Illuminating the Corporate Governance Black Hole: Contextualizing the Link to Performance

okay this is joint work with Ron Gilson who's here and Darius Talia from the Rutgers Business School so I'll just start with a little bit of background I think there's a general belief probably shared by most maybe everybody in this room that there's a connection of some sort between corporate governance structures of firms and their capacity to create value for shareholders and obviously we all also know there have been significant empirical efforts to try to confirm that relationship including work done by people in this room one of the important element one of the important approaches being the index studies where the index author posits a group of characteristics governance characteristics which they believe represents good corporate governance and up a score and then test that to see whether as an impact on some measure of corporate value such as Tobin's Q and the two best known ones are the are the Gompers index the G index and the one done by Lucian and his co-authors the the e index and they each find a highly statistically significant relationship between these index scores of and Tobin's Q so our objective here is to try to piece out a little bit more of why we see this relationship and that's that's an important question because these studies have been criticizing including by by Michael Klausner who's here in Catan and kahan who basically say they see no theoretical reason why we should see this relationship and what we want to do here is suggest to sort of try to begin to see reasons for the relationship and suggest and empirically confirm that signaling plays plays a a role here and the underlying idea here I'll dig into it a little deeper but that good managers choose structures that get higher scores because those structures indicate to to the market that they are higher quality managers beyond that we want to kind of make simply the larger point that governance is contextual because as we'll see we see that this phenomena seems to be stronger under some circumstances then and others and it suggests that these linkages are between governance structures and performance are more contingent and contextual and the directions of causation more varied than maybe people typically applaud so what what are the possible explanations for for seeing this relationship between good scores and higher q well one of them might be right that better governance structures lead to better corporate managers that that in essence the these features overtime filter out bad managers another explanation again kind of going from structure to value would be that regardless of the quality of the managers that managers who are operating in a better scored structure would do a better job they have a more incentive to do so or they're better monitored or they're getting better information or whatever the third possibility the one that we end up trying to test is that better managers lead to choose better structures the market sees that choice of the better structure and infers that the managers are better and that inference leads to a higher q that basically a change of structure is a signal of market quality so this third explanation gives rise to our tested hypothesis so let me just flush the hypothesis out a little bit more you kind of start with this proposition of there's an information asymmetry that managers know more about their quality and the market does that then that a change in governance structure which could represent for example a change in management's exposure to capital market discipline is a signal to the market of management quality and obviously the signaling Theory requires that you rule out cheap talk that somehow the choice of a better rated structure is more costly to poor poor managers than good ones and and this is what we'll try and take advantage of to see whether this is at work we want to look at situations where that asymmetry seems to be larger and we try and do this across time examine a compare what we might call normal years to the a period where we think the asymmetry was unusually large the period of 2000 and 2002 the period of the accounting scandals the paper has considered considerable a qualitative evidence that that people were surprised as one shoe after another dropped of companies that were highly respected people discovered that their accounting in fact was fraudulent or defective and that there was then a drop in trust in terms of the rest that there was some kind of pooling equilibrium in which people were more suspicious of the companies wondering whether there were more shoes to draw and then also in terms of trying to find variation in asymmetry across firms whether using evidence that firms with a lot of R&D that there tends to be more asymmetry of information associated with those okay so what's the empirical strategy well again we what we do is examine the normal periods nineteen in terms of this time difference nine the normal periods would be 1992 through 99 and 2000 three through six versus the scandal period and then we examine the difference between the two both using OLS regression and fixed effects or aggressions taking advantage of the fact that the fixed effect regressions only look at firms that change where we think that the change in structure is the signal so I just you know basic basic factors about the regressions the OLS regression the structure is the independent variable Tobin's Q is the dependent variable and we add ten family French industry controls for the for the fixed effects we look at change in governance scores and then also add change in a variety of other things that have been found to affect Q such as these these items here sort of summary statistics I'll skip by that our we first ran these regressions both OLS and fixed effects for the full sample period and like the earlier studies found highly significant statistical relationship between scores and inq the OLS figures now come the slides that are that are key to our findings we then divided and looked at the 2000 to 2002 period and the normal period now in each of these for both the G index and the e index we found a statistically significant impact of with with fixed effects using the fixed