Guan’s blog

Estimating a firm’s market value of common equity (MVE) at a given point in time is something that a finance researcher does quite often. It can be surprisingly hard using only the two commonly available datasets for financial and stock information, CRSP and Compustat.

The most foolproof way is to use the prcc_f and csho fields in Compustat Fundamentals Annual file (or their counterparts prccq and cshoq), which are the closing stock price and the number of shares outstanding (in millions of shares) on the balance sheet date.

For example, here are the values and corresponding market value of equity for Apple, Berkshire Hathaway and the New York Times Company on the latest datadate in my local copy of Compustat:

I can tell you that these numbers are, at least in my opinion, correct. This method is fine if you happen to need MVE on the annual balance sheet date, or the quarterly balance sheet date for firms that are in the Fundamentals Quarterly file, but what if you need MVE at the end of a different month or some arbitrary date? For example, you may want to construct portfolios sorted by book-to-market, and since accounting information is not released on datadate, you want to make sure you are not using information that is not available to outsider investors.

The simplest way is, of course, to look in the CRSP Monthly (or Daily) Stock File. The PERMNO for Apple common stock is 14593, so we could multiply abs(prc) (because bid/ask midpoint prices are reported as negative numbers in CRSP, don’t forget that!) and shrout, shares outstanding in thousands. For Apple on 2010-09-30, this gives us 283.75×915,970=$259.9 billion, the same result as Compustat.

What about Berkshire Hathaway? Remember that Berkshire Hathaway has two share classes traded on the New York Stock Exchange, A shares (PERMNO 17778) and B shares (PERMNO 83443). We simply have to multiply price and shares outstanding for each class and sum them. The generic way to do this is to start with a PERMCO, get every security that has share code 10 or 11 (common stock), and multiply price with shares outstanding. Here’s sample SAS code for this, in this case hard coded for Berkshire, but you can figure out how to merge with a separate list of PERMCOs and dates (and properly use group by):

proc sql;
select
    sum(abs(prc)*shrout/1000) as mve
from
    crsp.msf m,
    crsp.msenames n
where
    '31DEC2009'd ge namedt and '31DEC2009'd le nameendt
and shrcd in (10, 11)
and n.permno = m.permno
and n.permco = 540
and m.date = '31DEC2009'd;
quit;

How come we had to use 4 different figures to calculate Berkshire’s market value using CRSP data, and only two figures with Compustat data? This is because csho and prcc_f are stated in terms of only a single class of stock. Berkshire Hathaway had 1.552 million class A equivalent shares outstanding, weighted by the relative economic interest of the different share classes, a number that the poorly paid S&P research assistants probably found in Item 6, “Class A equivalent common shares outstanding, in thousands” in the Berkshire 10-K. (Some companies make this a lot harder to find, especially on the balance sheet date rather than reporting date.)

Now if you do this for the Times Company on 2009-12-31, you would find only one share class on CRSP, PERMNO 47466, and a market value of 12.36×143,715=$1,776 million. This is almost $10 million less than what Compustat has. What happened?

If you looked very closely at the CRSP names file, you would notice that PERMNO 47466 represents class A common stock. Some companies have only a single class of common stock that is still designated class A, but the Times Company has a separate class B that is not traded on any exchange and is not on the CRSP tape.

So how do we estimate MVE for the Times Company at the end of January 2010, which is not in Compustat? My solution is to multiply the CRSP market value with the last known ratio of Compustat to CRSP market value. This ratio was 1,786/1,776=1.0056. CRSP market value on 2010-01-29 was 12.92×143,715=$1,858 million. Multiplied with our ratio, this gives an estimate of total market value of equity of $1,867 million. This estimate doesn’t account for the special voting rights of class B stock, but it is probably the best practical value we can assign to Times Company equity at market prices.

My adjustment may not be such a huge deal for the Times Company, but I can tell you that at the end of 2002, Ascent Media Group (PERMNO 88354, GVKEY 006948) had a total market value of equity, as implied by issues on CRSP, that was 11.7 times the market value of issues on the CRSP tape.

From a James Fallows correspondent:

Medieval Latin is a much simpler form of Latin than classical, Roman Latin because it was everyone’s second language. Over centuries, Irish, Italians, and Norwegians all speaking to one another in the same language simplified and streamlined it, to make it easier for everyone to understand. It was a working language, the language of the church and professional elites everywhere. One aspect of understanding what Renaissance was and why it happened was that Petrarch and others like him noticed how much more complex, sophisticated, and, to their ears, beautiful, the classical Latin of Cicero and Ovid was. They then tried to recreate that flowery, rhetorical style in their own day and time, in large part because they thought that the beauty and sophistication of the language also helped it convey beautiful and sophisticated ideas.

There has been a lot of recent discussion in the blogosphere on Dropbox security and cloud security in general. GF has a good overview, Susan Orlean has a post on some of the scarier implications. I can sympathize with people who feel they have been received, but I personally don’t think their original wording constitutes a lie. I never read it before all the current brouhaha, but as someone who is a little familiar with practical security, I always suspected from Dropbox’s features that, in principle, they had access to my files. Indicia:

1. All your files are available on different platforms (desktop, web, iPhone) without much hassle or delay.
2. While the various clients (desktop, iPhone, etc) could probably do per-user encryption, the web interface doesn’t spend a lot of time decrypting files in JavaScript (or Java or Flash or some other way of doing it).
3. Even large files can be instantly moved to “Public”.
4. The fact that it’s possible to instantly share an entire folder with someone else.

