22:38that capital structure is going to
22:40matter it's going to matter mostly
22:41because of the existence of some
22:43frictions right what are those frictions
22:45potentially well it could be taxes it
22:48could be some costs of financial
22:50distress there could be some agency
22:51costs involved there could be some
22:54asymmetric information effects or ever
22:57selection problems that that create
22:59differences an optimal capital structure
23:01across firms right so what we'd like to
23:05know as researchers is what's important
23:08and what's not to the choice of the
23:12optimal financing your capital structure
23:14policy of the firm we also like to get
23:17at this the second issue at the bottom
23:20is the extent to which these are dynamic
23:22choices such that you know a capital
23:26structure decision today has an impact
23:29on your ability to finance your
23:30operations at some point in the future
23:32right or whether it's not really that
23:34dynamic and we can think of it more as
23:36you know in a static trade-off sort of
23:38sense right that's that's an additional
23:40issue on top of what the what the
23:44important frictions may or may not be
23:46okay so how can we get at the this
23:49fundamental question right well think
23:52about this in the sense of sort of a
23:54classic regression type of approach
23:56right what we're trying to explain the
23:59cross-sectional differences or really
24:02the determinants of the choice of
24:05capital structure or you know debt
24:08equity choice or whatever you want to
24:10call leverage it could even be have to
24:12do with with hybrid securities or any
24:15sort of structured securities on the
24:16left-hand side as well but we normally
24:18have in mind leverage as a function of
24:21some hypothesize determines I'm
24:23representing there as this vector X of
24:26hypothesized determinants right which if
24:29we just estimate that and you know LS
24:32sense then we've got the obvious
24:33identification problems
24:35could easily have some limited variable
24:38that that is correlated with both the
24:40the X X vector or parts of the X vector
24:44and the dependent variable the leverage
24:46variable so if we have panel data of
24:49course we can try to do something about
24:51that using panel data techniques those
24:54are imperfect as well for example we can
24:57put in firm fixed effects which I've got
24:59here as the gamma or time fixed effects
25:02with with the variable V but we still
25:06have these two primary issues that are
25:08gonna that are going to be hanging out
25:10there that is going to make this an
25:12imperfect way of trying to get whether
25:15there's a causal link between a
25:17hypothesized determinant and the
25:20dependent variable leverage do want is
25:22is again simply the standard and dodging
25:24ad issue of whether the the hypothesize
25:28determinant determinant is correlated
25:30with the aera turn my here with fixed
25:33effects we can't you know can't we can't
25:35rule that out obviously right the other
25:38issue that's going to come up in capital
25:40structure research anyway is the second
25:42one which is that you think about
25:44putting in a firm fixed effect for
25:47example then the identification that
25:50you're getting is coming from time
25:54series variation within a farm alright
25:57well that might be good for some
25:59problems but but bad for others and it
26:01might be bad for capital structure to
26:04the extent that there isn't that much
26:06variation in leverage over time if
26:08that's true I'm not saying to this chair
26:10because in fact it's not but if it was
26:13true there isn't much variation over
26:16time within a firm and that most your
26:18variation comes in the cross section as
26:20soon as you put in these firm fixed
26:22effects you're sort of like more washing
26:24away all the involve the interesting
26:25information in the test I and so you
26:28lose your ability to try to figure out
26:30what's it important determinant what's
26:32not so from an identification standpoint
26:35this approach is usually gonna fall
26:38short all right but that's not to say
26:40it's useless right you still you can
26:43sort of think of it as providing
26:45potentially some very useful descriptive
26:47information about what
26:48is correlated with the firm's leverage
26:50decision I could be connected with this
26:53existing theory to the extent that it is
26:56or is not consistent with what models
27:00would tell us are important or whether
27:02it's indicating something that's
27:03fundamentally different from the way we
27:05tend to think of as the hypothesized
27:07determinants but if you're going to try
27:09to write the paper as being this
27:12variable causes the difference in
27:14leverage that's going to be hard to do
27:15with this type of test if that's what
27:20you want to do then what you're really
27:22after is more of a randomized controlled
27:25experiment type of approach which as I
27:27mentioned before these don't really
27:29exist in in corporate finance right but
27:33there are times in which we have some
27:36so-called natural experiments that come
27:38pretty close very close in some cases
27:41due to the conditions that are there in
27:44a pure randomized controlled experiment
27:47where we have a group of firms or the
27:53little think of them as firms in this
27:54case is being treated in a random
27:56fashion and a set of of of control firms
27:59alright so think about a couple tests
28:02that I mentioned here that are trying to
28:04get at the impact of corporate taxes on
28:08the leveraged choice right again that's
28:10one of the hypothesized determinants and
28:12that's a point that's worth making here
28:14is that if you're going to go down this
28:16path of thinking in terms of something
28:18that looks like a randomized controlled
28:20experiment you're usually enough to
28:22narrow the question quite a bit in this
