## 3.1    The IMF Data Standards Initiatives

## Source details

**Canonical URL:** [3.1    The IMF Data Standards Initiatives](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025236-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025236-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025236-source-pdf.pdf.json)

---

### Framework and monitoring
- Operates across three tiers: the Enhanced General Data Dissemination System (e-GDDS), the Special Data Dissemination Standard (SDDS), and the SDDS Plus.
- Almost all IMF members participate in one of the three tiers, depending on statistical capacity.
- Tiers increase in stringency of dissemination recommendations and requirements; SDDS and SDDS Plus subscribers face more rigorous IMF monitoring.
- Performance under SDDS and SDDS Plus is publicly disclosed in Annual Observance Reports (AORs) which evaluate:
  - (i) data coverage;
  - (ii) the timeliness of data releases against standard-implied requirements and release lags;
  - (iii) the punctuality of data releases against self-committed, pre-announced data release commitments;
  - (iv) data dissemination through a National Summary Data Page (NSDP);
  - (v) metadata certification.
- AORs are publicly available on the IMF’s Dissemination Standards Bulletin Board and form the basis for the measurement concept used in the index construction.

### Source data used for indices
- AORs contain monthly information on data releases of real, fiscal, financial, and external sector data under two criteria:
  - (i) timeliness as the difference between the end of a reference period and the date of data dissemination;
  - (ii) punctuality as the difference between an economy’s Advance Release Calendar (ARC) date and the date of data dissemination.
- Under the punctuality criterion, both delays and early releases deviate from disciplined dissemination practices.
- NSDPs are used as the standardized IMF-monitored dissemination platform across all subscribers.
- Using AOR data from 78 subscribers, the authors construct a global composite data timeliness index and regional and domain indices available on a monthly basis since 2006.
- Examples of StandardReleaseLag_cat entries (from AOR tables) include: National accounts 90; Production index 45; Employment 90; Consumer prices 30; Central bank survey 15; Official reserve assets 7; Merchandise trade 60; IIP 90; External debt 90; and various financial-sector indicators with 90 or 120-day lags.

### Index construction: principles and formulas
- Four core principles guide construction:
  - (i) use publicly disclosed, regularly compiled records based on observable dissemination performance that enable cross-country comparison;
  - (ii) flexibility to accommodate varying subscription dates, different release frequencies (annual/quarterly/monthly), and changing index composition as subscribers progress;
  - (iii) international scope through broad geographic coverage and standardized measurement;
  - (iv) consistent benchmarking by relying on unchanged timeliness requirements, providing consistent measurement over time.
- Timeliness and punctuality indices per economy i, data category cat, and period t are constructed as:

  Timeliness:
  DT I(i,cat,t) =
  StandardReleaseLag_cat + Delay(i,cat,t)
  StandardReleaseLag_cat ·100    if 0≤|Delay(i,cat,t)|<StandardReleaseLag_cat
  0 if |Delay(i,cat,t)|≥StandardReleaseLag_cat

  Punctuality:
  DP I(i,cat,t) =
  StandardReleaseLag_cat − |Deviation(i,cat,t)|
  StandardReleaseLag_cat ·100    if 0≤|Deviation(i,cat,t)|<StandardReleaseLag_cat
  0 if |Deviation(i,cat,t)|≥StandardReleaseLag_cat

- Definitions and interpretation:
  - StandardReleaseLag_cat refers to the standard-implied release lag per SDDS/SDDS Plus data category.
  - For DTIs, Delay(i,cat,t) reflects actual delays in days; for DPIs, Deviation(i,cat,t) reflects deviations from schedule (absolute deviations capture both early and late releases).
  - The timeliness-based DTI caps reporting performance at 100 percent; early reporting is not rewarded (reflects timeliness versus accuracy trade-off).
  - No delays and schedule-adhering releases = 100 percent score.
  - Delays up to an extra standard-implied reference period produce linearly decreasing scores to 0 percent.
  - |Delay| ≥ StandardReleaseLag_cat (i.e., delays beyond two times the standard-implied release lag) are scored 0 percent, interpreted as releases past the next data release due date.
  - Administrative AOR markers mapped to numerical scores: ‘X’ (not disseminated) and ‘O’ (missing ARC entries) correspond to 0 percent (late dissemination past next release date). ’NYD’ and ’P’ treated as timely/punctual in index construction to avoid end-of-period bias.

### Aggregation to global and regional indices
- Composite indices constructed as cross-country means; aggregates also produced for four statistical domains (real, fiscal, financial, external) and five IMF regions: Sub-Saharan Africa (AFR), Asia/Pacific (APD), Europe (EUR), Middle East/Central Asia (MCD), and the Western Hemisphere (WHD).
- All subscribers and data categories contribute with equal weight; aggregates expand over time with increasing subscriptions.
- As of end-2023, 78 economies have subscribed to SDDS and SDDS Plus and are considered in aggregation.
- Sensitivity to changing subscriber composition is investigated in appendix E.

