Managing Government Cash

2. Cash flow forecasting

Abstract

Cash flow forecasting is crucial for governments to effectively manage liquidity and make proactive decisions to maintain target cash and liquidity balances by anticipating future funding needs. Robust forecasts of cash flows help anticipate potential cash shortages or excesses, reduce the need for large cash buffers, adapt financing operations in advance, and support predictability.

This chapter provides key features of common cash forecasting practices, which combine short-, medium- and long-term forecasts, frequently updated to reflect real-time conditions. It also discusses the challenges including dependence on third-party information, recruitment and retention of qualified staff and the inherent uncertainty in fiscal projections, as well as techniques that can help address these challenges, with country examples.

2.1. Introduction

Cash flow forecasting is crucial to any government's cash management strategy. This process, which encompasses the estimation of inflows and outflows from fiscal and financial flows, helps ensure there is sufficient cash to meet government obligations as they fall due, the core objective of cash management.

Cash flow forecasting is especially useful for governments in predicting potential cash shortages. Forecasting the future cash position can help to anticipate possible negative cash balances due to mismatches between the timing of revenue and expenditure flows owing, for example, to issues of seasonality and tax revenue collection dates.

Predicting cash balances allows the government to take timely action to ensure it can meet its payment obligations. Cash flow forecasting is often driven by forecasts concerning the government’s payment and revenue collection schedule, as well as borrowing and investment plans. Regarding borrowing plans, cash forecasts are also an essential input in the timing and scale of debt issuance and assist in the active management or investment of surplus cash.

This chapter examines the diverse practices of the Ad Hoc Group on Cash Management member countries concerning cash flow forecasting. It is structured as follows: the first section presents an analysis of cash flow drivers and information sources. The next section focuses on interactions with other bodies for information gathering. An overview of cash flow planning is presented in the next section. Modelling, uncertainty and forecasting performance are then explored, followed by an analysis of governance structures for decision-making. The chapter concludes highlighting the importance of cash flow forecasting as an input for decision-making.

The below schematic (Figure 2.1.) summarises the most common practices across countries covered throughout this chapter.

Figure 2.1. Streamlined summary of the most common cash forecasting practices

Note: This figure represents the most common practices in a stylised manner. Practices can significantly differ from these in some countries.

Source: Based on the OECD Ad Hoc Group on Cash Management Survey (2024).

2.2. Cash flow drivers and information sources

2.2.1. Drivers of cash flows

Cash flows are driven by fiscal and financial revenues and expenditures. On the fiscal side, this includes traditional budget expenditures such as salaries, pensions and social protection payments, and on the revenue side, tax receipts. Financial flows encompass mostly debt-related activities (e.g. debt issuances and redemptions) and short-term liquidity management, such as investing cash balances or using funding instruments like T-bills, repo operations, commercial papers and loans.

Additional cash flows can arise from non-debt financial activities such as derivatives that hedge financial operations (e.g. FX swaps, interest rate swaps) and privatisations. In that context, Credit Support Annex (CSAs), which require the posting or receiving of collateral, are of particular importance. These flows are uncertain as they are market-driven and, depending on the scale of the swap or hedge book, where the collateral posted is cash against domestic bonds, the swings in the cash inflows/outflows can be substantial.

2.2.2. Fiscal cash flows

Fiscal flow forecasting is generally the central challenge in cash forecasting, as debt flows, the main component of financial flows, are typically known in advance with a reasonable level of certainty. Fiscal cash flow projections are usually prepared by the fiscal authority or line ministries,1 with limited involvement from the cash management team, which primarily aggregates the data and may make small adjustments. These forecasts are naturally uncertain due to the variability in economic activity, affecting tax revenues and certain government expenditures, such as social protection benefits. In contrast, debt flows are managed by the DMO and follow the issuance calendar and redemption schedule. The remaining financial flows relate to short-term funding plans overseen by the cash management team.

Countries rely on historical cash flow data, information obtained from budget cycles and other data sources provided by central government bodies to forecast fiscal cash flows. These bodies include the fiscal authority, line ministries and government agencies. In some cases, information is gathered directly through centralised systems, while in others, cash forecasting depends on multiple sources, including manually provided data, budget projections (and revisions), historical trends and information systems. Although fiscal flow forecasts from the fiscal authority and line ministries can be smoothly integrated into cash management operations, cash management teams often use historical trends to break down these monthly or yearly forecasts into weekly or daily flows for short-term forecasts.