effects regressions of structure on on cue but importantly for each of these we found a highly statistically significant difference in the impact between the scandal years and the normal years but the impact was much greater during the scandal years now on the other hand for the OLS results we found again that highly statistically significant impact in both the scandal and the normal period but the difference between the two was very small far from being statistically significant so you know just to summarize then what we have found we found that for the full period high statistical significance ain't reached the sub periods that the difference in the fixed effects between the scandal period in the normal period was highly statistically significant and the difference in the OLS results between the two periods was not close to being statistically significant you know is this of some economic significance well it looks way for example if we took a firm Gompers that all kind of lumps the really the bad firms as they called it a dictatorship portfolio the good firms as a democracy portfolio if we went from a 14 score to a nine the lower the score the better the governance is rated 14 is kind of the best of bad firms 9 is the worst of good firms in the scandal period that would result for a median firm of one hundred and thirty two percent increase in q4 in the normal period only a thirty eight percent increase in Q similar kinds of results with the e index scandal thirty one percent normal only thirteen percent so what's our interpretation well again we first you know pay attention to the fact that the fixed effects results relate just affirms that change governance structure in a given year and we see this big difference in the changes impact on Q during the scandal period versus the normal period why might that be well one reason could be the market thinks during the scandal period that good management's more important or that these first two explanations of structure leading to better performance the filtering or the incentives of information work more effectively during that period of time and or that a change in structure sends a stronger signal concerning management quality during the scandal period well our OLS results remember relate to all the firms in the sample the substantial majority of which do not change in a given year or only a small minority do and the lack of difference between the scandal and the normal period OLS results suggest that the market is not thinking during the scandal period that these the causal effects from structure to performance is more important working better or whatever seems like the market still thinks that for firms that are continuing with the same the same structure that they're making the same inferences about what value will be and that seems to rule out the first explanation leaving the signalling explanation so you know the answer I think is that the change in governance structure itself conveys information to the market when information is more valuable in other words when it there's less good other information we see a bigger impact from this signal the change is a signal according to signal theory because the better rated corporate governance structures more costly to bad managers now a few robustus tests we change the definitions of the normal times two periods much shorter period before and after we get the same kind of results another robustness test we look just at the impact of adopting or or or or eliminating a staggered board or poison pill and the difference between the scandal period and normal period is highly statistically significant between these two periods when when we look at these Changez same thing with poison pill so it's consistent with our interpretation the full results of the index court about firms with with and without R&D well we have reason to think firms with R&D are more opaque than firms without our OD when firms change their governance structure what we found was that firms with significant R&D have a greater change in Tobin's Q during the scandal versus the normal period than the change in Tobin's Q for the non R&A firms and that the difference again is highly statistically significant which provides I think further support for our signalling hypothesis this time looking for differences in asymmetry across firms rather than across time so summary of the results I think we find evidence in support of the managerial signaling hypothesis by seeing that we see this larger impact went opens when on Tobin's Q when asymmetric information problems with respect to managerial quality or higher either across firms or across time and our basic conclusion then is the signaling Theory helps explain the the index the index study results that I think is an important result in and of itself because it's important to know the mechanisms by which information asymmetries which are generally not a good thing can be reduced suggest why as an aside the sharp asymmetry increases when gatekeepers fail and then kind of a larger point that corporate governance is more contextual than I think the index studies which look at average firms over a long period of time might take account of we are beginning to try to take account of that and notice them that linkages with performance are more complicated and more contingent than often thought great to be here thanks for having you discuss this paper this is paper called illuminating the corporate governance black hole contextualizing the link to performance I thought we should talk about the title first I'm caught I don't I think maybe the physics here wrong illumination means you shine a light on something and then you'll learn more about it and you'll learn more because of the reflection of light but black holes specifically do not reflect light and so I don't know if you saw this it was a big deal recently that they made uh the astronomers made a kind of picture of a black hole and I took forever and you can see on the picture that they're not actually seeing the black hole the point the black host is an event horizon inside that horizon the gravity is so high the light cannot escape so we can never see anything so specifically illuminating a black hole is of physically unappealing anyway how we get