There’s probably some way to do each of these with proper encryption. For the last point, they could perhaps have a separate key for each folder, then give others access to that key when a folder is shared. And with some creativity there are probably ways to implement every Dropbox feature in a fully secure way. But my guess was always that it would be a huge hassle to have that level of security with a service like Dropbox.

All in all, Dropbox just “feels” like a service where they have access to everything (with proper access controls of course). Kind of like I always knew that authorized Google employees could read my email even though I never read the terms of use.

Different kinds of cloud services will always have different security tradeoffs. I won’t encrypt everything in my Dropbox because I want easy web and phone access to the kinds of documents I use Dropbox for. My 1Password files are in Dropbox, but they have their own iPhone app that can decrypt them, and I can live without web access to my passwords.

I encourage everyone to be careful about their Dropbox use and there are certainly things I wouldn’t include in my Dropbox, but that I would include in my Backblaze backups. And I don’t even trust Backblaze fully.

GF describes a different kind of balance:

SpiderOak, by contrast, cannot disclose its customers’ files, even if it wanted to. That is because it lacks tools to tap any of the data it stores on behalf of users. However, this “zero knowledge” means that if a user loses his key, he can never again access those data.

This is more secure and somewhat less practical than CrashPlan’s tradeoff, also described in GF’s post, but you are still reliant on SpiderOak’s application. They could easily be compelled by a court to introduce a back door that sends your encryption keys back to the bad guys. What, I’m paranoid? It happened.

I believe there are ways to use the cloud in a fully secure way, but you can’t rely on the cloud vendor’s software. If you use open source software to encrypt your files and then upload them to Amazon S3, it would be very difficult for baddies to read your files. Duplicity is an open source backup program that can use this security model.

(This is based on some of my comments on Hacker News.)

Last month Felix Salmon urged people not to donate to Japan in the wake of the earthquake, and if you do wish to do something, to make unrestricted donations to major charities like MSF. You should then trust MSF to allocate donations to countries and particular causes and to deliver help effectively or partner with a local organization that can do so. I guess the idea is that most people are unable to judge the relative merits of various causes, or don’t have enough information or are too biased to do so, and that if you just chase the latest cause célèbre a lot of your donations will be wasted.

This sounds quite reasonable, but it makes me wonder about the governance of nonprofits. Fama and Jensen (1983), the definitive work on separation of ownership and control, stress the importance of residual claims. Owners of residual claims on the firm—shareholders in the case of a corporation—are in the best position to monitor the firm because as residual claimants they stand to benefit, in most cases, from improved governance.

Nonprofits don’t have residual claimants. This is a more fundamental definition of “nonprofits” since many nonprofits actually have profits, but almost all nonprofits are subject to a non-distribution requirement, meaning that profits cannot be distributed to members or anyone else. There are no residual claims on nonprofits by design.

In nonprofits that are primarily funded by endowments, especially those with self-perpetuating boards as is common in the United States, governance is more or less hopeless because there is no external governance mechanisms. Governance will be good if the board members happen to behave well. Fama and Jensen point out that some nonprofits have very extreme bonding mechanisms, such as the poverty and celibacy vows in the Catholic Church, as additional governance mechanisms.

In the case of nonprofits without large endowments that are mostly funded by current donations, you might expect monitoring to be done by donors. If the nonprofit doesn’t behave well, donors will stop donating and the managers will lose their jobs (“private benefits” in the governance jargon), so managers will tend to behave well.

It seems to me that this model of governance would work best if donors actually earmark and target their donations quite narrowly. Maybe your favorite charity should spend more money in Haiti than Japan, or perhaps move resources away from Haiti to more worthy countries. You could trust that the charity’s managers will do this, but why not force their hand by earmarking your donations to the most worthy country (not Japan)? Donors, acting through the market for donations, will then have maximum control over how charity is directed and delivered.

What, then, is the agency theory of nonprofits that is consistent with Felix’s (and others’) recommendation that you give unrestricted donations to large, well-known charities? How can unrestricted donations be good for governance, and how can it be good simply to give to the most reputable large charity?

I think that first of all you need relatively large information acquisition costs at the individual donor level, so it’s costly to judge whether Haiti or Darfur or some third country is most worthy. Second, there has to be some mechanism (and enough information) for choosing the “best” charity, which right now seems to be MSF. I know that outfits such as Guidestar claim to do this. So the assumption is that Guidestar is able to tell you “give to MSF but not Yéle,” but not how to earmark your donation. Is this plausible? Is there a better theory out there?

You may have read this piece by David Honig or this post by Alex Tabarrok that links to him.

Do you have a cell phone? How would you like it if the FCC required you to pay an extra $20 a month to get movie downloads, whether you want them not, or to allow your kids to access violent video games or adult content, whether you want them to or not, just so everyone would get what the government considers to be “the full Internet experience?” What if you’re low income, and you’d rather spend that $20 on books? Or warm clothes? Or food?

My point has already been made by people in the comments and elsewhere, but I like repeating other people:

Net neutrality does not imply that poor people have to subsidize people who use a lot of bandwidth. There is no reason to force anyone to pay an extra $20 a month for more bandwidth or unlimited bandwidth. Most of the costs of running a network are fixed costs and it is quite reasonable and non-evil to allocate those costs in such a way that people who use the network the most also pay the most.

What is unreasonable and harmful is to discriminate by content, by charging someone more simply because they watch video or, even worse, because they watch video not made by a conglomerate that also owns the internet provider, rather than charging them more because they use more bandwidth.