The TSA may cover only the central government or extend to other public entities, requiring data collection from these entities for cash flow forecasts. These could include social security agencies, pension funds, special-purpose funds and other state-relevant entities generating significant flows. This can also happen when these entities are not subject to centralisation but have their deposits held with the treasury.

There can be differences in the process and use of information from central government sources versus other public entities, making the latter more complex to incorporate into cash flow forecasts. The expected flows from these other public entities may be reconciled and incorporated into the overall cash flow forecast, considered through a different mechanism than those of the central government, or not included at all, as they may not be legally required to coordinate with the central government. The frequency of interaction with these public entities varies, sometimes weekly or monthly. In other cases, when they are less regular, they are often linked to specific financing operations.

2.2.3. Financial cash flows

In most cases, daily financial cash flow projections are readily available and updated within the debt management systems operated by the DMO. These systems can be integrated with cash management systems, allowing automatic data exchange for cash planning. Alternatively, if the systems are not integrated, debt-related information can be exported (e.g. in CSV or Excel files), processed, and fed into cash planning manually.

Due to the ease of extraction, financial cash flows are typically updated daily, or even intraday, to incorporate new debt and investment information into the cash forecast. Any updates to planned debt issuance are also promptly reflected in cash planning. While fiscal cash flows are often provided on a monthly or yearly basis and need to be disaggregated for short-term forecasts, daily financial flow data can be aggregated into monthly or yearly figures for long-term forecasting.

Unlike fiscal cash flow information, financial cash flows are highly accurate in daily terms across any time horizon. This accuracy stems from the predominance of fixed-rate debt in OECD countries' debt portfolios, typically comprising over 80% of the outstanding debt stock. However, in some countries, accuracy may be affected by a higher proportion of debt issued at variable rates (e.g. inflation-linked or floating-rate debt) and by retail debt products, where issuance amounts are more difficult to predict due to investor discretion (e.g. household savings-linked products or the ability of investors to purchase and redeem on demand (Foxall and Policino, 2025[1]) Where the DMO uses swaps and derivatives to manage risk on a large scale, two-way CSAs requiring the posting or receiving of collateral can also lead to unpredictable and large market-driven cash flows.

Changes in market conditions can affect financial flows. For instance, the monetary tightening cycle between late 2021 and 2023, which also included quantitative tightening (QT) programmes that are ongoing in many OECD countries,2 has impacted financial flows, making them less predictable in some countries and affecting cash management operations (Box 2.1). This includes changes to issuance plans, taxpayers' and subnational government's behaviour, and market conditions for short-term funding and investing.

Box 2.1. Impact of the recent tightening cycle and quantitative tightening on cash flows

Financial flows are generally more predictable than fiscal flows due to their alignment with scheduled debt issuances and redemptions. However, shifts in financial conditions can introduce unexpected impacts, affecting cash management operations. Recent monetary tightening (post 2022), including QT, has exemplified these challenges in a few OECD countries.

• In New Zealand, the central bank sells government bond securities directly to the treasury as part of its QT programme. This prompted the treasury to revise its issuance plans soon after the central bank’s announcement. This revision has been done in advance, as the central bank has published its complete bond sales plan, minimising undue surprises.

• In Sweden and Switzerland, tighter monetary conditions have led to structural changes in taxpayer behaviour. Under the negative interest rates prevalent over the previous decade, taxpayers delayed refund requests and paid taxes early to maximise the time their funds were held in the TSA at zero interest. In Switzerland, when interest rates turned positive, taxpayers began delaying tax payments until the last possible moment and requesting refunds earlier to invest these funds at positive market rates. In Sweden, taxpayers continue to leave overpaid taxes with the government as they can earn market rates (with a lag), giving them a positive interest rate.

• In addition, cantons (regions) in Switzerland used the TSA as a zero-rate deposit mechanism to park their cash balances when rates were negative. The cantons collect taxes from natural persons and also reimburse the withholding tax to them. In the negative interest rate environment, the cantons have refunded the withholding tax, but at the same time they have not reclaimed it from the federal government, thus accumulating credit balances with the federal government. However, they rapidly withdrew funds from there to benefit from positive interest rates in the market. This led to large and unexpected outflows and forced the cash management team to adjust their cash planning.

• In the United States, the QT programme of the Federal Reserve is considered in the cash forecast as it increases the marketable borrowing need from private market participants, potentially affecting bond yields and market conditions. Additionally, the current monetary tightening cycle has negatively impacted the Federal Reserve’s earnings, which are remitted to the Treasury and have historically amounted to around 0.5% of GDP annually over the past decade, directly impacting cash inflows.