### Stylized facts: global composite DTI (monthly series since 2006)
- Key level and trend facts:
  - Global DTI levels are generally high, ranging on average across subscribers between 92 and 99 percent from 2006 to 2023.
  - The index declined over time: first ten years characterized by high and relatively stable DTI levels; level dropped noticeably and persistently thereafter. Since 2017, the index is exposed to higher volatility.
  - More recently, it recovered though without fully recovering to pre-2017 levels.
  - Despite crises, the global DTI did not drop below 90 percent on average across subscribers, suggesting resilience; post-crisis improvements are becoming harder to maintain.
- Comovement with global disruptions:
  - Decreases and higher volatility during crises: GFC (end-2007—mid-2009); euro area crisis (end-2010—end-2012); Arab Spring (end-2010—mid-2011); first year of Covid-19 pandemic (Dec 2019—end-2020); early phase of Russia’s invasion of Ukraine (Feb 2022—July 2022).
  - COVID-19 induced a sharp decline followed by a V-shaped recovery; 2022 shows a distinct drop around Russia’s invasion of Ukraine.
  - Since the pandemic, increased comovement with global economic conditions observed (two-year rolling correlations based on Hamilton-filtered series increased up to about 0.75 by mid-2023).
- Trend-cycle decomposition:
  - Hamilton-type decompositions (appendix D) show declining trend especially after 2017 and widening, more persistent negative gaps since then.

### Data timeliness by statistical domain
- Domain-level findings:
  - Real sector data are released most timely.
  - Real, external, and financial subindices tend to track each other closely.
  - Dips in timely financial data dissemination tend to be larger during crises associated with financial distress (e.g., GFC, European sovereign debt crisis).
  - Fiscal data tends to be reported less timely; deterioration in fiscal timeliness is especially pronounced since 2017.
  - Differences among subindices have widened over time.
- Fiscal decentralization effect:
  - Economies grouped by tax revenue decentralization (IMF Fiscal Decentralization Dataset) show that decentralization is associated with longer delays in fiscal data dissemination.
  - Classification threshold: 0.29 percent of tax revenue decentralization as share of general government separates relatively decentralized (14 economies as of 2020 or latest) from relatively centralized economies (49 economies).
  - Lower fiscal DTI values on average for relatively decentralized economies; interagency coordination needs likely contribute to longer delays.
- Fiscal transparency relevance:
  - Greater fiscal transparency has been associated in the literature with better credit ratings, fiscal discipline, and lower sovereign financing costs for EMEs and DEs and AEs (literature cited in source).

### Data timeliness by economy groupings and regions
- AEs vs EMEs:
  - Composite DTI divergence since 2017: persistent gap opened up with AEs showing greater resilience and recovering to pre-2017 levels after the pandemic, while EMEs experienced more lasting impacts from Covid-19 and recent geopolitical tensions.
  - Over 2023 the gap narrowed considerably.
  - Before 2017, EMEs sometimes outperformed AEs (e.g., during European sovereign debt crisis).
- SDDS vs SDDS Plus:
  - SDDS Plus reporting performance started on a higher level in 2015; pronounced decline during 2017/18 documented for SDDS Plus as well, with subsequent settling on a lower level.
  - SDDS Plus established in 2012; first eight countries adhered in 2015; as of end-2023, 29 economies adhere to SDDS Plus of which 24 are AEs.
- Regional patterns:
  - Pre-2017 comovements across regions were high and homogeneous; regional divergence increases after 2017.
  - Pandemic impacted all regions with varying severity:
    - EUR subscribers showed the most resilient performance on average.
    - AFR, MCD, and WHD subscribers on average were more vulnerable, especially during the pandemic.
    - APD subscribers track the composite index closely.
    - Volatility highest for AFR group (fewest subscribers).
  - MCD subscribers showed the most persistent decline starting around 2015; WHD subscribers show most relative improvements since 2018.

### Composition and robustness (Appendix E)
- Main robustness findings:
  - Mean-based DTIs are importantly driven by weak performances at the bottom of the monthly distributions; downward spikes are driven by the bottom three to eight subscribers.
  - Key stylized facts remain robust to alternative index compositions and extending the set of economies with new subscribers does not materially drive global DTI dynamics.
  - Regional DTI observations can be very high, "often at or close to 100 percent".
- Distributional details:
  - More than 50 percent of subscribers maintain consistently timely data releases each month.
  - The persistent downward shift in performance since 2017 is visible up to the 25th percentile (up to about 19 economies) and pulls the mean toward the 25th percentile.
  - The 10th and 25th percentiles exhibit persistent declines since 2017; the 5th percentile shows recovery tendencies but sharp declines around periods of distress.
- Alternative composition tests:
  - Average performance among the initial set of subscribers and average performance among the SDDS Plus adherents as of 2023 backcasted show remarkable consistency in overall trends.
  - Addition of 16 new SDDS subscribers did not materially drive global DTI dynamics.

### Tighter fiscal stances, constrained resources, and donor funding
- Fiscal resources devoted to statistical services posited as a potential explanation for DTI dynamics; timely publication requires dedicated statistical services and resources.
- Budgetary envelope trends:
  - Overall own and donor-funded budgetary envelopes have shrunk with adverse implications for statistical services.
  - General government total and general public service expenditures grew during the GFC (end-2007—mid-2009) and the Covid-19 pandemic (Dec 2019—end-2020), but both shrank during the intervening years and "both shrank the most by 2016/17," suggesting constrained resources prior to the drop in the DTI in 2017/18.
- Donor funding patterns:
  - Donor funding for statistical capacity building directed toward EMEs declined substantially between 2013 and 2017, with a notable bounce-back in 2022 that coincides with a narrowing gap between AE- and EME-DTIs after 2022.
  - The donor-funding indicator covers 38 of the 78 economies in the study and predominantly covers EMEs in the sample.
- Evidence of underfunding and shrinking inflation-adjusted budgets for statistical offices cited from multiple sources (Grant 2012; Hartman et al. 2014; The Globe and Mail 2014; West 2017; Groshen 2021; UKSA 2022; Campbell 2023; American Statistical Association 2024; The Economist 2025).
- Low prioritization tendencies may constrain modernization and innovation efforts needed in the age of data science and AI (American Statistical Association 2024).