the point so this is work on the important question on what the relationship is between governors and value right I think it started this may not be the first paper but early work involved things like looking at the relationship between Q remember Q is the total value market value of the firm's assets relative to the book value relating that to some measure of governance mork's life from Vishnu use the percentage of board ownership and I think there's an upward slope here although obviously the shape is quite interesting composition metric work is about looking our returns here's the one of the main tables yes is perhaps the main table of their paper where they say the difference in return over their sample period between the portfolio of world government companies that democracy and the portfolio of poorly government firms dictatorship has an f of 71 meaning the return is higher now the return and Q are not completely different things if you think about a change in Q you're you're changing this ratio of the market value of claims to a book value this is approximately the return minus the growth of assets and because there's usually more going on with returns especially in the short-run you can think that this is basically love Olsen this is changes pretty much okay and that means they're connected because if you have a lot of positive changes then your level is higher and in fact you can see this in composition metric that they have a lot of results about the Q over time the Q slope over time and that's going on all right so why is there a positive relationship let's recap some of the possibilities good corporate governance can reduce risk or raise growth once we're done our Q there's some things that we we're not looking for so we're not looking for things that make the firm bigger okay just proportionately bigger there's a lot of economics literature with the span of control stuff where you say a better manager is someone who can produce the same thing for a bigger set of assets and that's also how sometimes the people think about asset management is that the better manager is someone who can manage 10 billion instead of 1 billion ok but so that's not what we're talking about here because we have some kind of relative value measure but growth would definitely work so if welcome and farms grow faster than their future profits and cash flows are higher and the market is going to assign a higher valuation relative to current book it calls me that high-value races governance it's some kind of luxury good so we find a correlation but the correlation is reflective of the fact that when you're doing well enough you can invest in you know charity sponsorships and good governance and then you can have more complicated dynamic theories like her Melinda my sparkler high-value entrenches management and then maybe down the line you get bad performance so this kind of story is more complicated because you might have different correlations at different horizons alright so this paper says when market is the market is pessimistic about management quality that's the governance crisis then governance changes or governance in general is a positive signal of value so basically matters more if you have a good high value on the G index or the e index at particular points in time and the point in time is the crisis in 2000 2002 with Enron a world calm and so on governance becomes more salient somehow more value relevant and so at that point while going firms should have relatively high prices the empirics is to compare Tobin's Q slopes on either the G or the index in this period compared to the crisis the non crisis years which is basically the composition metric sample and a few year four years after and let's we kept the finding so you run a regression with the dependent variable as Tobin's Q okay you have an index here it's G here it's e the results here are essentially the same let's look at the G results so in the scandal period a high values associated with the low value the coefficient is minus 11% in normal peers it's minus two and a half percent that difference is statistically significant and this is using Rene's preferred method of calculating the t-stat on the difference okay so this looks good firm fixed effects in here basically the indexes both of them predict Q with a bigger coefficient a bigger magnitude in this period than in other periods okay so a couple issues with this of course the this peak 2000 to 2002 of the impact of GE it's very close in time to the internet bubble and burst and so we want to think a little bit about that in fact already the composition metric paper has a time variation through the 90s they call it a slope goes up so these are they don't have a figure like this in the paper but they have a table where they show the Q slope year by year so they run a cross-sectional regression every year Q on the index and then they reported in a table the year by year coefficients and here's what they look like and I what I added is a trendline so you see it's going up so if you like this is a known fact that through the 90s the slope is going up in that sense maybe the reversal is particularly significant because that's not a finding we we know at least not from composition metric I think it does suggest that as we're looking at the time series it'll be nice to have other other sources of variation think about other times maybe there's been other crises that we could use or at least crises for particular industries that we could use when we should also see increased significance of governance in fact there's a little bit of a reversal already I think known this is on the words website words is Wharton's depository for financial data and they helpfully provide 1991 to 2002 crisis basically comparison of two portfolios democratic and dictatorship firms basically the composition metrics farms and here's the evaluation to the starting point and I think 1991 the blue line here is the dictatorship portfolio the green line is the democracy okay what you can see is this may be a small difference but there's a massive difference that opens up towards the end of the sample this is the same increase in the slope the Q slope on the index the composition metric found and interestingly it collapses back