Tag: Data Science / Quant

Data Spotlight: Dividend forecasts, market signals & more | Insights | Bloomberg Professional Services

[ad_1]

1. Short interest data to enhance fundamental signal

Factor investing can be strengthened by looking beyond traditional fundamentals to gain a clearer understanding of market trends. In this analysis, we explore how factor investing can be strengthened by looking beyond traditional fundamentals to gain a clearer understanding of market behavior. To do so, we incorporate short-interest data from S3 Partners, delivered through its collaboration with Bloomberg Enterprise Data. This dataset provides point-in-time coverage back to 2015 for more than 62,000 companies globally across a range of proprietary metrics.

Our analysis begins with a standard quality factor—consensus estimates for Return on Equity (ROE)—to identify high-quality companies. We then combine this with S3 short-interest metrics to highlight companies with more favorable market positioning. The study universe is the Bloomberg B500 Index over the past 10 years.

Each S3 variable is transformed into a daily-change factor, with bearish movements defined according to financial intuition: negative changes for Short Interest Availability and positive changes for all other measures. These bearish signals are intersected with the sector-level bottom quintile of ROE to construct the short-leg screen, applied point-in-time to the B500 constituents.

The long side maintains full, unfiltered exposure to the index. A standalone ROE backtest serves as a benchmark, allowing us to assess whether short-interest dynamics offer incremental value beyond fundamentals alone.

Chart 1 illustrates how combining Short-Interest Borrowing Cost with ROE can help uncover potential market opportunities.

Building on the results from the Borrow Cost + ROE signal shown in the Chart 1, the second chart broadens the perspective to evaluate all seven short-interest overlays within the same framework. While the initial analysis illustrated how a single signal can influence cumulative performance over time, this cross-sectional view compares the end-of-period outcomes for each combined factor as of 31 December 2024.

Applying the same long-index/short-screen methodology—where a bearish short-interest move is added on top of the bottom-quantile ROE filter—reveals a consistent pattern: six of the seven short-interest dynamics generated higher cumulative returns than the ROE-only benchmark.

Themes: Short Interest, Investment
Roles: Equity Portfolio Managers, Quants, Strategists
Bloomberg Datasets: S3 Partners Short Interest Data, Company Financials, Estimates and Pricing Point-in-Time

2. The value of dividend forecasts in modern investment decisions

Dividends, especially dividend forecasts, have long played a vital role in trading strategies, portfolio management and risk management practices in both sell-side and buy-side institutions. This has become even more relevant for institutional investors seeking to invest in markets that have been recently supported by dividend distribution policies, such as China.

A robust, forward-looking dividend forecast model with regulatory overlays to anticipate how companies are likely to adjust their dividend payouts could provide an edge in navigating today’s fast-changing financial markets and help investors make well-informed investment decisions.

In this study, we first look at how Bloomberg’s Dividend Forecast (BDVD) data can help with derivatives pricing. For derivatives desks, accurate dividend forecasts are essential because dividends directly affect option pricing, index futures, and structured products. Derivatives traders can adjust option fair values based on forecasted dividends rather than flat or assumed yields.

Chart 1 shows the index points difference for various A share and H share indices, representing the difference between projected and realized dividend. Our results show that Bloomberg dividend forecast could maintain an accuracy within roughly 10 basis points of the index level for a variety of China Indices.

For Buy-Side investors, Bloomberg’s Dividend Forecast (BDVD) data can provide additional insight to support dividend-focused strategies. In our analysis of a monthly-rebalance backtest from 2020 to 2025, companies in the top quintile of forecasted three year dividend growth exhibited higher cumulative returns than those ranked by three-year actual dividend growth.

This suggests that the market may place greater emphasis on the potential for dividend growth than on realized dividend payouts. Bloomberg’s dividend forecast data can offer portfolio managers deeper visibility into expected dividend trends, supporting more informed portfolio management decisions.