### Empirical applications — methods overview
- Investigates predictive relationships between DTIs and public finance outcomes and tests early warning properties for economic downturns.
- Methodology highlights:
  - Two-variable panel VARs and Granger causality tests study relationships between year-on-year changes in economy-level fiscal DTIs and changes in general public service expenditures, and separately with changes in central government debt (CGD).
  - ROC curves from panel logit models evaluate early warning properties for business cycle recessions and systemic crises.
  - Business cycle recessions derived from the Bry-Bochan algorithm on real GDP series from the IMF’s WEO April 2024 database for the sample 2006Q2-2023Q4.
  - Four years of lagged DTI-growth are considered (i.e., 16 quarters).

### Fiscal DTIs and public finance outcomes — key findings
- Temporal sequencing and predictive relationships:
  - Changes in general public service spending contain predictive information for fiscal DTIs; past changes in general public service spending help forecast fiscal DTIs while the reverse is not found.
  - The third lag of changes in general public service spending (three years ago) adds the most information and reflects a positive correlation with changes in fiscal data dissemination in period t.
  - This predictive relationship is significant only with respect to fiscal data dissemination, not for other sub-DTIs or the composite DTI.
  - Changes in past fiscal DTIs improve CGD growth forecasts across all sub-DTIs and the composite DTI; coefficient signs are predominantly intuitively negative, suggesting improving data release efficiency helps forecast future decreases in CGD.
  - The Granger causality Walt test statistic rejects the null hypothesis of no Granger causality at the 1% level. VAR coefficients in both specifications are estimated to be significant up to the first three own lags.

### Early warning properties — quantitative ROC/AUROC results (Appendix F highlights)
- Baseline ROC results (composite DTI as main explanatory variable; four years of lags included):
  - Business cycle recessions (sample 2006Q2-2023Q4): Area under ROC curve = 0.7221.
  - Systemic crises (sample 2006Q2-2019Q4): Area under ROC curve = 0.6855.
  - Both ROC curves lie above the 45-degree diagonal (random classifier), indicating DTIs contain useful predictive information.
- Selected Table 2 baseline and alternative logit specification results (preserving coefficients and AUROC values exactly):
  - Baseline: DTI growth – 16 quarters
    - Coefficient: -0.244***
    - AUROC: 0.722[0.7010.743]
    - Systemic crises coefficient: -0.009*** AUROC: 0.686[0.6110.760]
  - DTI growth: 12 quarters
    - Coefficient: -0.082*** AUROC: 0.714[0.6940.734]
    - Systemic crises coefficient: -0.071*** AUROC: 0.678[0.6080.748]
  - DTI growth: 20 quarters
    - Coefficient: -0.016*** AUROC: 0.728[0.7070.749]
    - Systemic crises coefficient: -0.543*** AUROC: 0.725[0.6530.796]
  - DTI and real GDP growth (16 quarters)
    - Coefficient: -0.207*** AUROC: 0.777[0.7580.796]
    - Systemic crises coefficient: -0.142*** AUROC: 0.750[0.6810.818]
  - DTI and inflation (16 quarters)
    - Coefficient: -0.263*** AUROC: 0.735[0.7140.756]
    - Systemic crises coefficient: -0.434*** AUROC: 0.784[0.7240.844]
  - DTI and private credit growth (4 years) (annual)
    - Coefficient: -0.027*** AUROC: 0.735[0.6980.772]
    - Systemic crises coefficient: -0.131*** AUROC: 0.864[0.7400.988]
  - DTI and public debt growth (4 years) (annual)
    - Coefficient: -0.023*** AUROC: 0.737[0.7020.772]
    - Systemic crises coefficient: -0.040*** AUROC: 0.791[0.6500.932]
  - Real subindex (16 quarters)
    - Coefficient: -0.257*** AUROC: 0.721[0.6700.742]
    - Systemic crises coefficient: -0.198*** AUROC: 0.680[0.6020.757]
  - Fiscal subindex (16 quarters)
    - Coefficient: -0.033*** AUROC: 0.717[0.6960.739]
    - Systemic crises coefficient: -0.243*** AUROC: 0.711[0.6340.787]
  - Financial subindex (16 quarters)
    - Coefficient: -0.090*** AUROC: 0.719[0.6980.740]
    - Systemic crises coefficient: -0.040*** AUROC: 0.687[0.6110.763]
  - External subindex (16 quarters)
    - Coefficient: -0.102*** AUROC: 0.720[0.6990.741]
    - Systemic crises coefficient: -0.774*** AUROC: 0.693[0.6170.769]
- Additional AUROC highlights reported:
  - DTI and Real GDP Growth (2006Q2-23Q4) — Area under ROC curve = 0.7770 (business cycle recessions).
  - DTI and Inflation (2006Q2-23Q4) — Area under ROC curve = 0.7354 (business cycle recessions).
  - DTI and Private Credit Growth (annual: 2007-22) — Area under ROC curve = 0.7348 (business cycle recessions).
  - DTI and Public Debt Growth (annual: 2007-22) — Area under ROC curve = 0.7368 (business cycle recessions).
  - DTI and Real GDP Growth (2006Q2-19Q4) — Area under ROC curve = 0.7495 (systemic crises).
  - DTI and Inflation (2006Q2-19Q4) — Area under ROC curve = 0.7842 (systemic crises).
  - DTI and Private Credit Growth (annual: 2007-19) — Area under ROC curve = 0.8642 (systemic crises).
  - DTI and Public Debt Growth (annual: 2007-19) — Area under ROC curve = 0.7914 (systemic crises).
- Subindex-specific signaling insights:
  - Business cycle recessions: real and external DTIs most informative (baseline subindex AUROCs around 0.7208 (Real), 0.7174 (Fiscal), 0.7187 (Financial), 0.7198 (External)).
  - Systemic crises: fiscal DTIs show strongest predictive ability (baseline subindex AUROCs around 0.6797 (Real), 0.7106 (Fiscal), 0.6867 (Financial), 0.6928 (External)).
  - Sovereign debt crises (subtype analysis): AUROC by subindex: Real = 0.6017; Fiscal = 0.8593; Financial = 0.8794; External = 0.8390.