and in a way this is what the author's that this fits the author's story because the Enron WorldCom governance crisis period is here and this is when the democracy firms are worth more and then afterward this kind of collapses back a little bit I think you can think I mean you can certainly quibble with exactly which farm is where and how do you treat firms to move from one portfolio to another and so on so I don't think this is necessarily the only way to do it but I think it's fairly obvious that the relative value can open up a lot in the late 90s and then close back down a little bit after that so you might think about other things that are happening at that time you know around 1999-2000 such as M&A you know we're worried that there's some governance core that governance is correlated with some price characteristic with time varying importance such as M&A okay maybe the probability of takeover bids is high in some periods and low in other periods if you're well governed maybe you're more likely to accept bids and then you're more likely to be a target and this kind of relationship between governance and takeover probabilities is discussed by Kramer's in 9 for example okay so how this is connected to M&A cycles one well in recessions there are no bidders there's no merger activity so anything that makes you differentially exposed to mergers is irrelevant in the short run close to a relevant in a recession but maybe important that in a boom and remember M&A is incredibly cyclical this is just something I grabbed off the Financial Times a couple of days ago it goes from 96 to 15 the orange line is number of deals the blue bars are deal values it doesn't matter which one you look at this 99 2000 period is very high than M&A volumes collapse but they come back up in five six and then they've been high since 14 again okay so suppose this has to do with I'm an ape takeover vulnerability or exposure or probabilities then we have a very strong prediction that the coefficient should come back up here and then again here okay so I think that kind of prediction is important here okay yeah so you might even think that they're cross industry differences because you have margin wipes by industry self involves mergers then of course industries exposed to merger waves should have different coefficients on the index then on margarine okay so I think it's reasonable to try to think about these kinds of tests and there are a few in the paper but I think a little more can be done so we have one striking fact which is that in valuations the value have on the indices is really more important much more important in this round millennium period than other times but in a way this is just one observation we would like to have theories that can make other predictions and we can test them as well and the authors do a little bit of this sake cutting up the G Index and picking or index and picking up particular components but we would like to kind of swim outside this pool look for other data and other predictions though I think yeah you might work yeah not accusing the authors of doing anything illegitimate here but at the level of the profession people study corporate governance empirically these indices have been extensively studied and so even though you have no questions about a particular result just because there's been a lot of papers in the same kind of data data area you should be more and more skeptical of no results or put differently this is a real high value if you can find predictions from your theory that are outside this more well-known data is there anything you can do you know collect some new data look in some other country and so on so I'm just trying to think of some things I think the changes in market valuation are gradual changes in governance can be abrupt sometimes and so the announcement returns could be helpful I'm not saying year-to-year changes in vision like something at the date when you announce that you're changing your governance at that time there should be as response it should be higher in that period when you think governance is more salient that is looking a little bit in a new the new data set okay you could use other governance metrics and I don't mean subsets of this and I mean anything else that you think is relevant from a signaling point of view or exactly whatever your theory says or of course are the countries now the G did G Index has no meaning in Germany presumably so you can't just you know extend the study to international data but you would have to look for something else say when there's a scandal in some other country whatever measures of governance are known to work in that country should be more significant around the scandal and I think that's a plausible theory I think it'll be nice to test it a little harder yeah I think we recognize which I suppose doesn't tend to be recognized that fixed effects sometimes also they have their time variant factors and therefore while we use fixed effects to control for in hidden variables that OLS doesn't reveal there's the potential for the reverse I'm not sure that the M&A activity going on and off should be that relevant one of the things we see in this period is that a lot of the governance changes that we see were firms that were adopting poison pills during that period of time so they were doing that even if M&A activity was was lo I mean obviously the other things we try to do in terms of the trend factors you're talking about is is have these short the shorter normal period and also examining this impact on R&D I didn't mention it in the slides but we also found just for the whole sample a difference between R&D intensive firms and other firms in terms of their the sensitivity of their cues to differences in corporate in in index scores that was not at a statistical it was it was not statistically significant but it I mean it was sort of at the 80% level and then we saw this result within the crisis period of a larger larger impact one one paper that I think is probably worth mentioning is the paper by M try and do this right Emmer amaryllis Lonnie and at all that shows that social ESG firms indict relationship between value and ESG firms which normally one doesn't see any relationship I think they were measuring access to