Building An All-Seasons Tech ETF Portfolio

Semiconductor inventories are down and the industry should benefit next year, analysts say, as demand for consumer electronics, especially cell phones — upper end models in the developing world and low-cost units elsewhere — rises. This, they suggest, could spark a rally in chip stocks that were previously pounded due to disappointing earnings news. But even if they're right about industry conditions, technology investors may want to consider a more balanced sector strategy.

Thus far this year, chip stocks have been the laggards of the technology sector. Table A shows how poorly the three semiconductor exchange traded funds (ETFs) have performed on balance.

Obviously, past performance is no assurance of future results. So a case can be made that however weak chip stocks previously were, it might still be worthwhile to get in now in anticipation of a rally. The problem is that buying and selling at the right time is easier said than done.

We studied performance trends of 21 technology ETFs in an effort to create an all-seasons portfolio that stands a reasonable chance of generating satisfactory performance without having to correctly guess which tech sub-sector will be hot at which point in time.

We started by examining share price changes in successive five-trading-day periods from the start of 2006. We then ranked the share price performance of the ETFs for each of the 40 periods (based on a descending sort, where larger numbers mean superior performance). Table B presents three views of the results:

• The column labeled "all" is the average of the rankings achieved by the ETF over the course of all five-day periods.
• Each column under the "market type" heading shows the average of the rankings achieved by the ETF over the course of different types of five-day periods.
  • The column labeled "strong" is the average of the rankings achieved by the ETF over the course of only those five-day periods during which tech ETFs as a whole appreciated at least 1 percent.
  • The column labeled "neutral" is the average of the rankings achieved by the ETF over the course only those five-day periods during which the average tech ETFs price change was between -1 percent and plus 1 percent.
  • The column labeled "weak" is the average of the rankings achieved by the ETF over the course of only those five-day periods during which tech ETFs as a whole fell more than 1 percent.

The table is sorted, from best to worst, based on all-period average performance ranks.

We quickly notice that semiconductor ETFs are clustered near the bottom of the all-period rankings while telecommunications funds are bunched near the top. Notice, too, how things change when we focus on different tech-market environments. During sector rallies, semiconductors tend to be among the performance leaders while during sluggish periods, telecommunications tend to shine.

Low correlations — asset groups whose prices tend to move in unrelated or even opposite directions — is the theoretical underpinning to diversification, and the data we see above suggest that diversification, even within tech, should be productive. So rather than trying to guess when the market will favor semiconductors, technology, or any other tech sub-group, we'll seek out a diversified portfolio. We also decided to use five ETFs.

Selecting individual funds to represent different areas within technology is not necessarily an exact science; different observers tackling the same task would likely make different decisions. Here are the choices we made:

• To represent telecommunications, we chose Vanguard Telecomm Svc (VOX), the all-seasons leader within that area.
• We selected PowerShares Dynamic Hardware & Consumer Electronics (PHW), the all-season leader among non-telecomm funds. It focuses on hardware and consumer electronics.
• We then added PowerShares Dynamic Software (PSJ), representing the software sub-group. It was the top all-season performer outside hardware and telecommunications.
• The remaining two choices represent areas where the art, rather than science, of asset selection comes into play. Had we stayed strictly on script, we probably would have skipped over the next all-seasons category leader, which was Select Sector SPDR-Tech (XLK) a general tech fund, and gone to PowerShares Lux Nanotech (PXN), and finished up with PowerShares Dynamic Networking (PXQ). Instead, we changed course:
  • We decided to go with the Select Sector SPDR-Tech fund. While it will replicate some of the exposure we achieve elsewhere, it also adds sub-sectors we'd otherwise miss. We note, too, that its all-seasons ranking is high; its weak-market ranking is even better. So if history holds true — an always iffy but commonplace assumption — it would seem like this fund may add a defensive quality to the portfolio.
  • Finally, recalling why we undertook this exercise in the first place — the potential for a strong near-term rally in semiconductors — we decided to add a fund from that sub-group. Among these, we chose the top all-seasons performer, PowerShares Dynamic Semiconductors (PSI).

Table C summarizes share price performance of the ETFs we chose.

Had we wished to go on past performance alone, we'd have simply put all our eggs in the Vanguard Telecomm Svc basket. Assuming we don't want to do that, and further assuming a standard allocation of equal dollar amounts for each fund, Table D shows that on the whole, our ETF portfolio would have made a reasonable showing in 2006. While there is no way to be sure that experience would be replicated going forward, given the generally all-seasons nature of our choices coupled with a judgment call to play a potential rally in semiconductors, it seems like at least one reasonable set of choices.

On closer review, we notice that the general sector wide fund, Select Sector SPDR-Tech, outperformed the overall portfolio for all but the year-to-date interval.

Is this a case of overkill? Perhaps we'd have been best off paying homage to standard mantras about longer holding periods and going with Select Sector SPDR-Tech.

We decided to examine the issue more objectively, using established portfolio optimization techniques; the jargon among those who've studied finance is "efficient frontier." This goes beyond a simple inquiry into relative performance and considers period-to-period variations ("standard deviation") and also, perhaps more importantly, analysis of the extent to which price movements for each asset tend to match or diverge from price movements of all other assets ("covariance"). The aim, here, is to identify a set of asset allocations that will, if history holds, maximize potential return while at the same time minimizing the likely fluctuations.