Themes: Quantitative Trading, Alpha Generation
Roles: Portfolio Managers, Quantitative Researchers, Derivatives Traders
Bloomberg Datasets: Dividend Forecast

Intraday liquidity for accurate backtests

When performing backtests, quantitative researchers often treat the closing auction price as an executable level. This approach implicitly assumes that sufficient liquidity is available in the auction to support execution at that price. There is extensive evidence that market impact cannot be ignored, as it varies significantly with trade size, market capitalization, and time of day, for example, see Direct Estimation of Equity Market Impact (Almgren et al). In practice, trading with minimal market impact during the closing auction typically requires that the order size represent less than 5% of the auction’s total volume.

As illustrated in Chart 1, smaller-cap companies may exhibit closing auction volumes that are insufficient to support execution without market impact. More realistic backtesting results may be achieved by incorporating pricing from different intraday windows, particularly those characterized by higher volume and lower volatility. For this purpose, we consider the volume-weighted average price (VWAP) observed during each selected session.

To support more effective execution, traders should take into account the available trading volume throughout the day, as this can also contribute to more realistic simulation results, especially in the context of trading smaller capitalizations. Chart 2 shows the distribution of available trading volume (right axis) across market-cap segments, alongside the realized 5-minute volatility (left axis) for each session. We observe that trading volume tends to be higher in the afternoon, while volatility is typically elevated in the morning around the market open. These patterns suggest that execution may be more favorable during the afternoon sessions.

This type of analysis enables tick-data users to conduct more granular assessments and can help refine the implementation aspects of their trading strategies, thereby supporting an improved risk-return balance.

Themes: Transaction Costs, Liquidity
Roles: Equity Portfolio Managers, Quantitative Researchers, Traders
Bloomberg Datasets: Tick History

How can we help?

Bloomberg’s Enterprise Investment Research Data product suite provides end-to-end solutions to power research workflows. Solutions include Company Financials, Estimates, Pricing and Point in Time Data, Operating Segment Fundamentals Data and Industry Specific Company KPIs and Estimates Data products, covering a broad universe of companies and providing deep actionable insights. This product suite also includes Quant Pricing with cross-asset Tick History and Bars. Additional solutions such as Geographic Segment Fundamentals Data, Company Segments and Deep Estimates Data and Pharma Products & Brands Data products will be available in 2025. All of these data solutions are interoperable and can be seamlessly connected with other datasets, including alternative data, and are available through a number of delivery mechanisms, including in the Cloud and via API. More information on these solutions can be found here.

Bloomberg Data License provides billions of data points daily spanning Reference, ESG, Pricing, Risk, Regulation, Fundamentals, Estimates, Historical data and more to help you streamline operations and discover new investment opportunities. Data License content aligns with the data on the Bloomberg Terminal to support investment workflows consistently and at scale across your enterprise.

[ad_2]

Bloomberg

Source link

December 5, 2025
Global insight: AI’s three revolutions for macro forecasting | Insights | Bloomberg Professional Services
[ad_1]
Filling data gaps

Economists have long mined text for research and forecasting. Early applications, from economic uncertainty (Baker et al. 2016) to central bank sentiment (Bulir et al. 2012), mainly relied on keyword counting.

Today, natural language processing and large language models unlock far richer signals. Sentiment scores of commentary in purchasing manager indexes can improve GDP nowcasts in combination with traditional data (de Bondt and Sun 2025). Topic modeling of news helps forecast aggregate stock market returns (Bybee et al. 2023). And language models turn headlines into forward-looking indicators of central bank policy — as with Bloomberg Economics’ indexes for the Federal Reserve, European Central Bank and Bank of England. These are available to Bloomberg Terminal subscribers via BECO MODELS CBSPEAK .

Satellite imagery is another important data source. Computer vision models can leverage interpretable features, such as roads, parking lots and crop fields, to provide information that official statistics either miss or capture too slowly. For example, we’ve built a monthly global GDP tracker using night-light data. Other researchers have found that retail parking occupancy measured from space can offer investors an edge (Katona et al. 2018), while snapshots of port-occupancy can help nowcast GDP and trade (Spelta et al. 2025).

Supercharged workflows
AI is changing how economists work as much as what they analyze.

Automation of routine tasks: Data cleaning, classification and feature engineering are increasingly handled by AI, freeing researchers to focus on interpretation and strategy.