### Stylized facts and policy-relevant implications
- Stylized facts:
  - Notable shift toward longer delays in official data release since 2017, particularly pronounced among EMEs, leading to a widening AE–EME DTI gap.
  - Real sector data is consistently published more timely than other data; fiscal data dissemination faces persistent challenges.
  - More coordination-intense government decentralization has, on average, been associated with longer delays in recent years.
  - Timely dissemination of official data tends to decrease during episodes of global distress.
- Policy implications:
  - Constrained fiscal resources devoted to statistical services can help explain deteriorations in data dissemination timeliness.
  - Maintaining robust data transparency frameworks and subscriber compliance with the IMF Data Standards Initiatives remains important.
  - Resilient compliance and sustained institutional capacity and sufficient resources are required to ensure timely data release when information is most needed to resolve macroeconomic imbalances and market disequilibria.
  - Decreasing compliance can be accompanied by challenges related to fiscal outcomes and economic downturns; maintaining donor and domestic funding for statistical capacity building is relevant for EMEs.

*Source: wpiea2025236-source-pdf (Appendices and sections quoted as provided in the supplied content).*

### 3.1    The IMF Data Standards Initiatives

### 3.1    The IMF Data Standards Initiatives

### The IMF Data Standards Initiatives: framework and monitoring
- Operates across three tiers: the Enhanced General Data Dissemination System (e-GDDS), the Special Data Dissemination Standard (SDDS), and the SDDS Plus.
- Almost all IMF members participate in one of the three tiers, depending on statistical capacity.
- Tiers increase in stringency of dissemination recommendations and requirements; SDDS and SDDS Plus subscribers face more rigorous IMF monitoring.
- Performance under SDDS and SDDS Plus is publicly disclosed in Annual Observance Reports (AORs) which evaluate:
  - (i) data coverage;
  - (ii) the timeliness of data releases against standard-implied requirements and release lags;
  - (iii) the punctuality of data releases against self-committed, pre-announced data release commitments;
  - (iv) data dissemination through a National Summary Data Page (NSDP);
  - (v) metadata certification.
- AORs are publicly available on the IMF’s Dissemination Standards Bulletin Board and form the basis for the measurement concept used in the index construction.

### Source data used for indices
- AORs contain monthly information on data releases of real, fiscal, financial, and external sector data under two criteria:
  - (i) timeliness as the difference between the end of a reference period and the date of data dissemination;
  - (ii) punctuality as the difference between an economy’s Advance Release Calendar (ARC) date and the date of data dissemination.
- Under the punctuality criterion, both delays and early releases deviate from disciplined dissemination practices.
- Table 1 (SDDS and SDDS Plus Timeliness in Days) lists StandardReleaseLagcat values for data categories; items in square brackets indicate categories additionally covered under SDDS Plus. (Examples of StandardReleaseLagcat entries in the table include: National accounts 90; Production index 45; Employment 90; Consumer prices 30; Central bank survey 15; Official reserve assets 7; Merchandise trade 60; IIP 90; External debt 90; and various financial-sector indicators with 90 or 120-day lags.)
- NSDPs are used as the standardized IMF-monitored dissemination platform across all subscribers.
- Using AOR data from 78 subscribers, the authors construct a global composite data timeliness index and regional and domain indices available on a monthly basis since 2006.

### Index construction: principles and formulas
- Four core principles guide construction:
  - (i) as data source publicly disclosed, regularly compiled records based on observable dissemination performance that enable cross-country comparison;
  - (ii) on flexibility through accommodating varying subscription dates, handling different release frequencies (annual/quarterly/monthly), and adapting to changing index composition as subscribers progress;
  - (iii) on providing international scope through broad geographic coverage and standardized measurement;
  - (iv) on consistent benchmarking by relying on unchanged timeliness requirements, providing consistent measurement over time.
- Timeliness and punctuality indices per economy i, data category cat, and period t are constructed as:

  Timeliness:
  DT I(i,cat,t) =
  StandardReleaseLag_cat + Delay(i,cat,t)
  StandardReleaseLag_cat ·100    if 0≤|Delay(i,cat,t)|<StandardReleaseLag_cat
  0 if |Delay(i,cat,t)|≥StandardReleaseLag_cat

  Punctuality:
  DP I(i,cat,t) =
  StandardReleaseLag_cat − |Deviation(i,cat,t)|
  StandardReleaseLag_cat ·100    if 0≤|Deviation(i,cat,t)|<StandardReleaseLag_cat
  0 if |Deviation(i,cat,t)|≥StandardReleaseLag_cat

  - StandardReleaseLag_cat refers to the standard-implied release lag per SDDS/SDDS Plus data category.
  - For DTIs, Delay(i,cat,t) reflects actual delays in days; for DPIs, Deviation(i,cat,t) reflects deviations from schedule (absolute deviations capture both early and late releases).
  - The timeliness-based DTI caps reporting performance at 100 percent; early reporting is not rewarded (reflects timeliness versus accuracy trade-off).
- Interpretation:
  - No delays and schedule-adhering releases = 100 percent score.
  - Delays up to an extra standard-implied reference period produce linearly decreasing scores to 0 percent.
  - |Delay| ≥ StandardReleaseLag_cat (i.e., delays beyond two times the standard-implied release lag) are scored 0 percent, interpreted as releases past the next data release due date.
  - This approach aligns with IMF Executive Board Decision Amendments for SDDS and allows mapping administrative AOR markers (‘X’ for not disseminated and ‘O’ for missing ARC entries) to numerical scores corresponding to late dissemination past next release date.