28:25case you're narrowing it down to one
28:27hypothesized determinant right that you
28:30have some control over in an in a
28:32natural experiment sort of setting
28:33alright so in the two papers that I
28:35mentioned here all right what you have
28:37are sort of staggered introductions of
28:40changes in in the corporate income tax
28:44rate I mean higher lung this paper it's
28:46it's changes in state corporate income
28:49tax race income tax rates and the pacio
28:52shew paper these are changes in rates
28:55across different countries right so what
28:59you have then is the ability to say
29:01right or the
29:02States or for those countries in which
29:04you have a change in the corporate
29:07income tax rate right but everything
29:09else is being held constant in a sense
29:11right do you see a change in in in the
29:16company's depth and the amount of data
29:17that's the way they've done them you can
29:19think of as changes in leverage ratios
29:21as well right so relatively clean and
29:27neat sort of set up right but this type
29:30of setup is always going to raise a
29:31couple of issues right the first one you
29:34have to think about always is well is
29:36this truly a randomized controlled
29:38experiment or not right what led to the
29:43so-called treatment in this case all
29:46right was there some sort of lobbying
29:47that took place this is not really
29:49random treatment at all in which case
29:51you've got a problem or is it somehow
29:55the case that general economic
29:58conditions are what led some states to
30:00change their corporate income tax rate
30:02in other state didn't if so it's eking
30:05out potentially the economic conditions
30:07that are driving the change in leverage
30:08and not the change in the tax rate so
30:11you could try to directly control for
30:13those things you can try to directly
30:15search and convince the readers that
30:18there wasn't any sort of lobbying effort
30:21right there somehow correlated with with
30:24the change in a tax rate but you have
30:26some work to do I think if you're trying
30:28to set this up to convince people that
30:29you that your so-called natural
30:32experiment really is approximating a
30:35randomized controlled experiment all
30:38right then of course you want to test
30:39parallel trends right were these how did
30:43these treated and control firms behave
30:45prior to the so-called treatment right
30:49were they moving in roughly the same
30:51direction so that seems like there is a
30:53change right around the treatment as
30:55opposed to well it looks like they
30:57started to deviate a couple years before
30:58this treatment in which case is probably
31:00something else that is driving the
31:03results right the third thing I want to
31:06point out I think it's perhaps the most
31:08important is that you really have to
31:09think about with your natural experiment
31:11you know what sorts of conclusions can I
31:14draw from the from the
31:16sort of thing and what sorts of
31:17conclusions can I not draw all right
31:19with it this sort of test this is really
31:21designed to isolate the causal impact of
31:25one factor like that's what your
31:28concluding something about you can't
31:30then say whether this is a first-order
31:32factor or a second-order factor or not
31:36you don't know how there's the
31:38importance of this factor compared with
31:40others unless you do something else
31:41maybe you compare the economic impact
31:44that you found for this fact it was
31:46somebody else's economic impact that
31:49they found for a different factor in
31:50another paper there's never one for one
31:53comparison there all right so it's going
31:55to be very hard to say well taxes are
31:57the most important factor that
31:59determines capital structure simply
32:01because you found that they are
32:02important in this experiment that's not
32:04a reason not to do this sort of thing
32:06it's more a statement about well what
32:08can you and can you not conclude from
32:10the sort of study you can conclude
32:12something potentially very decisively
32:14about the causal impact of one variable
32:17acts on the dependent variable ah that's
32:20extremely useful in our field night but
32:23don't try to take it further than that
32:25because that's when you'll get shot down
32:27because know you can't draw you any
32:29further inferences than that like the
32:31last thing I point out is that with
32:32natural experiments you do also always
32:35have to worry about external validity is
32:37there something unique about this
32:39experiment so that you know maybe taxes
32:42mattered in this very narrow sort of
32:44situation you studied but they don't
32:46matter more generally that's not a
32:48simple thing to really test in a study
32:52but you have to think about that and at
32:53least try to make an argument for why
32:56you think this has some externally
32:58beyond just the one experimental setting
33:02that you're you happen to be studying
33:04right so that that's in a sense like I
33:09said the holy grail of trying to get at
33:11causal impact is to do some sort of
33:13randomized controlled experiment sort of
33:17at the other end of the spectrum or what
33:20I would call the the sort of exploratory
33:22exploratory data analysis types of
33:25studies in these types of studies
33:28they're almost pure
33:29descriptive - sort of describing a set
33:33of facts and ultimately trying to
33:35describe them in a way that the
33:38professional think is extremely useful
33:39for understanding the fundamental
33:41question in this case how firms choose
33:44their their leverage ratios right it's
33:48not gonna be able to say that there's a
33:50causal impact if one thing aren't other
33:52doesn't ever there's always going to be
33:53an identification problem and studies
33:56like this right and so you have to be
33:58really careful when you're doing studies
34:00like this you're not claiming anything