Usually, optimization is done on the basis of expected return. For this exercise, rather than trying to predict share-price future performance, we decided to optimize instead on the basis of all-season performance rankings; in other words, we tried to maximize expected ranking while minimizing the extent to which rankings would vary from period to period. We can't say for sure, but we assumed there would be a better chance rankings could be sustainable over time than would be the case for a particular price-performance forecast.

The calculations produced an interesting result. Table E shows the recommended portfolio allocations.

The suggested allocations are eye-catching. Potential semiconductor strength inspired this exercise yet now, the optimization model suggests we completely omit the most relevant ETF. Also, it looks like we now have less diversification: we'd be downsizing our stake in the general fund and in the hardware-consumer electronics offering.

Even so, Table F shows that the optimization would have been well worthwhile in 2006.

The ultimate question is whether this is a reasonable strategy going forward.

The heavy weighting in the telecoms fund comes as no surprise given use of historic data. Interestingly, that same information confirms that the software ETF may be an adequate vehicle for diversification. As noted above, the extent to which pairs of assets experience price movements that are or are not in tandem figures heavily in the optimization process. We see for the telecom and software funds a -0.83 ranking correlation.

That is almost perfect diversification. A correlation of 1.00 would mean that the ranks of the funds move completely in tandem. A rank of -1.00 would mean the changes are completely opposed. The result we got is very close to the latter.

Reiterating that we cannot blindly rely on past performance, something that can never be repeated too often, there is still something to be said for the results of the optimization. It reminds us of the danger of going into semiconductors without a strong conviction future strength, if it materializes, is likely to be sustainable. It would be one thing if the number-crunching suggested a low weighting in chips, say 10 percent. But the number crunching came back with a zero! That's hard to ignore. Bear in mind that even the past period did include some intervals when chips performed quite well. It's just that in a big-picture sense, these weren't sufficient to overcome generally weak showings.

On reflection, there is reason to expect that such an outcome might be replicated in the future.

Chip-making is a generally commoditized business. It's not perfectly so; firms like Intel Corp. (INTC), Advanced Micro Devices Inc. (AMD) or Nvidia Corp. (NVDA) do create new variations. Ultimately, though, rivals catch up. And even when strong market shares can be sustained, the fortunes of the chip-makers still depend on prospects for the end products.

Speaking of the latter, recall that analysts who recommended chips did cite cell phones as an especially strong potential end use. That would also benefit many companies that appear in the portfolios of telecommunications ETFs. The latter portfolios would, however, also benefit from holding among firms whose revenues include more stable streams of service fees.

It would likewise benefit the hardware and consumer electronics ETF, the allocation for which was downsized. But we could pick up many of those benefits with the telecommunications and software ETFs, the ones that now would dominate the portfolio.

Bottom line: Even if you agree with the bullish chip scenario, there may be better ways to invest than loading up on big-name semiconductor stocks in the hope of catching a quick but potentially short-lived surge.

At the time of publication, Marc H. Gerstein did not own shares of any of the aforementioned companies or ETFs. He may be an owner, albeit indirectly, as an investor in an open-end mutual fund or another Exchange Traded Fund.

Note: This is independent investment and analysis from the Reuters.com investment channel, and is not connected with Reuters News. The opinions and views expressed herein are those of the author and are not endorsed by Reuters.com.

Comments

[SeekingAlphaEditors]

Marc, great writeup of an interesting and important portfolio challenge.

I'm wondering why you didn't include SMH in your overview/analysis of the semi ETFs? If I'm not mistaken, that's the largest ETF in the sector by assets.

[MarcGerstein]

Mick, good question.

If one determines ahead of time that it's to be semis, yes, SMH should be in the mix. But if one is looking for a diversified ETF portfolio, SMH, which strictly speaking is not en ETF but a HOLDR, might cause awkwardness since, as with all HOLDRs, purchases must be made in 100-share increments. Therefore, it may not mesh with allocation preferences. This may or may not be an issue, depending on the size of the portfolio ($ per security) and how firm one's convictions are regarding asset %s.

[JonD]

interesting article. I'm going to now go jump in my time machine to capture the excess return of this strategy.

seriously, I own VGT b/c it's the cheapest ETF on the board (by far) and also the most diversified (measured by the most total holdings in the portfolio and the least concentration among largest stocks) among the broad Tech ETF offerings.

the sector ETFs are interesting if you want to make a call on a certain sector short term but i don't think are true diversifiers long term as their components are well represented in the broader Tech offerings (except for PXN I suppose)

where did I put that flux capacitor now...? funny, I actually saw a DeLorean on the road the other day. pretty neat.

[Man Jiraa]

Great article on ETF.
Any thoughts with Income /Dividend/Current Income stream generating
ETF portfolio ?

[MarcGerstein]

I looked into the income ETFs but didn’t find as much opportunity to apply this sort of analysis. You do have a wide range of covariances, etc. but there’s more predictability here. If you expect rates to fall, longer-term instruments tend to outperform, etc, etc.

Seems to me, at least so far, that if you can correctly guess which way rates will go, pick one appropriate fund; that should do it. For those who are completely undecided, just stick with a middle of the road fund.

That said, I’m not sure any of these approaches are quite up to what one might accomplish with a good income-stock screen. In theory, higher yield means higher risk (dividend cut and/or poor growth), but from what I’ve been seeing, the market’s worries often tend to be disproportionate to real business threats. If you do a seekingalpha ticker search on WMT, you’ll see something I wrote in the past week on strip-mall REITS that look better the more Wal-Mart struggles; you have some nice yields there.