credit markets that during during the financial crisis that suddenly ESG scores did help predict access to financial markets to credit markets which which again would suggest that there are times where things that don't seem to be such important signals other times can become important at other times whenever I take questions oh my god Oh Merritt I I was called on in empirical and to theoretical questions on your signaling model so you're talking about differences but fixed effects is not differences fixed effects it's related but it's not the same so I wonder if that's just a terminological if you have just used the word fixed effects but actually ran first differences or if you haven't then you I think you should run first differences because your theory is that the effect should be visible instantaneously signaling model will bring out the effect instantaneously so the first difference would be the right way and I mean we're doing average differences not first difference is this what you say yes yes that's what you're doing and so like if the effect shows up two or three years later your estimate if it's fixed effect still picks it up whereas first difference would not and I think in your case it shouldn't because your theory is that it's like it's a signal like the market sees right away and then on the theory side so signaling requires a cost as you said in your talk and but I think you also related this to like mics and other people's points that these governance things are really irrelevant and so if they're really relevant then there's no cost of adopting them so then it will break it will break down and relate it there I was asking myself the question what exactly does the manager game from signaling now it's very obvious if you're at the IPO stage you want a signal quality because you get more money for the stock you're selling but if it's midstream and your firms are midstream then there's only I think two way if I have are in the manager I know I'm a good guy I can just like be entrenched stay around for five years and then like cash out my stock rewards five years later by which time the information will have come out anyway so if I want a signal now it has to be because either I'm planning to raise equity for the firm in which case I need the stock price to behind now I'm not quite sure I should I think the only way I would want to raise the stock price now okay so on the methodological front I have to admit this isn't my strong suit Darius seemed to think it didn't matter so much but it seems to me we should try and check those things on the theory points yeah let's let's start with is there a cost so I mean I think I we try and deal with that in the paper in in a couple of different ways one by looking at some of the elements in the index and suggesting why we think there is a story as to and Lucian may may want to add to this story but why there is a story as to a relation why there would be a relationship first of all one of the elements is standard boards and and Michael certainly I think things stack or boards are relevant Michael then goes on to point out well if you have a staggered board some of the other things that are measured like supermajority clauses or ability to have extraordinary shareholders meetings which are scored in the index would be irrelevant but you know close to half the firms don't have staggered boards and they could be relevant in in those circumstances and firms that do have staggered boards in in for example 80% of the cases don't have supermajority clauses so you know you may have one or you may have may have the other I mean I think Michael's right that it would one could be more sophisticated in constructing these interactions but I think there's some kind of a story as for poison pills I certainly understand the logic of why everybody has a shadow pill if they don't have it but you know there seems to be some empirical evidence that having having pills does matter and I can imagine a situation for example where if a pill if a firm withdrew uphill then the fact that it doesn't have a pill would make adopting a pill later when a crisis happens with there's a proxy fight or something like that more costly to to to affirm beyond that I guess you know we think there's a story there is no story as to why there is no story as to why we're getting results we haven't heard a story as to why we're getting results when why we're getting results I guess I mean somebody can beat up on the story as to why we get results but there's no story as to why we grant results if this isn't true and I guess finally well gee if there's no cost and we've demonstrated you could improve a share price this easily why didn't why didn't everybody do it now do your final point you know I teach corporate finance too and I wonder about those things on the other hand I think in fact managers do seem to care day to day week to week months a month about what their share prices are and you know there is signaling theory associated with debt also so at least we have the you know we have the respectability in our theory that like you know other people think that managers might do this even if they're not necessarily going to do an equity issue any time any time soon yeah so focus made 80% of my point but I'll just carry it a little bit further at least one of my many points and that is the variation you're seeing with fixed effects are firms dropping their pill I mean adopting pills you've said that at the end it's it's that's driving your results it's overwhelming they're a very small number of classified boards adopted in those periods and they're all by very large firms so that's not doing any work it's all firms that are adopting pills for whatever reason so your signal which I agree is not remotely costly seems to be firms choosing not to adopt a pill and thereby signaling their good quality or firms adopting a pill thereby choosing to signal their bad quality neither which is costly so the signaling story is is really I don't think it's right I kind of like your title for that reason but now one more thing you did say that I that I you at the end and responding the hoager you said well look look at these results I totally agree with you look at the results in all of these things that