34:02about a causal impact you're just sort
34:04of trying to lay out a set of facts that
34:08everyone should find important if what
34:10they're interested in is the question at
34:12hand in this case all firms choose their
34:15capital structure so let me give you an
34:16example of what I mean by that and and I
34:19should point out I should have said this
34:21earlier that you know I'm just using
34:23representative examples all the way
34:26through it's not to say I think these
34:27are the very best papers they're all
34:30good papers I think but you know they're
34:32representative examples the type of
34:33research that the research approach is
34:36that I'm trying to get at i but the
34:38other thing I want to point out is that
34:40you'll notice that every one of them has
34:43been published in a top journal right so
34:46it's not the case of top journals only
34:48published certain types of research they
34:50publish all types of research if this
34:52research conveys information that that
34:56referees and editors think are is
34:58particularly useful to know so let's
35:01think about this D'Angelo role paper
35:03that that came out in the Journal of
35:05Finance a few years ago and so sort of
35:08the background of this paper is that in
35:11general is there's this view in the
35:13capital structure literature that
35:15leverage is relatively stable over time
35:18for most firms right so if that's what
35:22you believe then that the implication of
35:25those that researchers need to focus
35:27their efforts on understanding the cross
35:29section explaining the cross section of
35:32leverage choices and all they're really
35:34doing this paper and it's this is a
35:36deliberately short description of their
35:40empirical approach because it's a really
35:42simple idea of what they're starting
35:44with it's not simple in the end what
35:46they're doing but it's very simple in
35:49terms of what they're studying they're
35:51really just trying to describe the with
35:54information in leverage ratios - I'm
35:58trying to say is this general view
36:00correct or not if the answer is no as it
36:04turns out to be no all right then what
36:06does that mean there should we look
36:08further and then try to figure out what
36:10does that mean for the fundamental
36:12question that we're trying to get out
36:13which is how firms should choose their
36:16their their leverage ratio right so so
36:20what are they fine my three main
36:22findings I would say in its paper right
36:24one is that there is substantial
36:27instability in leverage ratios for
36:30individual firms over time right it's
36:33not the case that they choose some
36:35target leverage ratio and just keep it
36:37there through time if you look at their
36:39actual leverage ratio and these are both
36:41booked and market it's not just market
36:44prices and their mean leverage ratios
36:45not leverage ratios are bouncing all
36:47over the place for individual firms
36:49through time so the basic idea of
36:52leverage stability seems to be incorrect
36:56right secondly what they find is that
36:59when it is stable for individual firms
37:02it tends to be during periods where
37:05those companies have really low leverage
37:07my MA times zero no debt at all in their
37:11in their capital structure I am third
37:14when they depart from periods of
37:17stability they came to do so in a way
37:20that's pretty strongly correlated with
37:24company expansion and contraction right
37:27and they're getting a part of that with
37:28some additional case based sort of
37:31evidence right so if you think about
37:34what they're doing the basic approach is
37:36pretty simple right the results that
37:39they have are purely descriptive in
37:41nature right which you might think if I
37:45describe to study that way to you that's
37:47it you'd say well how does this ever
37:49appear in the Journal of Finance how
37:51would I ever get a paper like this in
37:53the Journal of Finance well the
37:55the way you get into the general finance
37:58is that you make it clear that the that
38:01the results that you found have pretty
38:03strong implications for how we should
38:06think about capital structure and I've
38:08listed out three primary implications
38:10that they talked about in their paper
38:12that when you think about them you know
38:14that certainly seems important enough
38:16that the Journal of Finance should want
38:18to publish this particular result right
38:21one is that if we're going to start
38:23thinking about credible theories of
38:25capital structure then they're gonna
38:28have to explain not just the
38:30cross-section but they have to explain
38:31why there's so much time series
38:33variation as well right and you know as
38:36these two authors and I think they're
38:38correct they're you know look at the
38:39literature the literature hasn't said
38:41that much about the time series
38:43variation right so that's a big one I
38:46think secondly another big one is that
38:49it appears from their results that over
38:53a pretty wide range of leverage choices
38:56leverage per se seems to be of
38:58second-order importance to value right
39:02it looks like the firm's behave as if if
39:05they move their leverage around a lot
39:07it's not really changing their value
39:09much all right so it's not you know the
39:11standard inverted u-shaped curve that we
39:15draw for our students we talk about
39:18optimal capital structure in corporate
39:20finance but rather it looks like it must
39:22be almost kind of flat my over a wide
39:25range of leverage value doesn't seem to
39:27move that that much my and third it
39:31would appear then the main determinants
39:33of observed leverage ratios must be
39:35factors other than those that we
39:36traditionally think are important
39:38because those would have generated a
39:40curve that looks very different from
39:42from the curve that seems to exist all
