Lessons on accruals shifted retail buying only when they included the calculation

An accrual is the part of reported earnings that no cash has arrived for. Earlier work, which the team took as its starting point, found that retail investors weight reported earnings too heavily, bid accrual-heavy stocks up after an announcement and then sit through the reversal ^[13].

Two of the four arms exist to absorb confounds. One group of stocks got nothing but a note that earnings were coming ^[10]. That arm removes the effect of simply drawing attention to a date. A fourth group got no posts at all ^[11]. The finding rests on the gap between the arm taught both the concepts and the method for calculating accruals ^[9] and the arm given the date alone.

The return figures need care. The 4.2 and 3.9 point differences describe the stocks' price paths over the following year against the announcement-only control ^[15], not the profit in any investor's account. A stock bid up less at the announcement has less to give back, and the give-back is what registers as a lower one-year return in the control arm. The team measured prices and retail buying and selling ^[18]; the design does not follow individual portfolios.

Reach was large in total and thin per name. More than 30 million views across the platforms ^[12], spread across the roughly 3,500 stocks in the three arms that received any post, is about 8,600 views per stock ^[20]. Each stock also had its own page on the education website the team built ^[6].

Contamination here works against the researchers. The same Weibo and Stocktwits audiences saw posts about treated and control stocks, and an investor who has learned to compute accruals for one company can compute them for the next, so leakage should shrink the measured difference. On that reading, the closeness of the two markets, 0.3 points apart ^[21], is the sturdier half of the result, because the same lesson run on two different retail populations produced nearly the same number.

The design cannot separate skill from effort. The arm with the calculation instructions and the accrual tools ^[9] differs from the concepts-only arm ^[8] in both what it taught and how much work it invited, so a reader who looked harder and a reader who analysed better are indistinguishable in the order flow. The paper, by Zhao Xiaofeng of Lingnan University with co-authors at Henan University and The Chinese University of Hong Kong, Shenzhen ^[4], is in the Journal of Accounting and Economics ^[5], and the published summary does not report standard errors for the return differences ^[22].