Faster iteration: LLMs can increasingly act as research assistants — summarizing literature, coding econometric routines or stress-testing assumptions. Incorporating generative AI can cut project timelines from weeks to days (Korinek 2025).

Collaborative tools: AI-driven platforms integrate data, models and visualization, creating more transparent and reproducible research pipelines.

At Bloomberg Economics, we’re working faster across a broader suite of models, and spending more time contextualizing the results in our research.
Beyond linear models
AI is improving the predictive power of econometric models, enabling policymakers to be more proactive in monitoring financial risks. From classical machine learning to transformer-based foundation models, new methods capture signals — including nonlinear relationships — that can provide early warning systems for stress events.

Classical ML methods like Lasso, Ridge, and Elastic Net handle large datasets effectively, while tree-based ensembles such as Random Forest and Gradient Boosting capture nonlinearities that emerge during periods of high macroeconomic uncertainty (Aldasoro et al. 2025). Newer variants, such as the macro random forest, combine ML flexibility with the structure of economic models, outperforming traditional econometric techniques even when even when the time series are short (Chinn et al. 2023).

Deep learning models (neural networks) extend these gains to complex, high-dimensional tasks, such as predicting FX dislocations (Aquilina et al. 2025). These models excel at identifying nonlinear patterns and rare events, though they require substantial tuning (Athey and Imbens 2019).

Text-based large language models and NLP systems extract sentiment and information from unstructured sources like news, policy statements and corporate filings. This can boost forecast accuracy when combined with numerical data (de Bondt and Sun 2025). Similarly, time-series foundation models — such as TimeGPT (Nixtla), Moirai (Salesforce), and TimesFM (Google) — bring transformer architectures to economic forecasting. While they don’t yet outperform econometric mainstays like Bayesian VARs, combining their predictive flexibility with structural model discipline yields accuracy gains across variables and horizons (Carriero et al. 2025).

Although some studies have explored using LLMs directly as forecasters, their lack of a true “point-in-time” notion makes them unreliable for out-of-sample evaluation (Lopez-Lira et al. 2025). The frontier instead lies in hybrid approaches — blending the interpretability of structural economics with AI’s adaptive, data-driven strengths.

Central banks and international institutions are increasingly deploying these same AI models to anticipate financial stress. The ECB’s Cassandra system, for example, fine-tunes language models to analyze financial news sentiment and then applies boosting and neural network methods to flag early warning signals for banks (Petropoulos et al. 2025).

Similarly, BIS research (Aquilina et al. 2025) combines recurrent neural networks with LLMs to forecast and interpret episodes of FX market stress. The BIS system combines quantitative assessments with qualitative interpretation of financial news to predict stress events up to two months ahead.
[ad_2]

Bloomberg

Source link
October 28, 2025
Which ESG scores work best for portfolio construction? | Insights | Bloomberg Professional Services

[ad_1]

What are the key characteristics of Zero-Centered Scores?

Bloomberg ESG Scores measure best-in-class performance of a company’s management of financially material corporate sustainability issues. The issues deemed material in the Environmental (E) and Social (S) pillars are peer group-specific. The scores also factor in a company’s level of quantitative data disclosure. Since the average levels of company-reported data in the E and S Pillars vary considerably across industries, Bloomberg ESG Scores are not comparable across peer groups. That is, a score of 3, say, may indicate a laggard in one peer group, but an average company in another.

This is not an issue for analysts studying narrowly specified industries, but it poses a problem for portfolio managers or index providers whose tradable universe spans multiple sectors. To facilitate comparisons across broad market portfolios, practitioners often use peer group-specific percentiles as a way of identifying leading and lagging companies. Percentiles rank companies by their relative standing within a peer group, making them effective for filtering or screening exercises—for example, excluding the lowest 10% of companies from a portfolio.

Bloomberg’s Zero-Centered Scores, by contrast, go beyond ordinal ranking and provide an added element of the magnitude of a company’s outperformance or underperformance on sustainability relative to its peers, much like a Z-score. The Zero-Centered Score represents the difference between a company’s ESG Score and its peer group’s median ESG Score from the previous fiscal year, with the prior year’s median floored at 1.5. ZCSs can range from –10 to 8.5, with higher values indicating better outcomes.