39:45right so again when you think about that
39:47way you think well those those seem like
39:48pretty big implications and so knowing
39:51this set of descriptive facts is really
39:53important to our understanding of
39:55capital structure and therefore this
39:57paper rightfully gets a fair amount of
40:01attention even though you know there's
40:04no identification whatsoever there's no
40:06real sophisticated econometrics in this
40:08page
40:09for at all okay
40:12a fifth type of study that were again we
40:15don't see as much as we did at one point
40:18in time I was so-called clinical studies
40:21right so what I did I went back and
40:23collected from a jfe editorial in 1989
40:27what they described as a clinical study
40:30so to give you a little bit of
40:31historical background since none of you
40:33will remember this I don't think that
40:36the jfe a lot of time that Michael
40:39Jensen was leaving University of
40:41Rochester going the Harvard Business
40:42School he sort of found religion in
40:44terms of case studies and thought law
40:45okay we can learn a lot from case
40:48studies whereas before he didn't think
40:50that way and so he thought well maybe
40:52the jfe for which he was still the
40:54managing editor ought to start
40:56publishing nice alright and so they sort
40:59of put out this editorial describing
41:01what they viewed as a clinical study and
41:03what was it basically inviting the
41:04profession to to do this this kind of
41:07work if they thought it could be useful
41:10all right so what is the what is the
41:13clinical study and so you I mean you
41:14could read this read this for yourself I
41:17won't read the whole thing but but I do
41:19want to highlight a couple of things one
41:21is what I italicized these are really
41:25in-depth analyses of important
41:27dimensions of a phenomenon right so
41:30there's two aspects to that I think that
41:32are important one is the in-depth part
41:34right with a clinical study or a study
41:38of an individual company or a small set
41:40of companies you can go in a far greater
41:42depth as to what really was going on in
41:46the decision-making process and you
41:47could in a large sample study and maybe
41:49that would be useful to our
41:51understanding I and secondly there'd be
41:56important dimensions of a phenomenon
41:58which I which is again two aspects to it
42:00one could be you can get in depth on
42:02that on that phenomena but secondly you
42:05may be able to identify phenomena that
42:07appeared to be important to companies
42:09that we in the profession had not yet
42:12understood to be important and that's
42:14another way that case studies good could
42:16be useful right now highly also the last
42:19sentence the fact that
42:22that these could be potentially useful
42:25for a both empiricists and sort of guide
42:28their empirical analysis but also for
42:30theorists in terms of thinking about
42:32what factors are important and what
42:34factors are not because these are being
42:36drawn directly from the world and as
42:38such these calculations
42:40characterizations might be more
42:41productive than if they simply came from
42:43the researchers imagination so there's a
42:45little bit of a dig at the profession in
42:47the sense that we're just kind of making
42:48stuff up yes to what might or might not
42:51be important maybe we look at the real
42:53world to see what what appears to be
42:55important to the real decision makers
42:57now obviously if you're gonna do a
42:59clinical study you're gonna have two
43:02primary problems right you can only do a
43:04clinical study on a small set of
43:07companies maybe as little as one could
43:10be one case that you're doing an
43:13analysis on so you've always got a small
43:16sample problem that will then lead to
43:18some questions of external validity you
43:20say well sure that happened in this one
43:22case what why should I believe this this
43:26is this is sort of a general result and
43:28the answer is you shouldn't necessarily
43:30believe that it's a general result if
43:32you're going to do a clinical study I
43:34think it has to do it has to be done in
43:36one of two ways right one is the way
43:37they're talking about here in this
43:39editorial is that you're gonna try to
43:42illuminate something that we just didn't
43:45know about as a profession and by going
43:48in depth in a small set of observations
43:50you sure to say well in this particular
43:53situation this is what was important
43:56right so if that might be important in a
43:59more general sense than maybe we should
44:00think that way in terms of writing
44:02theories or that think that way in terms
44:04of how we construct some empirical test
44:06and see whether it's true in being a
44:10larger setting or not alright so that's
44:15one way you know second way you do it
44:17and the way I'm going to talk about is
44:18that I think this is a useful way to
44:21supplement some of the traditional
44:23studies that we do particularly those
44:25like your traditional lll so or panel
44:28regression sort of approach that suffer
44:31from your classic identification
44:34problems all right so we we
44:36identify some correlations in the data
44:37we don't know whether they're really
44:39causal or not so by simple regression
44:42techniques right but if we can look at
44:44things in a more clinical way for a
44:46smaller set we do know in those cases
44:49once you get into a sort of a case based
44:51situation you do have virtually perfect
44:53identification you can tell what caused
44:56what right so again that doesn't prove
44:59that that's true for the big sample that
45:01you're studying but it certainly goes a
45:02long way towards convincing the reader
45:05that it's clearly possible that that's
45:07what's going on in the larger sample
45:10data as well all right so I'll say an
45:13example of one of my own papers that did
45:15with Ken Lane and Lance fire on that we