### Aggregation to global and regional indices
- Composite indices constructed as cross-country means; aggregates also produced for four statistical domains (real, fiscal, financial, external) and five IMF regions: Sub-Saharan Africa (AFR), Asia/Pacific (APD), Europe (EUR), Middle East/Central Asia (MCD), and the Western Hemisphere (WHD).
- All subscribers and data categories contribute with equal weight; aggregates expand over time with increasing subscriptions.
- Sensitivity to changing subscriber composition is investigated in appendix E.
- As of end-2023, 78 economies have subscribed to SDDS and SDDS Plus and are considered in aggregation.

### Stylized facts: global composite DTI
- Global DTI behavior (monthly series since 2006):
  - Global DTI levels are generally high, ranging on average across subscribers between 92 and 99 percent from 2006 to 2023.
  - The index declined over time: first ten years characterized by high and relatively stable DTI levels; level dropped noticeably and persistently thereafter. Since 2017, the index is exposed to higher volatility.
  - More recently, it recovered though without fully recovering to pre-2017 levels.
  - The global DTI tends to comove with major global disruptions: decreases and higher volatility during crises (GFC end-2007—mid-2009; euro area crisis end-2010—end-2012; Arab Spring end-2010—mid-2011; first year of Covid-19 pandemic Dec 2019—end-2020; early phase of Russia’s invasion of Ukraine Feb 2022—July 2022).
  - COVID-19 induced a sharp decline followed by a V-shaped recovery; 2022 shows a distinct drop around Russia’s invasion of Ukraine.
  - Since the pandemic, increased comovement with global economic conditions is observed (two-year rolling correlations based on Hamilton-filtered series increased up to about 0.75 by mid-2023).
  - Despite crises, the global DTI did not drop below 90 percent on average across subscribers, suggesting resilience; post-crisis improvements are becoming harder to maintain.
- Trend and cyclical analysis (Hamilton-type decompositions in appendix D) shows declining trend especially after 2017 and widening, more persistent negative gaps since then.

### Data timeliness by statistical domain
- Domain-level findings:
  - Real sector data are released most timely.
  - Real, external, and financial subindices tend to track each other closely.
  - Dips in timely financial data dissemination tend to be larger during crises associated with financial distress (e.g., GFC, European sovereign debt crisis).
  - Fiscal data tends to be reported less timely; deterioration in fiscal timeliness is especially pronounced since 2017.
  - Differences among subindices have widened over time.
- Fiscal decentralization effect:
  - Economies grouped by tax revenue decentralization (IMF Fiscal Decentralization Dataset) show that decentralization is associated with longer delays in fiscal data dissemination.
  - Classification: threshold of 0.29 percent of tax revenue decentralization as share of general government separates relatively decentralized (14 economies as of 2020 or latest) from relatively centralized economies (49 economies).
  - Lower fiscal DTI values on average for relatively decentralized economies; interagency coordination needs likely contribute to longer delays.
- Fiscal transparency relevance:
  - Greater fiscal transparency has been associated in the literature with better credit ratings, fiscal discipline, and lower sovereign financing costs for EMEs and DEs and AEs (literature cited in source).

### Data timeliness by economy groupings and regions
- Advanced economies (AEs) vs Emerging Middle-Income Economies (EMEs):
  - Composite DTI divergence since 2017: persistent gap opened up with AEs showing greater resilience and recovering to pre-2017 levels after the pandemic, while EMEs experienced more lasting impacts from Covid-19 and recent geopolitical tensions.
  - Over 2023 the gap narrowed considerably.
  - Before 2017, EMEs sometimes outperformed AEs (e.g., during European sovereign debt crisis).
- SDDS vs SDDS Plus split:
  - SDDS Plus reporting performance started on a higher level in 2015; pronounced decline during 2017/18 documented for SDDS Plus as well, with subsequent settling on a lower level.
  - SDDS Plus established in 2012; first eight countries adhered in 2015; as of end-2023, 29 economies adhere to SDDS Plus of which 24 are AEs.
- Regional patterns:
  - Pre-2017 comovements across regions were high and homogeneous; regional divergence increases after 2017.
  - Pandemic impacted all regions with varying severity:
    - EUR subscribers showed the most resilient performance on average.
    - AFR, MCD, and WHD subscribers on average were more vulnerable, especially during the pandemic.
    - APD subscribers track the composite index closely.
    - Volatility highest for AFR group (fewest subscribers).
  - MCD subscribers showed the most persistent decline starting around 2015; WHD subscribers show most relative improvements since 2018.

*Source: IMF SDDS and SDDS Plus Annual Observance Reports (AORs) and author’s computations as described in the source content.*

### Appendix E investigates potential composition effects in detail. While the mean-based DTIs

### wpiea2025236-source-pdf - Appendix E investigates potential composition effects in detail. While the mean-based DTIs

### Composition and robustness
- Appendix E investigates potential composition effects in detail and finds that while the mean-based DTIs are importantly driven by weak performances at the bottom of the monthly distributions, the key stylized facts remain robust to alternative index compositions and extending the set of economies with new subscribers does not materially drive global DTI dynamics.
- Regional DTI observations can be very high, "often at or close to 100 percent" (note in text).