2.3. Interaction with other bodies for information gathering

2.3.1. Bodies that provide inputs to the cash management team

The bodies that provide most of the inputs for cash flow forecasting typically include the DMO, the tax authority, the budget office and line ministries. In some cases, the budget office centralises the consolidation of information (Figure 2.2., Panel A), while in others, the cash management team receives inputs from all these bodies (Figure 2.2., Panel B).

The tax authority often provides formal revenue forecasts, but there are exceptions, requiring the cash management team to produce their own estimates. This is the case in Iceland and Switzerland. In the latter, the cash management team maintains regular contact with the tax authority via phone or email to help accurately plan for major inflows and outflows.

In some countries, expenditure forecasts are compiled centrally by the budget office and then passed to the cash management team (e.g. Austria, Hungary, Italy and the United States). In contrast, in other countries, data collection depends more on direct communication between the cash management team and line ministries (e.g. Belgium, Brazil, Japan and Türkiye). Additionally, in some instances, the cash management team is responsible for estimating certain expenses, such as discretionary expenditures (e.g. Brazil).

In some cases, a broader range of entities outside the central government contribute to cash flow forecasts. For instance, state-owned legal entities provide inputs to the cash management team (e.g. on dividends, financing, and investment activities) in Austria, Belgium, Brazil and Türkiye. Subnational governments supply inputs in Belgium, Japan and Türkiye. In Belgium and the United States, economic study departments (e.g. macro-fiscal units) also assist cash flow forecasting by providing economic forecasts, such as GDP growth, inflation indices, and interest rates. Lastly, independent bodies, including social security agencies, development finance institutions, regulatory and supervisory agencies and central banks, are common contributors to cash flow forecasts.

Figure 2.2. Data gathering process

Source: OECD Ad Hoc Group on Cash Management

2.3.2. Models of data collection and exchange

The granularity of information shared can be categorised as either 1) detailed cash flows with specific dates, values and purposes and 2) aggregated figures like total revenue and expenses. Fiscal data tends to be shared less frequently and in a more aggregated form, unlike financial forecasts, which are typically more precise and regularly updated (refer to sub-section 2.2). When cash management teams only receive aggregate estimates, they must disaggregate these using other estimation methods.

Typically, the budget and its revisions serve as the foundation for cash forecasts. This is because the budget provides a binding plan for government spending and revenue collection, centralising expenditure plans across all ministries and agencies. As a result, cash flows and budget flows are well-aligned in most countries. In some cases, budget preparation and revision, along with medium-term fiscal frameworks (e.g. medium-term budget plans or Stability Programmes for EU members), are the only occasions when the fiscal authority formally shares information on forecasted revenues and expenditures.

However, budgets are not revised as frequently as cash forecasts, making regular communication essential for accurately forecasting cash flows and coordinating cash management and funding needs. In the absence of updated official information, the cash management team must often rely on their own estimations, drawing on historical data and growth assumptions, or gathering information directly from other bodies, such as the fiscal authority and line ministries. When information is centralised within the fiscal authority, budget office, or an Integrated Financial Management Information System (IFMIS), data collection for the cash management team becomes easier. Regular communication is, therefore, critical to ensuring cash flow forecasts are aligned with projections from the fiscal authority and line ministries.

Data collection involves both automated extraction from databases and manual inputs. Some countries extract data, typically financial flows such as debt maturities, from treasury management or accounting systems, supplemented with fiscal or budget data. However, only a few countries have fully automated extraction, and most countries still rely on Excel files, emails, meetings and calls to gather data. When cash management includes institutions beyond the central government, gathering data from sources outside these systems may be necessary, further complicating the process. Box 2.2 explores the means of data gathering in the United States while Annex 2.A compares the country’s cash management practices with Iceland.

Box 2.2. The United States’ data sharing practices in cash management

In the United States, the fundamental data for cash flow forecasting is gathered during the quarterly preparation of budget-related A-11 reports, with updates made when new information becomes available. The A-11 report provides the foundation for cash management forecasts by detailing government outlays. Agencies submit this report quarterly through the Office of Management and Budget’s Monthly Outlays Plan application. Agencies can also attach supplemental data, such as Excel files, to offer more detailed information. These attachments enable the cash management team to incorporate critical information not captured in the standard A-11 format. Agencies may also provide updates via email on an ad hoc basis.