The median company in a peer group has a ZCS near 0, outperforming companies have ZCSs greater than 0 and underperforming companies have ZCSs less than 0. Any two companies, from any peer groups, that have the same ZCSs can be considered to be performing equally relative to their specific peer averages. Corporate sustainability performance can thus be compared across all peer groups through this lens.

Each year’s peer-group medians are determined for an essentially fixed set of core companies. This provides year-over-year stability of ZCSs by virtue of a time series that is more robust to changes in the overall scoring universe (additions, removals etc.) than a time series typically constructed using Percentiles or ranks. This feature is particularly valuable for analyses of score changes over time (e.g., identifying improvers).

Percentiles and ZCSs have different scales. Percentiles span 0 to 100, and ZCSs can range from –10 to 8.5, though in practice the range is approximately -4 to 4. Nevertheless, these two metrics are highly correlated since ZCSs preserve the ordinal information captured by Percentiles. For many types of analysis, investors could use either measure and obtain similar results.

To illustrate this, we use point-in-time ZCS and Percentiles data retrieved via Bloomberg Query Language (BQL) for the subsequent analysis. We reproduce a chart we presented (as Figure 5a) in our earlier article and show it as Figure 1a here. It shows the historical returns and Sharpe ratios of quintile portfolios formed by sorting on ZCSs of companies in the Bloomberg WORLD Index that have High or Average levels of quantitative data disclosure, as defined in the previous article.

Figure 1b shows results for the same set of companies, but for quintile portfolios formed on Percentiles. In both cases, the results are similar: the quintile portfolios of companies with better sustainability performance (i.e., higher ZCSs or Percentiles) exhibited higher returns than those with worse sustainability performance. Though not shown here, the same pattern is seen in market value-weighted quintile portfolios.

To understand how Percentiles and ZCSs differ we examine how their values are distributed. Figures 2a and 2b show histograms of the distribution of all companies that have Bloomberg ESG Scores in June 2025, using ZCSs and Percentiles as the ESG–score metric, respectively. Percentiles, by definition, follow a uniform distribution, with approximately the same number of companies in each quantile.

By contrast, the ZCS distribution is bell–shaped, with a concentration of companies near a ZCS of –1 and very few companies with very low (–4) or very high (4) ZCSs. This reflects that few companies underperform or outperform their peer averages by a significant amount. Thus, in this example, ZCSs distinguished marginally better performance from exceptional outperformance and could have helped portfolio managers calibrate portfolio tilts.

The scatter plot in Figure 3 makes the differences in the distributions more evident. The two metrics are highly correlated and follow a linear trend for the most part. However, there is some dispersion of ZCSs at any given Percentile.

How do Zero-Centered Scores improve portfolio optimization?

We now present the results of two portfolio optimization exercises that used Zero-Centered Scores and Percentiles as their ESG signals, respectively. Once more, we limit our universe to companies in the Bloomberg WORLD Index that have ESG Scores based on High or Average quantitative data disclosure.

We utilized Bloomberg’s PORT Optimizer and Bloomberg’s Multi-Asset Class Fundamental risk model (MAC3) to maximize each portfolio’s ESG signal (ZCS or Percentile, respectively) while limiting ex-ante annualized tracking error volatility (TEV) to the WORLD Index to 3% and simultaneously constraining total active factor risk exposures to near zero.

This allowed us to create two portfolios that closely track the WORLD Index benchmark while varying individual security weights to maximize the ESG signal (ZCS or Percentile). Additionally, we utilized the risk model to do this in a manner that prevents any incidental active risk factor exposures—such as country, industry or style (e.g. momentum, value)—between the portfolio and the benchmark. Thus, any differences in performance between the two portfolios and the benchmark index should have been due primarily to the effect of security selection effects resulting from the use of different sustainability metrics. Note that for a more comprehensive description of the “Selection Effect”, please see the return attribution analysis in our prior blog post.

Figure 4a summarizes portfolio performance statistics relative to the benchmark, Figures 4b and 4c show the portfolio performances for the period from 17 March 2017 – through 30 June 2025.