45:18published in the jfets just last year so
45:21basically what we were doing here was
45:22studying capital structure decisions of
45:25US firms in this period of 1905 to 1924
45:30now why did we choose that period
45:33initially we chose that period because
45:35we realized that's when corporate income
45:37taxes were first introduced in the US
45:40and we thought oh this could be really
45:42cool experiment I you know ignorance
45:45about history we are forgetting the fact
45:47that well World War I want happens
45:49around that time as well and baby that
45:51was important too and it turns out it is
45:55quite important in terms of investment
45:57opportunity so you have this very large
45:59but transitory shock to the investment
46:02opportunity set of these US firms and so
46:05now we have two sort of factors that are
46:07arguably pushing capital structure in
46:10different ways and so what do we find is
46:13that well first of all it looks like the
46:15tax part has very little impact on the
46:17leverage choices right and we find
46:20pretty strong correlation between
46:23leverage and an investment and negative
46:27correlation between leverage and cash
46:29flows so it looks like the evolution of
46:32the investment opportunity set and the
46:33cash flows what's really driving the
46:36leverage ratios but we don't have
46:38identification there in any way what
46:40what so up might so one solution to that
46:45is to go the clinical realm
46:48right because not only do we not have
46:50identification we do have some other
46:52events around that time that could have
46:54been the causal determinants of why we
46:57saw these leverage changes that we did
46:59such as the the panic of oh seven the
47:01Federal Reserve got created in 1913
47:04there's a big depression following the
47:06war all these things could have been the
47:08main the main factor that was pushing
47:10these leverage ratios around so so what
47:13do we do so since we had a relatively
47:15small sample I think 57 observations in
47:19this case of publicly traded US
47:20companies we're able to look for each of
47:23those 57 at in a reasonable detail some
47:27clinical evidence of that that provided
47:30a direct link between the investment
47:33shock of World War one and investment
47:36itself and leverage alright so how did
47:39they actually finance the investment
47:41shock that came about because of the war
47:44and was that the reason that leverage
47:46changed during that period of time a
47:48second part of that is that we were able
47:51to see how the companies sort of use and
47:54built financial flexibility so that
47:56they'd be able to take advantage of
47:58opportunities that might arise such as
48:01this one so so how do we how do we
48:03depict this I don't expect that you can
48:04really can read that especially in the
48:07back but this is always going to be a
48:08challenge if you're sort of using
48:10clinical evidence to sort of supplement
48:13your baseline descriptive correlations
48:16sort of sort of evidence as well how do
48:18i summarize a clinical evidence in a
48:21reasonable way and in a paper and what
48:24we did is that we sort of put together
48:26this particular paper in which I'll just
48:29point out what what we have here is that
48:31we've got the set of companies and this
48:34is a set of companies that the biggest
48:36leverage changes you know what was their
48:38leverage at the start prior to the war
48:40how high did the leverage you get during
48:42the war and then how much did it go down
48:45following the war and then we have two
48:47columns they're sort of the the case
48:49study evidence that says well what
48:51actually happened to cause their
48:53leverage to go up there in the war and
48:56what happened when it came down
48:59following the war so we can provide
49:01this direct link between the investment
49:03shocks of the war and the leverage
49:06increases and then the actual actions
49:08that cause the the deal everyn
49:11deleveraging so we provide this as sort
49:14of a summary table right that that gives
49:17the overview and then usually why I
49:19think you would do in what we did was
49:20create in your net appendix that has
49:22very detailed case case descriptions
49:25about everything that that's happening
49:27to the supplement this right so again
49:30this is not I don't think if you just
49:33did this evidence by itself it would
49:36necessarily carry a paper right but if
49:39you've got correlation of it it's not a
49:41fairly large sample and then you can
49:43supplement it with case based evidence
49:45that can really get more decisively at
49:49the identification issue then you have a
49:51much more convincing package of results
49:53that you can you can convey some
49:56information with okay last one I'll talk
49:59about then we'll have some time for Q&A
50:03is a structural estimation approach
50:05right and again this is going to be a
50:07totally different approach than some
50:10basic exploratory data analysis and it's
50:12certainly not a natural experiment
50:13either all right so let me quote Tony
50:16Whitehead here since she's sort of the
50:18the the master of the structure
50:20estimation approach as to what it's
50:24trying to do and this is from Tony's
50:25presentation from several years ago and
50:27exactly one of these tutorial sessions
50:30all right what are you trying to do with
50:32structural estimation right you're
50:34trying to first of all to fit some sort
50:36of optimization model directly to the
50:39data and assess the quality of the fit
50:42so you can ultimately identify
50:45parameters that that govern the
50:47technology the preferences and the
50:50institutional features right so
50:55ultimately you're saying whether the
50:57optimal decisions provided by this
51:00particular model that you've put
51:02together resemble the actual decisions
51:05by the firm so do you think of this as a
51:08combination of a theoretical model and
