This subsection reports the empirical findings in two stages. First, we apply Natural Language Processing (NLP) to the Governor’s monthly speeches to construct sentiment (-1 to +1) and clarity (0 to 1) indices and describe how these evolve across policy cycles. Second, we use these communication measures in moderated regression models to test whether they dampen or amplify the effects of the BI-Rate and year-on-year CPI inflation on Jakarta Composite Index (IHSG) volatility.

The descriptive statistics in Table 2 show that all series exhibit sufficient time variation while remaining within plausible macro-financial ranges. Monthly IHSG volatility (σ) has a mean of 0.0085, suggesting generally stable equity returns in normal times. However, the maximum value of 0.0407, observed at the height of the COVID-19 sell-off, indicates that the market can become around five times more volatile when systemic risk surges, whereas the minimum of 0.0038 corresponds to periods of well-anchored expectations and benign macroeconomic conditions. The policy rate (BI-Rate, X₁) averages 5.46% (median 5.63%) and moves between 3.50% and 7.75% (σ = 1.36%), reflecting Bank Indonesia’s active adjustment of its main instrument to balance inflation and exchange-rate objectives. Inflation (X₂) averages 3.65%, with values between 1.32% and 8.36% (σ = 1.72 per cent), capturing both low-inflation environments and pronounced cost-push shocks. Taken together, these patterns indicate that the sample spans tranquil, tightening, and stress periods, providing a suitable setting to examine how communication interacts with monetary policy in shaping stock-market volatility.

On the communication front, the sentiment index (Z₁) is generally positive, with a mean of 0.17808 and σ = 0.10363, yet the lower bound of -0.11500 indicates moments when BI adopted an unmistakably cautious or even pessimistic tone, typically in the face of global turmoil or sharp domestic price spikes. Finally, the clarity index (Z₂) averages 0.73728, moves within a narrow band (0.50500-0.81000), and exhibits a modest standard deviation of 0.04388. This profile suggests that while the BI’s rhetoric occasionally dips in readability, it remains consistently high by international standards, providing a relatively stable informational channel through which tone can fluctuate. Collectively, these statistics justify the study’s two-stage approach: first interrogating how sentiment and clarity evolve, then examining how both attributes moderate the transmission of monetary and inflation shocks to stock-market volatility.

This section evaluates the informational content of Bank Indonesia’s policy statements over 2014-2024 using two linguistic dimensions extracted from the Governor’s monthly speeches: sentiment and clarity. Sentiment captures the tonal polarity of the message, while clarity reflects how easily an informed, non-specialist audience can process the text. Together, these attributes summarize the communication channel through which BI’s policy moves reach financial markets, and the subsequent sub-sections link their dynamics to salient macro-financial events.

The sentiment expressed in Bank Indonesia’s monetary policy communication over the 2014-2024 period exhibits clear temporal shifts, reflecting the central bank’s adaptation to evolving macroeconomic conditions. As quantified in this study, sentiment scores were extracted from Governor’s speeches and visualized in Figure 1.

The plot traces the monthly trajectory of sentiment scores over the eleven-year period. The scores range from a low of -0.12 (January 2014) to a high of 0.38 (January 2020), with most months post-2017 consistently recording values above 0.20. Early years are characterized by volatility and caution, with frequent neutral and negative readings: for instance, June and July 2015 registered near-zero or negative sentiment (-0.01 and -0.07, respectively), reflecting the macroeconomic pressures of that period.

A pronounced structural shift is observed from 2017 onwards, where almost all months exhibit positive sentiment. For example, November 2017 reaches a notably optimistic score of 0.35, and this trend of positivity persists into the subsequent years. The heatmap also reveals subtle seasonal effects, with January, April, and November frequently marking higher sentiment values. For instance, January 2017 and January 2020 scored 0.36 and 0.38 respectively, often coinciding with major policy announcements or economic outlooks.

