Cost stickiness, earnings forecast accuracy, and the informativeness of stock prices about future earnings: evidence from China

The study examines whether and how firms’ cost stickiness influences the information content of stock prices about future earnings in China’s capital market. Traditional asset pricing often abstracts from cost behavior, yet prior research shows costs decrease less when activity levels fall than they increase when activity rises by the same amount (i.e., sticky costs). The paper predicts that lower cost stickiness reduces earnings volatility, improves investors’ and analysts’ ability to forecast future earnings, and thereby increases the informativeness of current stock prices about future earnings. It tests this prediction by linking current returns to future earnings via the future earnings response coefficient (FERC) and by assessing stock price synchronicity as a proxy for the incorporation of firm-specific information. The context of China is apt due to strong analyst influence on investors, relatively stable macro conditions, and pronounced differences in cost behavior between state-owned and non-state-owned enterprises. The authors posit three hypotheses: H1: cost stickiness reduces the earnings response coefficient (ERC) and FERC; H2: the effect on FERC is weaker for state-owned enterprises; H3: cost stickiness increases stock price synchronicity.

The paper builds on extensive literature documenting cost stickiness across settings and cost categories (Anderson et al., 2003; Dierynck et al., 2012; Banker et al., 2013) and its implications for operating efficiency and risk (Anderson et al., 2003; Yao, 2018) and for analysts’ forecast behavior and accuracy (Weiss, 2010; Ciftci et al., 2016; Banker and Chen, 2006). Stock price informativeness is linked to the pricing of current and future earnings through ERC and FERC (Collins et al., 1994; Lundholm and Myers, 2002; Ayers and Freeman, 2003; Choi et al., 2019), and stock price synchronicity captures the relative incorporation of firm-specific versus market/industry information (Roll, 1988; Morck et al., 2000; Durnev et al., 2004). Prior work also highlights the role of analysts and institutional investors in information incorporation (Piotroski and Roulstone, 2004) and differences in transparency and political influences for state-owned enterprises (Bushman et al., 2004; Shleifer and Vishny, 1994). This study contributes by directly linking cost stickiness to ERC/FERC and synchronicity, and by contrasting effects between state-owned and non-state-owned firms in China.

- Sample and data: Annual observations for Chinese A-share manufacturing firms (C1311–C4390) from 2009–2021. Data sourced from CSMAR and iFinD. Exclusions include ST/*ST firms. Observations are limited to firm-years where costs and sales move in the same direction. Final main sample includes 15,976 firm-year observations for forecast-accuracy tests; FERC tests report subsamples (e.g., full sample N=8,418; SOEs N=4,021; non-SOEs N=4,397). Consensus analyst forecasts are computed from all forecasts issued in the month preceding the earnings announcement. - Cost stickiness (Sticky): Firm-level measure adapted from Weiss (2010). Activity change is proxied by change in sales. Total cost change is ΔCost = (Sales − Earnings) differences across periods, with Earnings measured before non-recurring items. The metric compares the cost increase in the most recent quarter with rising activity to the cost decrease in the most recent quarter with falling activity over the last four quarters. Observations with costs moving opposite to sales are excluded. Lower values indicate stronger cost stickiness. Variable winsorized at 1% tails. - Stock price synchronicity (Syn): Following Morck et al. (2000), regress weekly (and separately daily) stock returns on market and industry returns to obtain R², and transform via log(R²/(1−R²)). Two measures: Syn1 (weekly) and Syn2 (daily). - FERC model (returns-earnings association): Rit = β0 + β1Xit−1 + β2Xit + β3Xit+1 + β4Rit+1 + εit. Extended with Sticky and its interactions with Xit−1, Xit, Xit+1, Rit+1 plus controls (Lnsize, MB, Earnvol, Retvol, Analyst) and year/industry fixed effects, to test how cost stickiness moderates ERC (Sticky×Xit) and FERC (Sticky×Xit+1). - Synchronicity model: Syn_{t+1} = β0 + β1Sticky_{it} + controls + Year + Industry + ε. Controls include Lnsize, Roa, Lev, SelManexprat, Loss, Seasonvol, Ihold, Dividend, MB, Retvol, Dturn, Analyst, Big4, ABACC, State, with fixed effects. A dynamic model examines changes: Syncha_{t+1} = Syn_{t+1} − Syn_t regressed on Stickycha_t = Sticky_t − Sticky_{t−1} and controls. - Analyst forecast accuracy (mediator): Forecast error FE = (foreEPS − EPS)/Price_{t−1}×100; accuracy proxied by absFE = |FE|. Regress absFE_{t+1} (and FE_{t+1}) on Sticky and controls (information environment and uncertainty proxies: Vsale, Oplev, FI, DTE, Surprise, Earnvol, Retvol, Seasonvol, CFvol) with year/industry fixed effects. - Mediation tests: Include absFE and its interactions in the FERC and synchronicity regressions to test whether forecast accuracy mediates the impact of cost stickiness (Sobel-type approach per Baron and Kenny, 1986). - Robustness: Alternative stickiness measures (e.g., eight-quarter M_Sticky; He et al., 2020 specification), lagged Sticky_{t−1} to focus on prior decisions, additional controls (accrual quality per Dechow–Dichev, management forecasts, market beta, Altman z-score), firm fixed effects, first-difference GMM, and propensity score matching to mitigate endogeneity and selection concerns. Results remain consistent across checks.

