MGIC Investment Corporation (MTG) - Normalized ROE does not look attractive

This could turn out to be a multi-part project. I just started looking at this company last week so my insights would be limited. However, I’m writing anyways as it helps me gather my thoughts and move forward.

An Inherently Unattractive Industry

MGIC Investment Corp (MTG) is a Private Mortgage Insurer (PMI). This is one crappy industry. Results are incredibility cyclical and sensitive to some assumptions. On the competitive front, we have already seeing new players trying to compete on not just price, but underwriting terms (NMIH Holdings is an example). Now it looks like the 7 players will not be putting a united front with respect to proposed capital requirements.

Why would anyone buy a PMI then? The standard long thesis says 1) FHA will be giving away market share to the private sector, this coupled with an improving mortgage market will lead to high volumes. 2) Lower losses from legacy vintages running off, as well as fixed cost/operating leverage would lead to a spike in earnings.

Some parts of this hypothesis are valid but I think the sell side tends to bake in both 1) mortgage industry recovery, AND 2) losses stay low at post crisis levels, when in fact the two may not be compatible. While losses are sure to come down from present levels in the next couple years, some analysts seem to assume the pristine underwriting quality of 2009-11 vintage will continue forever. Given that 1) first time homebuyers will be needed to drive housing recover and 2) they tend to be of lower credit quality, the assumption of “2009 forever” is clearly unrealistic.

So you have multiple offsetting factors at play and investing in the sector requires one to say “ok, new businesses will have higher losses at some point, but I don’t think it will be that bad, and meanwhile this thing is dirt cheap and I like the risk and reward”.  This reminds me of my write up on Santander Consumer USA. While there’s a place in your portfolio for a speculative play like this, these could be good ideas but not the best ideas.

Normalized Losses and ROE

Unattractive industries demand great valuations. I want to get some sense of economics and value before filing this away. To do that, some conception of a “normalized return” is needed. Anyone who has followed the industry would know that current loss levels are far from “normal”, since MTG is still working off legacy businesses. So blindly applying some P/E or P/B ratio to next year’s forecast would be meaningless. 

Below are my estimates of new businesses economics when losses normalize, under the current capital regime, and then under the proposed PMIERs.  I’m thinking about this at the opco level so that’s why there’s no interest expense.

 

Two part discussion here, first on impact of capital requirement, then on normalized losses.

First, capital requirements. Management said that if PMIERs goes through, capital requirements for recent businesses probably correspond to 11.5-12x in the old risk to capital framework and that gets them to low teens ROE before reinsurance.  

From 2Q14 transcript: “…Under the proposed eligibility requirements, the mix in the first half of the year seems to require a risk to capital of about 14-to-1 at time of origination, i.e. they are all current. But as we know, even with the high-quality profile, some will go delinquent. So, if you factored that in and probably goes to 13-to-1. And then if you want to add some room away from the capital requirement just to give yourself some margin you probably talking 11.5-to-12-to-1.

And by our calculations, on a direct basis before any reinsurance and whatnot, we think that delivers a return in the lower double-digits. On the current or prior to eligibility requirements that we are issued here, we were think and closer to 18-to-1. And if you give yourself a little room and whatnot operate around 16-to-1. We think those returns are kind of back as Curt said where they probably should be for the overall risk of the business in the mid-teens.”

So let’s say losses will revert higher in new vintages going forward, I think a 10:1 risk to capital, single digit ROE after reinsurance is probably reasonable.

How about normalized losses? In the 2Q14 call, management said the 2009-2011 books are running about 15-20% loss ratio. This is how I got the 35bps loss as % of RIF assumption (50bps premium * 17.5% loss ratio / 25% RIF = 35bps). Keep in mind 2009-2011 vintages are loans with pristine underwriting standards. Going forward as the mortgage industry reach down the credit spectrum, it’s reasonable to think that losses will be higher. How much higher? To give an idea of how volatile these items are, below are historical loss ratios from 1996 to 2006 before the whole industry blew up. I used 50bps credit loss in the above table as a placeholder, but the ROE sensitivity table is all over the place.

So you got a cyclical, competitive industry getting hit with higher capital requirements. What should investors demand? The CEO gave some jumbled answers on the 2Q14 call but I think he meant to say mid-teens return overall and high teens for low FICO/High LTV businesses:

GS analyst: “Now, but if you were just, say, isolated – let's say you were the only player in the industry in a very hypothetical scenario, I mean, what required return would you want to get on those lower FICO, higher LTV buckets? I mean, would you be looking at low-teens? High-teens to account for some of the greater volatility in those buckets?

Curt S. Culver: “Yes, but I, on your question, I think for the lower FICO you need a mid-teens minimum return, given the variability on that business and how quickly things can change. So, certainly it demands a higher return. The returns on the other business will be, I think low-to-mid teens so that certainly would require in my opinion, a high-teens return.

So management want double digit returns but the analysis above shows that ROE with 1) normalized credit loss, and 2) new capital regime will likely be in the high single digits. And MGIC trades at 3x book value when the only thing we can really count on is volume growth. Surely there are better ways to play a housing market improvements? 

Up to now, I have referred to new business economics at the opco level, assuming equity capital = investment assets. In reality there's a mix of vintages books, the investment portfolio is much higher and there's interest expense from the holdco debt. To value MTG you have to take those into account. I won't bore anyone with the model here but my calculation shows that MTG is about fairly valued right now ($8.4 per share).

A Cash Flow Model

REIT Analyst did an SA article on MGIC last week. Specifically, he actually tried to project out the cash flows of mortgage insurance premiums and losses, then calculate a present value. Financials analysts as a group tend to stay away from cash flow statements, so what he’s doing is very different. That and the result of over 100% upside got my attention.

I can understand why his valuation is so much higher than market and what I have above. My guesses are 1) discount rate used - the market is rightly demanding more than 10%, 2) Not all investments are excess. Put another way, they are operating assets required to back the day to day MI business, so the investment income stream needs to be discounted together with other operating cash flow streams.  3) subtle assumption difference in premium, loss curves, reinsurance...can all make a big difference. 

Nevertheless, some sort of cash flow analysis would be useful to quantify the positive effect of legacy vintage running off - a big part of the long thesis. A real deep dive here would mean building a full cash flow model myself. At this point it is not a high priority given all the negatives I discussed earlier.

For now, I say we give this guy the sensitivity table and a set of darts, and call it a day.