During the COVID-19 shock, a period of sharply elevated global uncertainty, BI’s sentiment scores show an immediate response. Figure 1 shows that sentiment jumped to 0.38 in January 2020 but then dropped sharply in April (0.03) and May (0.03) as the economic impact of the pandemic became clearer. Despite these shocks, sentiment did not remain depressed; BI soon reverted to a more optimistic tone, with scores returning to the 0.10-0.30 range in the second half of 2020 and thereafter.

The overall trend, as visualized in Figure 1, demonstrates considerable fluctuation in the early years, with sentiment regularly oscillating between negative and positive values. Yet, from 2017 onwards, sentiment not only stabilized but also trended upwards, clustering in the positive zone. From 2021 to 2024, it stayed consistently positive, with monthly scores ranging between 0.15 and 0.35, except for minor dips that coincided with episodes of heightened external uncertainty, such as global inflation surges or political transitions (e.g., the slight decline at the end of 2024). The lowest score of -0.12 in January 2014 aligns with a period of significant economic strain, including rupiah depreciation and current account deficits. Conversely, the highest sentiment values, such as 0.38 in January 2020, occur in contexts where the central bank sought to reinforce public confidence or signal robust policy measures.

These detailed sentiment dynamics underscore the evolving role of BI’s communication as an active instrument in monetary policy. Figure 1 reveals a marked improvement in communication tone since 2017 and the agility with which BI adapts its messaging to macroeconomic developments. This shift, from cautious, sometimes negative sentiment in earlier years, to a more measured and positive outlook in recent years, demonstrates a strategic evolution towards greater transparency and consistency in Indonesia’s central banking communication.

Beyond sentiment, the clarity of central bank communication is closely linked to policy transparency, communication credibility, and the management of market expectations. This study quantifies clarity using linguistic metrics: average sentence length; lexical density; and the proportion of complex words, applied to the full text of BI’s Governor speeches from 2014 to 2024. Figure 2 shows that clarity is generally high but temporarily declines when new instruments are introduced or uncertainty spikes, before returning to its usual range once the policy path is better understood. These movements suggest that BI varies the density and complexity of its language to provide credible guidance and to anchor expectations as macro-financial conditions change.

The time series analysis reveals that clarity scores remained consistently high, with values typically ranging from 0.70 to 0.80 throughout the period. This suggests that BI maintained a generally clear and accessible communication style, even during episodes of economic turbulence. Notable exceptions, occasional sharp dips in clarity, such as in early 2015 (score around 0.58), mid-2018 (0.63), and mid-2020 (0.51), coincide with periods of heightened external shock or crisis. These temporary reductions in clarity may reflect the inherent complexity and uncertainty of policy responses during such times, as the bank grappled with communicating unprecedented challenges, such as global financial market volatility and the COVID-19 pandemic.

Despite these sporadic declines, clarity rapidly returned to its previous high levels. For instance, after the marked dip in 2020, corresponding to the initial pandemic shock, the clarity scores quickly rebounded above 0.70, reflecting BI’s commitment to transparent and comprehensible communication even in crisis management. The peak clarity scores, exceeding 0.80 (e.g., late 2017 and parts of 2018), correspond to periods where BI delivered especially concise and structured policy guidance, likely in response to positive macroeconomic developments or major strategic announcements.

Over the full observation window, the clarity scores display modest cyclical fluctuations but no significant long-term decline, indicating institutional learning and adaptation in crafting effective public messages. Clarity generally stays in the upper part of the 0-1 scale, with most observations clustered around 0.70-0.78, and only temporary dips during episodes of heightened external uncertainty, such as the monetary tightening episodes of 2014-2015 and 2018 or the COVID-19 shock. The overall high and stable level of clarity suggests that BI’s communication not only became more positive over time (as reflected in the sentiment scores) but also more systematically clear, which supports policy transparency, strengthens public trust and enhances the predictability of its monetary stance.

