The Schwab Trading Activity Index came in at 57.50 for August, down from 59.80 in July. That is the first monthly decline since April, and it arrived in a period when the S&P 500 gained 2.96%, its strongest showing since May. A behavior index that falls while the market it measures rises is telling you something the index level alone will not.
The composition of the flow is where the reading gets interesting, and where the obvious interpretation of it is wrong. The five names most heavily net sold by Schwab clients were Palantir, Microsoft, ServiceNow, Salesforce and Oracle. The five most bought were SpaceX, Micron, Nvidia, Intel and Alphabet. Put those lists side by side and the story writes itself: retail took profits in software after a sharp rebound and rotated the proceeds into semiconductors. Joe Mazzola, Schwab’s head trading and derivatives strategist, said roughly the first half of that himself, describing clients as taking profits in high-beta software while staying engaged with growth and innovation leaders.
The second half does not survive contact with the sector data.
Only three S&P 500 sectors attracted net buying from Schwab clients in August: Industrials, Utilities and Real Estate. Information Technology saw the largest net selling of any sector, followed by Communication Services and financials. Semiconductors sit inside Information Technology. Alphabet sits inside Communication Services. So four of the five most-bought individual names live in sectors that were, on net, being sold. This is not a rotation into chips. It is a book getting smaller in technology overall, with what remains concentrated into a handful of names that clients were willing to keep or add.
There is also a timing problem with the phrase “buying the dip.” The semiconductor dip was a July event, and a violent one. The Philadelphia Semiconductor Index fell 21% in July, its worst month since October 2008, with intraday swings of at least 2% on every session of the month. Micron dropped 29%, its steepest monthly decline in more than a decade, erasing about $374 billion of market value. Intel fell 35%. By the time the August measurement window was underway, the SOX had already rallied more than 20% off its July 29 low and climbed back out of bear-market territory. Clients buying Micron and Intel in August were buying a recovery that was well advanced, not a falling knife.
Which leaves a buy list containing two different behaviors dressed as one. Micron and Intel were the largest single-name casualties of July, and buying them is a drawdown trade: the position is sized against how far the price fell, not against a revised view of the business. Nvidia and SpaceX are the opposite, momentum names that had not been damaged and were being added on strength. A list that holds both is not evidence of a thesis. It is evidence of two groups of clients doing different things in the same month, which is exactly the kind of dispersion that pulls a median-based index lower even as the tape goes up.
The vehicle data supports that reading more directly than the single-name data does. Two ETFs ranked among the top five net buys and four appeared in the top ten, which is a high share for a list dominated in most months by individual equities. Call buying stayed strong, concentrated in companies tied to tech-heavy ETFs, but the flow shifted away from the put buying that characterized earlier periods. Clients wanted defined-risk exposure that kept them in the market without rebuilding equity positions at full size. Broad instruments, capped downside, sector-level selling, concentrated single-name adds. That is a book being managed, not a view being expressed.
The macro backdrop during the window explains the caution without needing to invoke sentiment. The July nonfarm payrolls report, released early in the period, showed a decline of 23,000. Second quarter GDP came in at 1.5% on a seasonally adjusted annual basis. Monthly inflation growth moderated over the summer, which should have been the friendly part, except that oil fell early in August and then rebounded as Middle East tensions reemerged, dragging Treasury yields up with it through the month. Hawkish remarks from Chairman Warsh landed on the final afternoon of the measurement period and lifted September hike odds sharply, too late to register in the score but not too late to shape what clients did next. Gen X clients again led net buying by age cohort, and self-directed traders remained considerably more skewed toward buying than self-identified investors, which is the usual split and tells you the marginal buyer in this data is the more active account.
For anyone holding the names on either list, the position math matters more than the flow. Micron and Intel are being bought at prices that already reflect a substantial bounce, which means the July drawdown has been paid for once by whoever sold it and is being paid for again by whoever is buying now at a higher entry. Nvidia sat on the buy list going into its late-August print, and the subsequent chip rally rewarded that positioning, the one place in this data where the crowd was early rather than late. Salesforce sat on the sell list and then rallied more than 20% on its second quarter beat. If the STAX window closed before that move, as the measurement calendar suggests, then a meaningful slice of the software selling was executed directly into the news that reversed it.
That is the trade to watch in the September print. The single-name lists are five observations and they will churn. The sector flow is the entire sampled book, and it is currently telling you that clients reduced technology exposure while keeping a small number of technology positions. If Information Technology returns to net buying in September, the August selling was profit-taking around a rebound and the chip names were the keepers. If the sector stays in net selling while Micron and Nvidia remain on the buy list, then what this data captured was not conviction in semiconductors at all, but the last names to be sold.