51:11an empirical test because it's is using
51:13actual data and sort of calibrating
51:15to some features of the data well how do
51:18how exactly do you do let's think about
51:21one example of this from the capital
51:24structure literature this Tony's paper
51:25with Harry and Linda D'Angelo alright so
51:28in this particular model what the firm
51:31is doing is choosing endogenously its
51:34level of investment they call it K in
51:36the model its amount of debt or net debt
51:39which they call Pete so debt minus the
51:41cash and they're choosing those two
51:44things to maximize the value of the firm
51:47which is the present value of all the
51:49future cash flows all right without
51:52value is a function of not just the
51:54investment and the financing but also
51:57these shocks to the investment
51:58opportunity set that they call Z all
52:02right
52:02so it's and that's the way most of these
52:05structural models will work is that
52:08there'll be some value function that is
52:10being optimized and the values generally
52:12give me the present value of the future
52:13cash flows of the enterprise what's
52:16going to what's going to differ is what
52:19you think the key drivers of that value
52:21are all right and secondly what I'm
52:23calling the key ingredients of this
52:25particular model in this model right
52:27you're going to be governed by the
52:29sources and uses constraint which I hope
52:31every model would have alright but then
52:33they're gonna have to specify something
52:35about the cost functions for raising
52:38capital in its different forms right the
52:41way they do it is to say well there's no
52:43issuance cost for debt all right there's
52:46a convex equity issuance cost function
52:49right so the cost of equity is going to
52:51be pretty high for reasonably large
52:53amounts of equity issuance cash balances
52:56are going to be cost so you're not gonna
52:58be zero cost there's gonna be some cost
53:00to holding cash so it's never gonna be
53:02the case that you can just you know
53:04build up your cash balance and be able
53:06to finance everything that way all right
53:08and you're gonna be limited in the
53:10amount of debt that you can take on
53:12right so the way they're gonna they do
53:14it is to simply say there's a maximum
53:16amount of debt that you can take on an
53:19implicit in that is that if you go above
53:21that you've got infinite distress costs
53:24right so with those key ingredients in
53:28mind and those are
53:29they're gonna be key to the implications
53:31of the model what do they get out of
53:33that model so with that sort of setup
53:36they can say we if the firm is going to
53:38be optimizing its value across the
53:41choice of investments and financing
53:43right then we can say something about
53:46how leverage is going to change through
53:49time was the dynamics of the leverage
53:51ratio conditional on this random set of
53:55investment shocks is going to come in
53:58right how is the leverage ratio going to
54:03change through time or vary across
54:06different equity issuance and the cost
54:09of cash balances how important is the
54:13variation of investment shocks in the
54:16extent to which those are correlated
54:17through time right and how will leverage
54:23adjust toward this target through time
54:26will this be a rapid adjustment or time
54:28will this be a slow adjustment three
54:30times these are going to be direct
54:34implications that can be derived
54:36directly from the model conditional on
54:38that setup of them all that I gave all
54:40right then they can all take this they
54:43can take these all to the data and see
54:45whether the model seems to explain what
54:47we actually observe pretty well or not
54:50all right so if the model was set up
54:53perfectly then they could say well then
54:56these are the factors that actually
54:57matter in practice if it looks like they
54:59don't matter in practice it could be
55:00because maybe tomorrow wasn't set up
55:02exactly right they've left out some key
55:05ingredients for example but that that's
55:08sort of the limitation of the approach
55:10made it's very precise in some ways i
55:14but tied very closely to the setup of
55:17the model itself it could be the case
55:19that the model is oversimplified meaning
55:22the virtue of the structural estimation
55:23is that there can be lots of moving
55:26parts in practice that influence the
55:28leverage ratio for example and you're
55:29trying to model all those together
55:31that's great night but in order to be
55:34able to actually solve that problem
55:36you've got to simplify it to a certain
55:39extent and perhaps you oversimplify it
55:41and you left out some key ingredients
55:43right and if so the model is going to
55:46have less predictive content as as a
55:48result
55:49alright secondly it's if you're worried
55:52about it identification to begin with
55:54you should still be worried about
55:55identification now you're not really
55:57getting identification from the
55:58structural estimation approach I got all
56:06kinds of things going on there and then
56:09the third part is that these are
56:10technical technically very challenging
56:12estimation estimation is really
56:14difficult knowing what's the right
56:16likelihood function I one of the right
56:18moment conditions to set at the
56:20beginning these are not simple things to
56:22necessarily get done but they're doable
56:24all right and so this is a yet another
56:27way that we can try to get at the
56:29fundamental question of what factors
56:31really matter in how firms are choosing
56:34the capital structure or whatever is the
56:35fundamental question that that you're
56:37trying to get at all right so let me
56:40just try to wrap up with a couple of
56:41closing thoughts here and then let's
56:43open it up to any questions that you
56:45might have all right so first of all I
56:47reiterate again and I think this field