### Tighter fiscal stances and constrained resources
- Fiscal resources devoted to statistical services are posited as a potential explanation for DTI dynamics; timely publication of aggregate macroeconomic and financial data requires dedicated statistical services and resources.
- Budgetary envelope trends:
  - Overall own and donor-funded budgetary envelopes have shrunk with adverse implications for statistical services.
  - General government total and general public service expenditures grew during the GFC (end-2007—mid-2009) and the Covid-19 pandemic (Dec 2019—end-2020), but both shrank during the intervening years and "both shrank the most by 2016/17," suggesting constrained resources prior to the drop in the DTI in 2017/18.
- Donor funding patterns:
  - Donor funding for statistical capacity building directed toward EMEs declined substantially between 2013 and 2017, with a notable bounce-back in 2022 that coincides with a narrowing gap between AE- and EME-DTIs after 2022.
  - The donor-funding indicator covers 38 of the 78 economies in the study and predominantly covers EMEs in the sample.
  - Since 2017, EMEs’ DTIs decreased and recovery to pre-2017 levels remained incomplete.
- Evidence of underfunding and shrinking inflation-adjusted budgets for statistical offices is cited from multiple sources (Grant 2012; Hartman et al. 2014; The Globe and Mail 2014; West 2017; Groshen 2021; UKSA 2022; Campbell 2023; American Statistical Association 2024; The Economist 2025).
- Low prioritization tendencies may constrain modernization and innovation efforts needed in the age of data science and AI (American Statistical Association 2024).

### Empirical applications — overview
- The paper investigates predictive relationships between DTIs and public finance outcomes and tests early warning properties for economic downturns.
- Methodology highlights:
  - Two-variable panel VARs and Granger causality tests are used to study relationships between year-on-year changes in economy-level fiscal DTIs and changes in general public service expenditures, and separately with changes in central government debt (CGD).
  - ROC curves derived from panel logit models are used to evaluate early warning properties for business cycle recessions and systemic crises.
  - Business cycle recessions are derived from the Bry-Bochan algorithm on real GDP series from the IMF’s WEO April 2024 database for the sample 2006Q2-2023Q4.
  - Four years of lagged DTI-growth are considered (i.e., 16 quarters).

### Fiscal DTIs and public finance outcomes — key findings
- Temporal sequencing and predictive relationships:
  - Changes in general public service spending contain predictive information for fiscal DTIs; specifically, past changes in general public service spending help forecast fiscal DTIs while the reverse is not found.
  - The third lag of changes in general public service spending (three years ago) adds the most information and reflects a positive correlation with changes in fiscal data dissemination in period t, suggesting a lagged improvement (or deterioration) in fiscal data dissemination after changes in public service spending.
  - This predictive relationship is significant only with respect to fiscal data dissemination, not for other sub-DTIs or the composite DTI.
  - Changes in past fiscal DTIs improve CGD growth forecasts across all sub-DTIs and the composite DTI; coefficient signs are predominantly intuitively negative, suggesting improving data release efficiency helps forecast future decreases in CGD.
  - The Granger causality Walt test statistic rejects the null hypothesis of no Granger causality at the 1% level. VAR coefficients in both specifications (fiscal DTI-CGD and fiscal DTI-public service spending) are estimated to be significant up to the first three own lags.

### Early warning properties — quantitative results
- Baseline ROC results (composite DTI as main explanatory variable; four years of lags included):
  - Business cycle recessions (sample 2006Q2-2023Q4): Area under ROC curve = 0.7221.
  - Systemic crises (sample 2006Q2-2019Q4): Area under ROC curve = 0.6855.
  - Both ROC curves lie above the 45-degree diagonal (random classifier), indicating DTIs contain useful predictive information for economic developments.
- Sub-DTI diagnostic signals:
  - Real and external DTIs are more informative for business cycle recessions.
  - Fiscal DTIs show the strongest predictive ability for systemic crises, especially sovereign debt crises.
- Model comparisons and complementarities:
  - In business cycle specifications, including only lagged DTIs is almost as informative as including lagged real growth at many thresholds; adding both can show marginal gains for some thresholds.
  - In systemic crises specifications, private credit growth is identified as the single best early warning indicator; however, extending private credit growth models by including lags of the fiscal DTIs improves early warning properties along many thresholds, reflecting fiscal DTIs’ usefulness in indicating sovereign debt crises.
- Data and sample notes:
  - Systemic crises data comes from Laeven & Valencia (2020) amended by Nguyen et al. (2022) for 2018-19; systemic crises observations are available on an annual basis for 2018-19 and are assumed active in all quarters of a respective year in quarterly specifications.
  - For a subset of economies with only annual series, quarterly series are interpolated using cubic splines on 2005-2023 samples for turning-point detection (EGY, KGZ, LKA, MKD, MNG, MUS, NAM, SAU, SEN, SLV, SYC, URY, and WBG).

### Stylized facts and policy-relevant implications (from conclusion and empirical evidence)
- Stylized facts documented:
  - A notable shift toward longer delays in official data release since 2017, particularly pronounced among EMEs, leading to a widening AE–EME DTI gap.
  - Real sector data is consistently published more timely than other data; fiscal data dissemination faces persistent challenges.
  - More coordination-intense government decentralization has, on average, been associated with longer delays in recent years.
  - Timely dissemination of official data tends to decrease during episodes of global distress.
- Policy implications:
  - Constrained fiscal resources devoted to statistical services can help explain deteriorations in data dissemination timeliness.
  - Maintaining robust data transparency frameworks and subscriber compliance with the IMF Data Standards Initiatives remains important.
  - Resilient compliance and sustained institutional capacity and sufficient resources are required to ensure timely data release when information is most needed to resolve macroeconomic imbalances and market disequilibria.
  - The IMF Data Standards Initiatives continue to be relevant provided subscribers comply; decreasing compliance can be accompanied by challenges related to fiscal outcomes and economic downturns.