Taken together, the evidence from sentiment and clarity dynamics points to an increasingly deliberate communication strategy. More cautious tones tend to cluster around tightening phases, while more positive sentiment accompanies easing or more accommodative periods, indicating the use of language as a tool for expectation management. At the same time, the sustained clarity of BI’s messages, particularly after crisis-related dips, shows a consistent effort to keep policy objectives, risk assessments and strategic priorities understandable for a broad range of stakeholders, thereby reinforcing the institution’s broader credibility in monetary policy. Having established these dynamic patterns and the informational content of BI’s communication, the next section examines how these linguistic features interact with key macroeconomic variables by analyzing their impact on financial market volatility through regression analysis.

This section presents the results of a moderation regression analysis aimed at empirically identifying the influence of macroeconomic policy variables, namely, the interest rate and annual inflation, as well as the quality of Bank Indonesia’s communication, on Indonesian stock market (IHSG) volatility from January 2014 to December 2024. The analysis utilizes a monthly panel comprising 132 observations, with the dependent variable defined as monthly IHSG volatility (σ), and the main independent variables including the interest rate (X₁), annual inflation (X₂), and two key communication dimensions: sentiment (Z₁) and clarity (Z₂).

Prior to estimation, all independent and moderating variables were processed using mean-centering. This approach, which subtracts each observation from its respective variable mean, serves to mitigate potential multicollinearity among predictors and to facilitate the interpretation of regression coefficients, where the constant term represents IHSG volatility under average conditions of all predictors. Interaction terms were then constructed by multiplying the mean-centered principal variables (interest rate or inflation) with the mean-centered moderators (sentiment and clarity), enabling direct interpretation of the moderation effects when variables deviate from their means.

Two modeling strategies were applied. First, the Ordinary Least Squares (OLS) method with Huber-White (HC1) robust standard errors was employed as the baseline, given initial diagnostic tests (White test) indicated general heteroskedasticity in the residuals, particularly for Model (2). The HC1 adjustment ensures the robustness of standard errors, thus validating statistical inference despite the presence of heteroskedasticity. Second, as a robustness check, Robust Least Squares (RLS) estimation based on M-estimation with a bisquare weighting function was performed. This technique automatically reduces the influence of extreme observations, such as the significant volatility spike in March 2020 during the COVID-19 pandemic, thus ensuring that the regression results are not unduly distorted by outliers.

Table 3 shows that the BI policy rate in Model (2) has a negative and statistically significant coefficient (-0.0382, p < 0.05), indicating that higher policy rates are associated with lower IHSG volatility. Periods of policy tightening therefore tend to calm equity markets, consistent with the idea that decisive rate hikes can anchor inflation expectations and reduce uncertainty about the policy stance. In Model (3), the coefficient on inflation is also negative and significant (-0.0180, p < 0.10), suggesting that better-controlled inflation coincides with more stable stock-market conditions rather than with heightened turbulence.

The communication variables are likewise economically and statistically important. Sentiment enters with a negative and significant coefficient in both models (-0.0089 and -0.0090, p < 0.05), indicating that more positive language in BI speeches is linked to lower stock-market volatility and can reassure investors. Clarity also has a negative and highly significant coefficient (-0.0211 and -0.0262, p < 0.01), implying that, on average, clearer communication is associated with calmer markets. Taken together, these results suggest that both the tone and the readability of BI’s messages matter for how strongly markets react to macroeconomic news.

The interaction terms highlight how communication reshapes the impact of policy shocks. For the policy rate, the interaction with sentiment is positive but statistically insignificant, indicating that variations in the optimistic or pessimistic tone of communication do not materially change the market’s sensitivity to interest-rate movements. By contrast, the interactions between the policy variables and clarity are positive and strongly significant in both models (1.1071 and 0.9511, p < 0.01). This means that higher clarity magnifies the market response to changes in the BI-Rate and to movements in inflation. Intuitively, clarity acts as a double-edged sword: in normal times, clear communication reduces noise and volatility, but when policy moves are large or unexpected, very explicit messages make the signal harder to ignore, so price adjustments become sharper and short-term volatility can temporarily rise.

To ensure the validity and reliability of the regression results, this study conducted a series of robustness checks to mitigate the potential distortion caused by extreme observations or outliers. Robustness was assessed by comparing the baseline Ordinary Least Squares (OLS) estimates with Huber-White (HC1) robust standard errors to Robust Least Squares (RLS) estimates based on M-estimation with a bisquare weighting function, for both Model (2) and (3).

