What Can a Word Count Say About Gambling? Measuring Gambling Discourse in Chinese Investor Texts, and the Gaps Between Discourse and

Research linking China's gambling culture to stock-market behaviour measures gambling propensity through regional lottery sales, which is a coarse province-year proxy. Investor-generated text seems to promise a finer measure: retail investors articulate their dispositions daily on stock forums. This note builds the instrument and then argues, against its own initial promise, that a text measure can only report gambling discourse and not gambling disposition. It specifies why. Applied to two public Eastmoney Guba corpora (2,131 deduplicated documents), an untiered gambling lexicon finds gambling discourse in 7.1% of documents; once disposition-marking vocabulary (以小博大 "small stake, big win") is separated from trading mechanics whose gambling reading is contestable (涨停 "limit-up", 满仓 "full position"), this drops to 0.0%. Three gaps then separate any surviving density from a claim about persons. First, language versus speaker: Chinese carries gambling-derived idiom independent of any speaker's disposition. Second, register versus disposition: adopting a community register is a performance rather than a state. Third, speech versus behaviour: the crash-risk outcomes at stake are behavioural. None is crossed by counting words. The contribution is fourfold: a tiered lexicon with negation-aware scoring; a quantification of the contamination effect; an explicit validity framework that names the three gaps; and a research programme whose designs close them one at a time. Code, data pointers, a browser demo and results are public.