In the back-tests, the portfolio optimized to maximize the Zero-Centered Score (ZCS) delivered an annualized return of 11.68%, outperforming the benchmark by 0.52% annualized over the period. By contrast, the Percentile-optimized portfolio largely tracked the benchmark and did not show sustained outperformance.

These results exclude transaction costs; adding turnover constraints or other cost controls would likely reduce realized excess returns. Given identical tracking–error limits and near-zero active factor-risk constraints for both portfolios, the performance gap most likely reflects the incremental sustainability-related information captured by ZCS rather than differences in factor exposures.

Key takeaways: ESG score selection makes a difference for portfolio construction

For investors, the choice of inability metric matters. Peer Group Percentiles are simple and effective for screening, but they can fall short when applied in portfolio construction. Zero-Centered Scores, by contrast, provide richer information that enables more stable comparisons across industries and time, and—as the back–tests showed—could enhance portfolio performance. Investors looking to integrate sustainability considerations into systematic processes may therefore benefit from relying on ZCS as their primary input. Put simply, when it comes to ESG scores, measuring how much better or worse a company is than its peers can make a difference.

Disclaimer

Nothing in the Services shall constitute or be construed as an offering of financial instruments by Bloomberg, or as investment advice or recommendations by Bloomberg of an investment strategy or whether or not to “buy”, “sell” or “hold” an investment. Information available via the Services should not be considered as information sufficient upon which to base an investment decision. Bloomberg makes no claims or representations, or provides any assurances, about the sustainability characteristics, profile or data points of any underlying issuers, products or services, and users should make their own determination on such issues. All rights reserved. ©Bloomberg.

[ad_2]

Bloomberg

Source link

October 1, 2025
Data Spotlight: Revenue surprises, tariffs impact & more | Insights | Bloomberg Professional Services

[ad_1]

1. Tariffs: Using supply chain and facilities data to evaluate market shifts in North America

In an earlier study, we looked at the two weeks following the 2024 U.S. election and found that companies with U.S.-centric supply chains outperformed their globally exposed peers. That analysis highlighted how supply chain data can reveal investor sentiment during periods of policy change.

In this expanded study, we extend the analysis through July 2025 and across North America markets to capture the effects of the tariff war. Using Bloomberg’s Supply Chain and Facilities datasets, we grouped companies of North America based on presence of US operational exposure or not and examined how performance diverged since the US elections and around key trade announcements.

Chart 1 shows the overall average US operational exposure for a selection of Bloomberg Indices for North America. Outside the US, Canada and Mexico are quite exposed countries to the U.S. making them sensitive to any change of trade policy.

Focusing first on U.S. domiciled companies, Chart 2 shows relative performance of companies with high exposure to the U.S. versus those with low US exposure for companies of the Bloomberg US 1000 Index (B1000) comprised of the largest 1,000 market capitalization in the US. Interestingly, companies with very high U.S. exposure have been performing well from the US Election to April 2025 as the Policy of the Administration has supported US based industries. However after April 2, Liberation Day tariffs, we note a peak and a reversal of this movement – probably highlighting the administration willing to make deals that can be profitable to US companies with global footprint.

In addition to U.S. companies, we have examined the rest of North America’s largest markets (Canada and Mexico). Chart 3 summarizes the cumulative performance of companies exposed to the US against those without US exposure: it appears that U.S. trade war is translating into negative equity returns for companies in their neighbor doing business with them.

Looking beyond North America, we observe a consistent trend globally (Chart 4).

Themes: Macro Investing, Tariff
Roles: Equity Portfolio Managers, Quants, Strategists
Bloomberg Datasets: Supply Chain, Facilities

2. Tracking when guidance moves markets: the Japanese case

During each earnings season, companies release actual financial results and often provide forward-looking guidance for upcoming quarters or the full fiscal year. While markets – and especially systematic players – have traditionally focused on the difference between reported earnings and consensus expectations because of a lack of availability of company guidance in a machine readable format, our research underscores the increasing importance of monitoring guidance surprises — instances where a company’s outlook materially deviates from market forecasts.