56:49is full of interesting important
56:52questions corporate finance is a very
56:54dynamic field it's changing all the time
56:57there's a lot of interesting stuff going
56:59on I think it's very worthy of our time
57:01to investigate these questions and so as
57:04I said being I'm glad to see so many
57:06people here who seem to be interested in
57:09these questions I but doing research in
57:13this in all fields it's not a
57:15one-size-fits-all endeavor right there's
57:18lots of different approaches that can be
57:20used to answer very different sorts of
57:23questions they're all useful approaches
57:26right none of them is perfect but they
57:30all have their virtues well they all can
57:32provide information that can be useful
57:33to the profession and I would encourage
57:36you sort of let the research question
57:39that you're asking
57:40guide the approach now I cringe when I
57:44see things working the opposite way
57:46people say well I can see this great
57:49data set out there only sort of retrofit
57:51a question to this data that's usually
57:54dangerous all right let's start with the
57:57what's an interesting question and then
58:00think about what's the best way I can
58:02think of to answer that particular
58:04question all right if we folks and a
58:08certain one I worry a little bit about
58:10with uh with with our attention to
58:13identification in the profession now is
58:15that I think I mentioned before that I I
58:17fear sometimes especially younger
58:20scholars feel like well if I don't if I
58:23really don't have good identification in
58:25my study I really can't do that sort of
58:28study I'll never get this study
58:29published and net result of that is is
58:33potentially damaging to our profession
58:35in the sense that I think it really
58:37narrows the set of questions that you
58:39can really ask and we're missing a lot
58:42of information if that's the approach we
58:43take as a as a profession so I encourage
58:47you to think beyond that even though
58:49there's plenty of room and a lot of
58:51usefulness for approaches like natural
58:55experiments that that do get at the
58:57identification question I'd encourage
58:59you to think about combining different
59:02empirical approaches in the same paper I
59:04mean I think you often we get we get the
59:07blinders on and we think okay I'm doing
59:09this study this way I but you know
59:13there's nothing that prevents you from
59:14say I'm gonna do this law it's a large
59:17sample more like a correlation sort of
59:19study but I can do I have a natural
59:23experiment on a smaller state could have
59:25some external validity problems that I
59:27can combine with that that bigger
59:29samples they're combined some clinical
59:31evidence to along with some descriptive
59:34analysis now you just have to think
59:36through well if I do this additional
59:38thing in this additional way is this
59:41going to add something useful to my
59:42paper if it's gonna make the paper more
59:44convincing why wouldn't you do it it
59:47seems like it if we learn something you
59:49want to do it alright and then lastly
59:51you know don't be afraid to point out
59:54the limitations and the caveats to the
59:57approach that you use there no way is
59:59perfect I haven't seen a perfect
1:00:02research study yet they all have their
1:00:04problems and I think sometimes
1:00:07understanding and articulating what the
1:00:10limitations are
1:00:11goes a long way to sort of convincing
1:00:13referees and editors that hey you know
1:00:16you know what you're doing and you know
1:00:18what you can say you know you can't say
1:00:19it's when you try to push it too far
1:00:20that you get this pushback you say oh
1:00:23they must not know if you're doing so
1:00:25therefore I'm gonna reject this paper
1:00:26and so don't be afraid to point out
1:00:29what's limiting as long as you can
1:00:31convincingly argue that you do have
1:00:33something that's useful then there
1:00:35should be room for in the profession all
1:00:37right so let me stop there and I'm happy
1:00:40to take any questions
1:00:44yep nope my students would agree or not
1:00:54[Laughter]
1:01:22yeah I still and I in some ways I
1:01:25hesitate to do but I always start with
1:01:28teaching them sort of a static trade-off
1:01:31sort of approach but the class I teach
1:01:34now is purely cases so they've already
1:01:39seen sort of the fundamentals in their
1:01:41finance one class and so they're
1:01:44allegedly they understand the basic
1:01:46theory or the static trade-off view but
1:01:48I still go back to it I tell them well
1:01:50so this is what it's dad's bonala so
1:01:51let's let's look at case settings in
1:01:53which things like financial flexibility
1:01:56might be important which introduces this
1:01:58notion of dynamics to it and so it's not
1:02:02you know I'm not trying to teach them
1:02:04theory in that sense but I do try to
1:02:08impress upon them some of these factors
1:02:11that look like as have come out in the
1:02:12research in the last ten years through
1:02:15specific cases I know if they think
1:02:18answering your question or not
1:02:26yeah oh yeah yeah yeah for sure well
1:02:33that's part of the difficulty with
1:02:35teaching cases in the first place and
1:02:36used to try to disabuse them of that
1:02:39notion from day one and say look there's
1:02:41ambiguity in real-world decisions and
1:02:44you're gonna be frustrated every time
1:02:47you do a case because it's gonna look
1:02:48like I have not given you enough
1:02:49information how it should make this
1:02:51decision I said well welcome to the
1:02:53world so it's going to be from here on
1:02:56out but that doesn't mean they like that
1:02:58aspect there they are frustrated by that
1:03:00you're right yeah absolutely right
1:03:07anything else yeah
1:03:18I actually at the great point and I
1:03:41actually think we've kind of gotten