Diagnostic tests indicate that the specifications are statistically reliable. White’s test points to heteroskedasticity in Model (2) (p = 0.005), which is addressed through Huber-White (HC1) robust standard errors, while Model (3) does not exhibit heteroskedasticity (p = 0.139). The Durbin-Watson statistics (1.73 and 1.70), together with the Breusch–Godfrey LM tests, show no evidence of residual autocorrelation, and the maximum VIF values (4.68 and 3.28) are well below conventional thresholds, suggesting the absence of problematic multicollinearity. Robust Least Squares estimates with bisquare weighting confirm the OLS results: key coefficients preserve their signs and significance, and visual inspection of standardized residuals shows that only one extreme outlier around March 2020 is present and is effectively down-weighted. Overall, the coefficient patterns are stable across estimators, indicating that the main findings are not driven by estimation method or by a handful of extreme observations.

In sum, the regression evidence reveals that both monetary policy actions and the quality of Bank Indonesia’s communication exert significant influence on stock-market volatility. Higher policy rates and better-controlled inflation are associated with calmer markets, while more positive and clearer communication further dampens volatility on average. At the same time, the interaction results show that very explicit communication can amplify short-term market reactions to large or unexpected policy moves, underscoring the need for BI to balance transparency with flexibility when conveying its policy stance to financial markets.

The results of this study show that Bank Indonesia’s communication plays a strategic moderating role in shaping the impact of monetary policy, specifically interest rate and inflation changes, on stock market volatility. This is consistent with the communication analysis in section 3.1, which documented systematic shifts in both sentiment and clarity in response to evolving domestic and global conditions.

First, the sentiment analysis indicates that a more positive tone in BI’s policy messages is associated with lower stock market volatility. This pattern aligns with evidence that reassuring or optimistic central bank communication can mitigate uncertainty and dampen excessive market reactions ((Ehrmann & Talmi, 2020); (Nicolay & Oliveira, 2019)). Conversely, more negative or cautious language tends to heighten perceived risk and is linked to higher volatility, in line with studies showing that pessimistic communication increases risk premia and amplifies market stress (Bennani, 2020)(Correa et al., 2021).

Second, communication clarity exerts both direct and moderating effects. On average, clearer and more accessible BI statements are associated with calmer markets, supporting the view that simple, well-structured messages help investors interpret monetary policy more accurately and reduce noise in price formation (Bulir et al., 2012)(Vyshnevskyi et al., 2024). At the same time, the positive and significant interaction between clarity and policy variables suggests that clarity can behave like a double-edged sword. In normal times, higher clarity helps stabilize volatility, but when policy moves are large or unexpected, very explicit communication can increase the salience of adverse news and temporarily intensify market reactions ((Ferrara & Angino, 2021); (Vyshnevskyi et al., 2024)).

Third, the direct effects of the policy variables themselves are consistent with a stabilizing role for credible monetary policy. Higher BI-Rate levels and better-contained inflation are associated with lower IHSG volatility, indicating that decisive tightening and well-anchored inflation expectations reduce macroeconomic uncertainty and support more orderly price adjustments. These results are consistent with previous evidence that greater transparency and clearer central bank communication in emerging markets can reduce financial-market volatility and disagreement about inflation expectations ((Ahokpossi et al., 2020); (Bulir et al., 2012); (Nicolay & Oliveira, 2019); (Vyshnevskyi et al., 2024)).

Overall, the findings reinforce the view that central bank communication is not merely an auxiliary element of monetary policy, but an integral instrument with a tangible impact on financial market stability ((Ahokpossi et al., 2020); (Blinder et al., 2024); (Ehrmann et al., 2023); (Nicolay & Oliveira, 2019)). In the Indonesian context, high-quality communication, combining a measured positive tone with well-calibrated clarity, can help reduce uncertainty, shape how markets respond to interest rate and inflation shocks, and thereby support BI’s broader objective of maintaining financial stability.