*Source: wpiea2025236-source-pdf (Appendices and sections quoted as provided in the supplied content).*

### References

### wpiea2025236-source-pdf - References

### Appendix A — Details and Examples of AOR Tables 1 and 2
- AOR Tables 1 and 2 (illustrated with the Swiss 2023 AOR) record:
  - ’check marks’ = timely/punctual data releases through the NSDP (delay of 0 days).
  - Delays recorded as negative entries measured in days.
  - ’X’s = data not disseminated through the NSDP (associated with 0 percent timely/punctual release).
  - ’O’s = missing ARC dates (associated with 0 percent timely/punctual release).
  - ’NYD’ = data not yet disseminated; ’P’ = release pending. These two are treated as if timely/punctual in index construction to avoid bias for observations toward the end of the reference period.
- Source example cited: AOR 2023 for Switzerland (IMF 2023).

### Appendix B — SDDS Subscribers and SDDS Plus Adherents (as of end-2023)
- Coverage: list of SDDS subscribers and SDDS Plus adherents reported “as of end-2023.”
- Examples (preserving original entries and codes):
  - Argentina (ARG) WHD EME 1-Nov-1999
  - Armenia (ARM) MCD EME 7-Nov-2003
  - Australia (AUS) APD AE 23-Jul-2001
  - Austria (AUT) EUR AE 10-Jan-2017 (SDDS Plus)
  - Belgium (BEL) EUR AE 26-Jan-2023 (SDDS Plus)
  - Canada (CAN) WHD AE 11-Apr-2017 (SDDS Plus)
  - China, People’s Republic (CHN) APD EME 7-Oct-2015
  - France (FRA) EUR AE 18-Feb-2015 (SDDS Plus)
  - Japan (JPN) APD AE 18-Apr-2016 (SDDS Plus)
  - United States (USA) WHD AE 18-Feb-2015 (SDDS Plus)
- Notes on coding:
  - Income level codes: AE = advanced economy; DE = developing economy; EME = emerging market economy (following IMF economy classification).
  - Region codes: AFR = Sub-Saharan Africa; APD = Asia/Pacific; EUR = Europe; MCD = Middle East/Central Asia; WHD = Western Hemisphere.

### Appendix C — Punctuality-based Global Composite DPI
- Composite DPI constructed from IMF SDDS and SDDS Plus AORs and author’s computations.
- Gray-shaded historic episodes used in charts: 
  - GFC (end-2007—mid-2009),
  - euro area crisis (end-2010—end-2012),
  - Arab Spring (end-2010—mid-2011),
  - first year of Covid-19 pandemic (Dec 2019—end-2020),
  - early phase of Russia’s invasion of Ukraine (Feb 2022—July 2022).

### Appendix D — Trends and Gaps: Econometric Lens
- Trend-cycle decomposition approach:
  - Hamilton (2018) trend derived from an autoregressive process including four most recent lags and a constant; cyclical component = unexpected deviations from a two-year-ahead forecast.
  - Quast & Wolters (2022) propose forecast-averaging (one-to-three-year-ahead forecasts) to smooth noise.
- Key numeric trend finding:
  - Both trend measures show an overall declining trend, especially after 2017, from around 96-97 percent to around 94-95 percent.
- Dynamics of gaps:
  - Gaps between index levels and the underlying trend widened in recent years and became more persistent (did not revert or close as quickly as earlier in the sample).
- Estimation details (as noted):
  - For Hamilton’s filter, forecast horizon of 24 months.
  - For modified Hamilton filter, forecast horizons between 12 and 36 months are considered.
  - Estimated coefficients informed by the whole sample period despite one-sided filters.

### Appendix E — Distributional, Robustness, and Sensitivity Perspectives
- E.1 Distributional findings (global composite DTI dispersion):
  - More than 50 percent of subscribers maintain consistently timely data releases each month.
  - The mean-based composite index is importantly driven by weaker performances; downward spikes are driven by the bottom three to eight subscribers.
  - The persistent downward shift in performance since 2017 is visible up to the 25th percentile, which corresponds to up to about 19 economies, and pulls the mean more toward the 25th percentile.
  - The 10th and 25th percentiles exhibit persistent declines since 2017; the 5th percentile shows recovery tendencies but sharp declines around periods of distress.
- E.2 Role of economy composition for aggregate indices:
  - Two alternative compositions examined:
    - Average performance among the initial set of subscribers (excluding those that subscribed after 2006).
    - Average performance among the SDDS Plus adherents as of 2023 backcasted over the whole sample.
  - Key finding: remarkable consistency in overall trends across compositions; SDDS Plus subscribers maintain higher performance, particularly for real sector data timeliness.
  - Conclusion: addition of 16 new SDDS subscribers did not materially drive global DTI dynamics; financial sector performance largely mirrors composite dynamics.