- ERC and FERC effects (H1): Cost stickiness reduces the informativeness of returns about current and future earnings. In the full sample, Sticky×Xit is positive and significant (0.249, p<0.01), and Sticky×Xit+1 is positive and significant (0.065, p<0.01). Given that lower Sticky indicates stronger stickiness, these coefficients imply that greater cost stickiness diminishes both ERC and FERC. - Ownership heterogeneity (H2): For state-owned enterprises (SOEs), Sticky×Xit is positive and significant (0.252, p<0.05), but Sticky×Xit+1 is not significant, indicating cost stickiness lowers ERC but does not significantly affect FERC. For non-state-owned enterprises, Sticky×Xit is positive and significant (0.264, p<0.05), and Sticky×Xit+1 is positive and significant (0.089, p<0.01), implying reductions in both ERC and FERC. This supports H2 that the impact on FERC is weaker for SOEs. - Stock price synchronicity (H3): Higher cost stickiness is associated with higher synchronicity (i.e., less firm-specific information in prices). In regressions of Syn1_{t+1} and Syn2_{t+1}, Sticky has negative and significant coefficients (−0.003 to −0.005, p<0.01), consistent with the coding where lower Sticky values reflect greater stickiness; thus, greater stickiness increases synchronicity. Dynamic specifications show that decreases in Sticky (greater stickiness) are associated with increases in synchronicity changes (Stickycha coefficients ≈ −0.003 to −0.006, p<0.01). - Mediation via forecast accuracy: Cost stickiness significantly worsens analysts’ forecast accuracy. For absFE_{t+1}, the Sticky coefficient is significant (e.g., −0.456, p<0.05; more negative Sticky implies larger forecast errors). Including absFE in ERC/FERC models reduces the magnitude and/or significance of Sticky×X coefficients (e.g., Sticky×Xit falls from 0.276 at p<0.01 to 0.192 at p<0.05 in the full sample), indicating partial mediation. In SOEs, absFE×Xit is negative and significant (−0.007, p<0.01). For synchronicity, adding absFE attenuates the Sticky effect and absFE is positively associated with higher synchronicity (e.g., 0.096–0.112, p≤0.10–0.05), indicating mediation. - Additional evidence: Descriptive statistics show high synchronicity in China relative to the U.S. and greater stickiness in SOEs than non-SOEs. Control variables generally load as expected (e.g., institutional ownership lowers synchronicity; larger size increases synchronicity). Robustness tests (alternative measures, fixed effects, GMM, PSM) uphold the main results.

The findings support the central hypothesis that cost behavior, specifically cost stickiness, materially affects the information efficiency of capital markets. Greater stickiness elevates earnings volatility and forecasting difficulty, leading investors to rely less on current earnings for valuation (lower ERC) and to incorporate less information about future earnings into current prices (lower FERC). The effect is attenuated for SOEs, consistent with higher regulatory influence, political objectives, soft budget constraints, and lower transparency that already complicate forecasting; additional stickiness has less marginal impact on FERC for SOEs. Elevated stickiness also increases stock price synchronicity, indicating that firm-specific information is less incorporated into prices, and that market/industry factors dominate. The mediation analyses demonstrate that reduced analyst forecast accuracy is an important channel linking cost stickiness to lower ERC/FERC and higher synchronicity. Collectively, the results highlight cost management behavior as a determinant of price informativeness, with implications for resource allocation efficiency in the market.

Using Chinese A-share manufacturing firms from 2009–2021, the paper shows that cost stickiness reduces the contemporaneous and future earnings response of stock returns (lower ERC and FERC), with the FERC effect concentrated in non-state-owned enterprises. Cost stickiness also raises stock price synchronicity, indicating reduced incorporation of firm-specific information. Analyst forecast accuracy partially mediates these relationships. The study extends the literature by linking a management accounting construct (cost behavior) to financial market information efficiency measures (ERC/FERC and synchronicity) and by documenting ownership-based heterogeneity. Policy and practice implications include enhancing disclosures related to cost structure and adjustability (e.g., capacity, labor contracts, equipment utilization) and investor education about cost stickiness to improve forecast accuracy and market information efficiency. Regulators might encourage or require disclosure for highly sticky-cost firms to mitigate information frictions.

- Generalizability: The sample focuses on Chinese A-share manufacturing firms (2009–2021) and excludes ST/*ST firms; results may not generalize to other sectors, markets, or periods. - Measurement constraints: Cost stickiness relies on sales as an imperfect proxy for activity and on total cost derived from accounting data; classification and pricing effects may introduce noise. Although alternative measures are tested, residual measurement error may remain. - Endogeneity and omitted variables: Despite firm fixed effects, GMM, and PSM, unobserved factors (e.g., supply chain dynamics, trade openness, governance practices) could jointly influence cost stickiness, forecasts, and price informativeness. - Analyst data limitations: Most analysts in China issue annual forecasts only, and consensus formation within a narrow pre-announcement window may not capture all information dynamics. - Interpretation of interaction signs: Because the stickiness measure is more negative for higher stickiness, coefficient interpretations hinge on coding; while addressed in the paper, this may complicate cross-study comparisons.