The 4.3% print includes a tariff refund. Budget against 1.8%, not the tailwind

The federal government has returned more than $100 billion to U.S. businesses and importers that paid the global tariffs ^[1], and Apollo chief economist Torsten Slok estimates that money will add roughly 0.2 percentage point to third-quarter GDP growth ^[5]. The Atlanta Fed's tracker has the quarter running at 4.3% ^[6], against 1.5% in the second quarter and 2.1% in the first ^[7][8]; the question for anyone setting a 2027 capital plan is which of those numbers describes the machine and which describes the accounting.

Start with what is visible. Forty S&P 500 companies have booked $9.6 billion of refunds, with Apple alone reporting nearly $2.2 billion, according to a Wall Street Journal tally that also lists Nike, FedEx, Amazon and General Motors among the top recipients ^[2][3]. That averages about $240 million per company ^[1], but Apple is roughly 23% of the tallied total on its own ^[2], and the whole S&P 500 tally is under a tenth of the money returned so far ^[3]. The refund is real cash and it is lumpy.

It is also not fully yours if you received it. Some consumers are suing companies to claim a share, and Amazon, FedEx and UPS have said they will return funds to customers ^[11]. Bank of America wrote earlier this month that retailers are spending refund money on promotions and on offsetting freight and other supply-chain costs ^[12], and expects some retailers to try to claw back tariff money from brands through direct payments or future purchase-order terms ^[13]. The bank's remaining scenarios are investment in the business, including AI and tech, or returning capital to shareholders ^[14]. In other words, a good deal of the earnings benefit is either a pass-through obligation or a discount already given away.

Now the growth arithmetic, because the honest version is less dramatic than the headline. Strip Slok's 0.2 point out and the quarter still tracks at about 4.1% ^[5], versus a 2.8-point acceleration from Q2 ^[4]. Refunds are a small slice; the rest is attributed to the AI spending boom, tax cuts from the One Big Beautiful Bill Act, and reshoring of manufacturing ^[9]. But the Q2 base was itself distorted, depressed by heavy AI-related imports ^[7], which means both prints are measuring trade timing as much as output. The first-half average of 1.8% ^[6] is the duller and more defensible planning base, and the refund contribution is a level effect: present in Q3, absent in Q4, and a mechanical subtraction from the growth rate thereafter.

Slok's read on labour is the load-bearing part of his bull case, and it is his read alone. He attributes July's weak report to seasonal-adjustment quirks in government and hospitality payrolls, arguing that excluding those sectors the economy added 70,000 jobs in line with consensus rather than losing 23,000 ^[10]. He points to jobless claims near 200,000 a week and six months of rising job openings ^[15], and concludes that the market is underestimating growth and that rates will stay higher for longer ^[16].

Three things to watch. The refunds so far are about 60% of the $166 billion collected under the International Emergency Economic Powers Act, which the Supreme Court struck down in February ^[17], leaving roughly $66 billion still to flow ^[7] and the timing of that disbursement determining how far the tailwind stretches. Second, whether Q4 trackers show the 0.2 point simply dropping out or compounding into a negative base effect. Third, disclosure: which companies are carrying refunds as retained margin and which are carrying them as a liability to customers or a credit owed to a brand partner.