### Appendix F — Early Warning Properties: Additional Results, Alternative Specifications, and Robustness
- Purpose: assess predictive content of DTI growth for economic downturns and systemic crises using logit specifications and ROC/AUROC analysis.
- Table 2 — Baseline and alternative logit specification results (preserving coefficients and AUROC values exactly):
  - Baseline: DTI growth – 16 quarters
    - Coefficient: -0.244***
    - AUROC: 0.722[0.7010.743]
    - Systemic crises coefficient: -0.009*** AUROC: 0.686[0.6110.760]
  - DTI growth: 12 quarters
    - Coefficient: -0.082*** AUROC: 0.714[0.6940.734]
    - Systemic crises coefficient: -0.071*** AUROC: 0.678[0.6080.748]
  - DTI growth: 20 quarters
    - Coefficient: -0.016*** AUROC: 0.728[0.7070.749]
    - Systemic crises coefficient: -0.543*** AUROC: 0.725[0.6530.796]
  - DTI and real GDP growth (16 quarters)
    - Coefficient: -0.207*** AUROC: 0.777[0.7580.796]
    - Systemic crises coefficient: -0.142*** AUROC: 0.750[0.6810.818]
  - DTI and inflation (16 quarters)
    - Coefficient: -0.263*** AUROC: 0.735[0.7140.756]
    - Systemic crises coefficient: -0.434*** AUROC: 0.784[0.7240.844]
  - DTI and private credit growth (4 years) (annual)
    - Coefficient: -0.027*** AUROC: 0.735[0.6980.772]
    - Systemic crises coefficient: -0.131*** AUROC: 0.864[0.7400.988]
  - DTI and public debt growth (4 years) (annual)
    - Coefficient: -0.023*** AUROC: 0.737[0.7020.772]
    - Systemic crises coefficient: -0.040*** AUROC: 0.791[0.6500.932]
  - Real subindex (16 quarters)
    - Coefficient: -0.257*** AUROC: 0.721[0.6700.742]
    - Systemic crises coefficient: -0.198*** AUROC: 0.680[0.6020.757]
  - Fiscal subindex (16 quarters)
    - Coefficient: -0.033*** AUROC: 0.717[0.6960.739]
    - Systemic crises coefficient: -0.243*** AUROC: 0.711[0.6340.787]
  - Financial subindex (16 quarters)
    - Coefficient: -0.090*** AUROC: 0.719[0.6980.740]
    - Systemic crises coefficient: -0.040*** AUROC: 0.687[0.6110.763]
  - External subindex (16 quarters)
    - Coefficient: -0.102*** AUROC: 0.720[0.6990.741]
    - Systemic crises coefficient: -0.774*** AUROC: 0.693[0.6170.769]
- Statistical notes and interpretation:
  - *, **, and *** indicate significance at the 10, 5, and 1% levels.
  - Baseline DTI-growth coefficient of -0.24 (business-cycle-recession exercise) is statistically significant at the 99% confidence level; increasingly timelier data releases associated with decreasing odds of a business cycle recession.
  - Negative coefficient estimates across different DTI-growth lag structures. Including up to five years of lags adds limited precision; four lags (16 quarters) is reasonable.
  - Adding macroeconomic variables generally improves predictive power (increases AUROC). For business cycle recessions, lagged real GDP growth increases AUROC most; for systemic crises, adding private credit growth improves predictive ability the most.
  - Caveats:
    - Fewer systemic crisis observations (often annual) reduce precision in systemic-crisis exercise.
    - Interpolating annually available series would introduce hindsight; annual specifications used for private credit and public debt (data available in IMF GFS and global debt databases).
    - Coefficients for private credit and public debt specifications may be hard to distinguish from zero for systemic crises; Hartwig et al. (2021) suggest private credit growth measures can have mixed signaling properties across economies.
- ROC/AUROC highlights from figures and exercises:
  - DTI and Real GDP Growth (2006Q2-23Q4) — Area under ROC curve = 0.7770 (business cycle recessions).
  - DTI and Inflation (2006Q2-23Q4) — Area under ROC curve = 0.7354 (business cycle recessions).
  - DTI and Private Credit Growth (annual: 2007-22) — Area under ROC curve = 0.7348 (business cycle recessions).
  - DTI and Public Debt Growth (annual: 2007-22) — Area under ROC curve = 0.7368 (business cycle recessions).
  - DTI and Real GDP Growth (2006Q2-19Q4) — Area under ROC curve = 0.7495 (systemic crises).
  - DTI and Inflation (2006Q2-19Q4) — Area under ROC curve = 0.7842 (systemic crises).
  - DTI and Private Credit Growth (annual: 2007-19) — Area under ROC curve = 0.8642 (systemic crises).
  - DTI and Public Debt Growth (annual: 2007-19) — Area under ROC curve = 0.7914 (systemic crises).
- Subindex-specific signaling insights:
  - Business cycle recessions: changes in real and external DTIs associated with the highest AUROC statistics; baseline subindex AUROCs around 0.7208 (Real), 0.7174 (Fiscal), 0.7187 (Financial), 0.7198 (External).
  - Systemic crises: decreasing fiscal DTIs associated with the highest predictive ability; baseline subindex AUROCs around 0.6797 (Real), 0.7106 (Fiscal), 0.6867 (Financial), 0.6928 (External).
  - Sovereign debt crises (subtype analysis):
    - AUROC by subindex: Real = 0.6017; Fiscal = 0.8593; Financial = 0.8794; External = 0.8390.
    - Interpretation: financial, fiscal, and external subindices are strongly indicative for sovereign debt crises (AUROC well above 0.8); real DTIs less indicative (AUROC = 0.6017).
  - Currency crises and banking crises: mixed evidence; fiscal DTIs convey most indicative information for currency crises (AUROC range 0.61 to 0.67 across subindices); for banking crises, real and especially external DTIs convey signaling information.
- Overall conclusion from Appendix F:
  - DTIs contain useful predictive content for business cycle recessions and systemic crises across specifications and subindices; adding macroeconomic controls generally improves predictive performance but the main result on DTIs’ predictive usefulness remains robust.

*Source: Behind Schedule? Assessing Global Developments in the Provision of Economic Statistics — Working Paper No. WP/2025/236*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025236-source-pdf.pdf_