Foreign Stock Investing and Diversification (EEM, EFA)

Awhile back there controversy here at Seeking Alpha after Michael Panzer posted an innocuous chart of iShares Emerging Markets (EEM) vs iShares MSCE EAFE (EFA) suggesting that the performance gap between EEM and EFA was narrowing. I’ve found that the solution to most problems in life is muddy up the water, and statistics are great way to do that. Thanks to data from Yahoo Finance and software from SPSS (SPSS), I've drawn some conclusions about the relationships among SPX, VIX, EFA, EEM.

The first grand discovery, which many people has also discovered in the past few weeks, is that global stock markets are very correlated and they tend to be able to explain each other's variance. If you do factor analysis, the first extracted factor explains 97% of the data.

This makes sense because of the global nature of the world economy. The US and EFA economies tend to be synchronized with each other and emerging markets economies depend on exports to the developed world. The net result is that investing in foreign countries doesn't provide much if any diversification against serious market moves. At this point it seems that main benefit of investing in foreign assets is the implied hedge of foreign currency dividends against a weaker dollar. A foreign bond fund would accomplish most of the same effects with much less risk.

The other grand discovery is that EEM is strongly negatively correlated with VIX (Implied S&P 500 Volatility). This also isn't surprising because high VIX tends to be associated with market stress. For the past few years hot money has been flowing into emerging market investments through hedge funds, mutual funds, and ETFs. Much of this investment is being made without thoughtful assessment of the extra risks of investing in emerging markets. That risk includes a high beta towards the US economy and investor risk tolerance worldwide. Everyone looks like an investment genius when markets are going well. But when investors decided that volatility is a bad thing, the outflow from EEM was nasty.

The take home from all this is that you can't rely on common stocks to diversify away the inherent risks of investing in common stocks. To do that you need something other than common stock, i.e bonds (TIP). Even things like REITs (ICF) or Gold (GDX) don’t give as much diversification as you would expect.

With world markets as interconnected as they are, you can't expect EFA to be a tower of strength when SPX is crumbling.

Comments

[Geoff Consi...:]

Hi MP:

While my own articles mirror many of your points--that international investing does not provide nearly as much diversification as many people think (see articles posted here going back to January)--I am trying to understand how you can say global markets can be used to explain 97% of each other's variance. In short, this implies that you can come up with some portfolio that is so highly correlated with the S&P500 as to have an R^2 of 97%. This may be possible over some periods of time, but I think that just because you can come up with an optimal linear model for a certain period of time will tend to overstate the case. If what you are talking about is a principle components analysis (PCA), then this definitely overstates the case. The weights in PCA models tend to be very sensitive to the specific data set. The bottom line is that I would be interested in what you are trying to get at. In my own work, I find that international funds have high Beta but low R^2. While I could jigger some 'optimally weighted' portfolio in a set of markets to explain variability in another, I don't think that this would mean much from the perspective of practical investing knowledge.

2006 Jun 20 03:18 PM

[WC Whiner:]

All you need to know about his data set is found in the three letters "EEM". That fund started in May 2003. It is my distinct suspicion that as a result of the very short time frame, our intrepid analyst used daily price movements. Me, I am a fan of using monthly numbers (which tend to flatten out daily noise to a certain extent) and time periods as long as I can get, to maximize the market regimes I can analyze. 5/03-present includes precisely two regimes: a trending global equity upswing favoring emerging markets, and a trending global equity downswing hitting the same group the hardest. Of course his PCA has .97 explanatory power -- it just says very, very little.

Let me grab some slightly longer time frames and run a simple monthly regression. While non-US equity investing can never diversify away your systematic market risk, the idea that it is of little or no benefit does not pass the smell test. Lessee... scribble scribble..

OK, I can't get total returns on short notice. I can grab GFD's price indexes through my old university library. Here's an ordinary least-squares run of the S&P versus GFD's World-ex-US and Emerging Markets price indexes, for which I have concurrent returns from July 1920, so over a thousand months and many different market regimes:

> print(summary(lm(SPXpr ~ xUSpr + GFD.EMpr, data=emp)), digits=1)

Call:
lm(formula = SPXpr ~ xUSpr + GFD.EMpr, data = emp)

Residuals:
Min 1Q Median 3Q Max
-2e-01 -2e-02 -5e-04 3e-02 4e-01

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.003 0.002 2 0.03 *
xUSpr 0.442 0.042 10

Adjusted r-squared here is 0.2. A little different than .97, don't you think?

2006 Jun 20 06:58 PM

[WC Whiner:]

Wow, that table didn't work.

Check out the R output at www.bignose.org/~wcw/GFDEMregression....

As I said, .97 it surely, surely ain't.

2006 Jun 20 07:04 PM

[WC Whiner:]

Holy jeepers, I think he was regressing prices, not returns. Could you be a little more clear on what you mean by, "if you do factor analysis?" The only way I get .97 with daily numbers from Yahoo back to EEM inception is with price regression. Using returns (including VIX "returns"), I only get daily r-squareds around 0.7.

2006 Jun 20 07:28 PM

[Geoff Consi...:]

Ouch! I sure hope its not prices, but that would explain a lot. In my own article on this issue, I make the point that while Betas are high, R-squared tends to be low:

etf.seekingalpha.com/a...

This is a unique feature of foreign ETF's.

MP: Please tell me it wasn't prices you were regressing.....

2006 Jun 20 08:40 PM