To explore this further, we use Bloomberg’s Company Financials, Estimates and Pricing Point-in-Time dataset to examine the frequency of earnings guidance issuance across various regional indices (Chart 1). The findings reveal that companies in Japan are significantly more likely to provide forward EPS guidance compared to their counterparts in the U.S., China, and Europe — highlighting a notable regional difference in corporate disclosure practices.

We further used the data from the Japanese equity market to examine how equity markets respond to earnings guidance surprises — defined as the difference between a company’s issued EPS guidance and the consensus EPS estimate for the next fiscal year.

Our findings (Chart 2) show that positive guidance surprises tend to yield immediate next-day positive performance, with the magnitude of the surprise closely correlated to the size of the price move. In contrast, negative guidance surprises tend to trigger immediate declines in stock price — even when reported results exceed expectations.

This shows that investor sentiment can be more sensitive to forward-looking outlook than to trailing performance, with guidance acting as a forward-looking shock that reshapes market expectations and valuations.

Bloomberg Company Financials, Estimates and Pricing Point-in-Time product provides a comprehensive, point-in-time history of company-reported metrics, consensus estimates, and management guidance as well as pricing information. This data enables investors to backtest stock performance accurately around earnings releases, helping investors understand how actuals, consensus estimates, and company-issued guidance interact to drive market reactions.

Themes: Quantitative Trading, Alpha Generation
Roles: Equity Portfolio Managers, Quantitative Researchers, Traders
Bloomberg Datasets: Company Financials, Estimates and Pricing Point-in-Time

3. Analyzing transaction data analytics and estimates to anticipate earnings surprises

Analysts estimates set investor expectations for a company’s performance each period, and earnings surprises often trigger significant stock price movements. If company performance trends can be evaluated ahead of earnings release, it may create opportunities to identify and respond to surprise-driven price actions. Using Bloomberg Second Measure’s near real-time transaction data (available on a 3-day lag via feeds), investors can gain early insights into company performance well before official reports. When combined with consensus estimates from Company Financials, Estimates and Pricing Point-in-Time dataset, it empowers investors to build actionable trading strategies.

In our study, looking at a quarterly rebalanced backtest from Q2 2020 to Q1 2025 we see that companies in the top quintile of revenue surprises—where transaction data analytics from Bloomberg Second Measure show stronger sales than market expectations—generated higher cumulative returns than those in the bottom quintile.

The Quintile 1 basket delivered robust long-term performance, while the Quintile 5 basket (representing the most negative surprises) showed lower performance. A long–short strategy that takes a long position in Quintile 1 and shorts Quintile 5 produced modest but consistent gains, reinforcing the idea that upside surprises offer a stronger signal than downside disappointments.

As shown in Chart 2, there are a variety of sectors covered in this analysis. This type of analysis can be refined based on a dedicated sector analysis: indeed this type of strategy may perform differently according to the industry.

These results underscore the value of alternative data might have in anticipating market-moving fundamentals before official disclosures.

Themes: Equity Fundamentals, Alpha Generation
Roles: Equity Portfolio Managers, Quantitative Researchers, Traders
Bloomberg Datasets: Company Financials, Estimates and Pricing Point-in-Time, Bloomberg Second Measure

How can we help?

Bloomberg’s Enterprise Investment Research Data product suite provides end-to-end solutions to power research workflows. Solutions include Company Financials, Estimates, Pricing and Point in Time Data, Operating Segment Fundamentals Data and Industry Specific Company KPIs and Estimates Data products, covering a broad universe of companies and providing deep actionable insights. This product suite also includes Quant Pricing with cross-asset Tick History and Bars. Additional solutions such as Geographic Segment Fundamentals Data, Company Segments and Deep Estimates Data and Pharma Products & Brands Data products will be available in 2025. All of these data solutions are interoperable and can be seamlessly connected with other datasets, including alternative data, and are available through a number of delivery mechanisms, including in the Cloud and via API. More information on these solutions can be found here.

Bloomberg Data License provides billions of data points daily spanning Reference, ESG, Pricing, Risk, Regulation, Fundamentals, Estimates, Historical data and more to help you streamline operations and discover new investment opportunities. Data License content aligns with the data on the Bloomberg Terminal to support investment workflows consistently and at scale across your enterprise.

[ad_2]

Bloomberg

Source link

September 23, 2025