that use the the indices particularly the GaN position metric and index they're implausible but they're there so I agree with you it is worth trying to figure out why on earth they're there but I don't think you've done [Laughter] just particularly the story which is but it's an invitation circumstances where to say the causation of we're not making the returns one we're we're not making the returns one that course that should be DuPont right we're making the slope points well we'll probably don't see very many firms that shift I guess I'm you know we don't see firms that have five factors change in years if it's signalling then the whole effect comes in at the time that you put out the signal in the market is efficient and so the best way to then test your theory would be to go on announcements of the changes that you are talking about and if you do that the evidence would be completely inconsistent with what you find here in the Sun that you find an effect of I mean more than a hundred percent virus that quietly in the trouble it means that you have no more return for Levitt form would be of the order of 150 to 200 percent which honestly we never found now on source the changes in those factors people have trouble finding evidence using advanced studies that there is a significant that fact so then he nobody finds an effect that is pen important so he said and so it's it's just not compatible with the way to interpret evidence as signaling I just don't see I mean I think you have to work with Santa the huge effect that you are going 2000-2002 with kind of the events can be evidence of changes in those aspects of parties after what signaling which has a positive field it reduces infernal symmetry the other is the managers are willing to drop take over the fences when they know that the future is good and therefore in the future they will not have to be afraid of a takeover in both of cases there is a facility but in the first one the market learned something in the second one just the researchers later on finds that they did something in situations when they had this information asymmetry now because you are focusing on long windows you are really testing the second theory the second theory is the one in which I'm willing to drop for the P because I know that in two years I've been good shape okay and you are not testing and in that case maybe you are confirming that theory but you don't find anything about market learning you're just finding maybe the market as finding that they're doing something and the researchers understand it and the only way to test the positive signaling hypothesis would be to look at the announcement and by the way here it's pretty unlike the classified boards where there is a long window until the market learns about it gradually in the poison pill they announce it in a given day when they do the poison pill so you could you could find it the other two things one Mike with his known modesty didn't mention his paper but you should we know from his work with Emiliano cotton that much of the firm fixed effect regressions in this area don't work very well because there are different waves that happen of firms of different size and once you control for that in different periods large firms and small firms moved or differently so for example if the first drop poison pill in the early 2000 are larger or smaller then mark Mike's work suggests that maybe all the effect is driven by how large versus small changes in value during this period and the last thing is you might want to look Alma China I have paper the jfe about learning about corporate governance and we find out we run 2002 this point in which returns stop being quoted with governance but Hughes continues but lots of things in the change and you want to see how those results relate because they happen around the same time like we find for example that analysts surprises about firms with good governance and bad governance change around the time things like that but in some sense that's right that's the end of that's the end of our peer right so I guess I could see looking taking our normal period as you know before and after rather than before and after but it seems like that makes a lot of sense that during a period in which there was a lot of uncertainty we then see a change and analyst surprises go down and I think I think you're saying something consistent with I have to I have to look at your paper but I think you're saying something consistent with our results in terms of at least of the period 2000 to 2002 versus after that but do I just so I understand the aleutian are you is your is your bigger is your bigger is your first point again the first difference is versus average differences know that if you are getting your results and let's suppose that they are accepted then what you're confirming is what I call the inside information hypothesis under which they drop takeover defenses because of inside information the future will be rosy rather than persuading the market hypothesis because that I pothinus should be tested by looking at returns in small windows that's the remain holger point at the announcement of the pill because you are looking at 2/3 like Cramer's you're looking at 2/3 your changes in queue what you are confirming if you get there if you if the results are robust you're confirming the inside information hypothesis and the second point is whether the results are robust or not you need to do the Catan and Klausner methodology of controlling also for the fact that there are large and small firms and they change in relative very differently in different for example I know that the years 2000 to 2003 the small firms fell much more than the large firms so just the fact if there is any difference between their tendency large and small firms to adopt poisoned piece that might fully drive your results in the same way that patent Klausner showed that the difference in the valuation of large and small because there were incidents of yeah I recognize the cue has come under attack but it is a measure and it does permit this this piece to be culpable with the earlier pieces thank you you [Applause]

Maurice Vega

Leave a Reply

Your email address will not be published. Required fields are marked *

Post comment