1:03:43there already in in a lot of our fields
1:03:46a lot of our subfields within corporate
1:03:48finance because there are a lot of
1:03:50studies that yeah the first time you see
1:03:52the instrument used and you think that
1:03:55it's really clever that's a great
1:03:57instrument no that's I buy this
1:04:00instrument and then you see a paper
1:04:03six months later that's using the same
1:04:05instrument for something else and you
1:04:08know you think well that paper ought to
1:04:10be rejected because immediately the
1:04:12exclusion restriction must not apply so
1:04:16I think we've gotten there already
1:04:19honestly and because of that there is a
1:04:23limitation on how far you could go with
1:04:25that type of study yeah I'm not sure
1:04:29what else to say about that
1:04:31[Music]
1:04:48[Music]
1:05:28yeah there's a couple issues there one
1:05:31is that I think in any of these studies
1:05:34and this is more important than ever I
1:05:37think given the the size of our data
1:05:41sets that we work with now is that
1:05:42economic magnitudes are hugely important
1:05:46so you know it's it's not hard in a lot
1:05:51of settings to get statistical
1:05:52significance given the size of the data
1:05:54sets so economic magnitudes are really
1:05:58quite important the other aspect to what
1:06:01you're saying though I think speaks to
1:06:04what you can and cannot conclude from
1:06:06some of these different empirical
1:06:08approaches like I used the example of
1:06:10natural experiments trying to get at the
1:06:12importance of taxes for capital
1:06:14structure and I think first of all
1:06:16logically you I think also think of
1:06:18course taxes should matter and you do
1:06:21natural experiment and yes they do
1:06:23matter and in the direction that you
1:06:26think they would but the real bigger
1:06:29question I think is are they a seriously
1:06:33important determinant of what the
1:06:35leverage ratio ultimately is the answer
1:06:38seems to be no and that's getting out of
1:06:40your issue of you know are you really
1:06:42increment incrementally explaining much
1:06:46of the r-square with this particular
1:06:49variable an answer seems to be no now
1:06:51you could do it directly like I think
1:06:53you're talking about or or indirectly
1:06:55with some of these exploratory paper
1:06:58like one I didn't didn't mention there's
1:07:00a paper by John Graham's with mark Leary
1:07:03and Michael Robert so they look at
1:07:06capital structure over decades in the
1:07:09u.s. in which there is huge changes in
1:07:12the tax code during that time and you
1:07:15just don't see leverage movie with the
1:07:18tax code and the way you think it would
1:07:20not you know what one obvious inference
1:07:22I think is that all taxes just don't
1:07:25look like they're they're early
1:07:26first-order so I I agree with you
1:07:29totally I think that is ultimately what
1:07:33you're trying to get and simply finding
1:07:35statistical significance on a variable
1:07:37is not enough to really tell us
1:07:40what's important and what's not it
1:07:42should be it's a first step
1:07:44it better be statistically significant
1:07:46in that kind of test but that's not
1:07:48enough to say it's really important you
1:07:57know I'm not I'm not sure we can move
1:08:12more than we already have now I think we
1:08:14have moved quite a bit already it's an
1:08:17interesting question though because I
1:08:19don't know if you saw you know typically
1:08:22the day after the Nobel Prize is awarded
1:08:25they'll be in editorial Wall Street
1:08:27Journal from somebody else who's sort of
1:08:30seen the phrases of the Nobel laureate
1:08:32and why their work is so important and
1:08:35this year was different in the sense
1:08:37that there was an editorial that was
1:08:38basically saying I wish they had done
1:08:41something bigger than what they did and
1:08:43their criticism is is a bit like what I
1:08:46was talking about earlier is that in
1:08:49that sort of type of experiment that is
1:08:51trying to approximate a randomized
1:08:53controlled experiment you have to narrow
1:08:56the question down quite a bit and so you
1:08:59can get the marginal effect of this one
1:09:02variable but you might be missing the
1:09:05bigger picture as to what's really a
1:09:06primary driver and my personal view is
1:09:12that you know both both of these types
1:09:15of empirical approaches are are very
1:09:18important in our field and I would hate
1:09:20to see us moving too far in the
1:09:22direction of one versus the other
1:09:24because they both provide very useful
1:09:26information that we should want to know
1:09:28and so too I you know I don't think
1:09:34we're gonna necessarily move further in
1:09:36that direction in our field because only
1:09:38because I think we have moved quite a
1:09:40bit in that direction already and we
1:09:42should
1:09:48yep yeah I think that that's a good
1:10:00point
1:10:01aye aye so far anyway I don't think of
1:10:05those things as necessarily a different
1:10:08empirical approach as as much as a
1:10:12different way of getting the data that
1:10:14we can implement in these empirical
1:10:16approaches but I already think those
1:10:18those two have made you know I've made
1:10:22progress tremendously more efficient in
1:10:24in corporate finance studies so far and
1:10:27I fully expected that will continue
1:10:38yes all right well thank you very much
1:10:43for coming appreciate
1:10:49[Applause]

About this transcript

This page contains the full transcript of Conducting Empirical Research on Corporate Finance by Financial Management Association International, generated from the public captions YouTube serves with the video. The transcript has 11,478 words across 1,671 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 translate the transcript into 100+ languages.

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, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.