In addition to the quarterly A-11 reports, agencies also report daily cash outlays and debt forecasts. Daily cash outlays are submitted via the Cash Position System, reviewed by the Cash Reporting Branch, and forwarded to the Office of Fiscal Projections. Debt-related data, such as pro-forma financing updates, is shared daily by the DMO. The treasury and the Federal Reserve also monitor intra-day cash flows against projections to track significant variances from forecasts. As a result of this daily exchange of information, the cash management team can fine-tune the cash position forecast in a timely manner.

Some aspects of data integration are automated. For instance, the DMO receives daily automated feeds of cash position forecasts, which it uses to update pro-forma financing estimates. These estimates are returned to the Office of Fiscal Projections to refine the forecast overall cash position.

Although the cash management team relies on two primary institutional mechanisms to ensure data reporting compliance, enforcement is largely reputational in practice. First, Office of Management and Budget Circular A-11 requires agencies to submit outlay reports to improve the treasury’s forecasting of daily cash balances and borrowing requirements. Second, the treasury financial manual mandates reporting large transactions over USD 50 million. While these guidelines are binding, agencies are encouraged to meet reporting deadlines through a scorecard prepared by the Cash Reporting Branch, highlighting any non-compliance.

The frequency of information sharing varies across bodies and countries and is typically supplemented with ad hoc interactions for significant revisions. Monthly or bi-monthly data exchange with the fiscal authority or line ministries is the most typical arrangement (e.g. Austria, Belgium, Brazil, Japan and Türkiye). In some cases, certain entities, such as the fiscal authority or the budget office, the central bank or the DMO, share information daily (e.g. Belgium, Italy, Japan and the United States). In other countries, data are primarily shared during key periods of the budget cycle and in other ad hoc circumstances, such as in Ireland, Luxembourg and Portugal.

2.3.3. Institutional framework for data exchange

Cash forecasting requires a timely flow of information and an effective coordination mechanism among relevant parties, including line ministries, the central bank, budget offices and the fiscal authority. The information supplied and its frequency can be defined formally or informally, depending on whether it is based on a binding legal or administrative mechanism, or an alternative non-binding arrangement, such as a memorandum of understanding (MOU) or service-level agreement (SLA).

Formal obligations are rare, and in such cases, informal mechanisms become essential. Building strong relationships between organisations and counterparties is also crucial in these instances, regardless of the formality of the mechanism.

Approximately half of countries do not rely on legal or quasi-legal provisions for data sharing. As a result, cash management teams are, in many countries, responsible for gathering information without any legal requirement or incentives to ensure timely, accurate and reliable reporting. In some cases, the absence of legal enforcement is due to the centralisation of information by the MoF (e.g. Austria). In other cases, without formal obligations, the frequency and quality of data can vary, largely depending on the cash management team’s initiative and the willingness of other entities to co-operate. This challenge is particularly pronounced for entities outside the central government (e.g. state-owned enterprises (SOEs) and subnational governments).

Some countries with legal or quasi-legal frameworks stipulating the frequency and type of data to be shared are Brazil, Hungary, New Zealand, and the United States. For example, Brazil’s "responsibility matrix", established by the Budget Execution Board and created by a presidential decree, determines which government bodies are responsible for providing forecasts of revenues and expenditures. These updates are submitted bi-monthly, in line with the Fiscal Responsibility Law. Similarly, in Hungary, budgetary entities face penalties for failing to provide timely or accurate forecasts to the State Treasury. In New Zealand, the treasury is empowered under the Public Finance Act to require information, including cash flow forecasts, from government entities. In the United States, the Office of Management and Budget Circular No. A-11 and the treasury financial manual require agencies to submit reports on federal outlays, with the Cash Reporting Branch using a scorecard system to incentivise compliance through reputational risks (see Box 2.2 for details).

Belgium, Ireland and Switzerland are examples of countries without legal provisions; however, they have formal agreements, administrative guidelines, or MOUs. Belgium, for example applies a similar approach, where public entities benefiting from a credit line with the Belgian Debt Agency must report expected cash flows under their credit agreements. In general, public entities do not have a legal obligation to report expected cashflows to the treasury department, but are advised to do so through informal means of communication. Ireland uses an MOU between the treasury and the MoF’s Budget Office to ensure the timely exchange of fiscal estimates. Similarly, in Switzerland line ministries are not legally obligated to report forecasts, but the Swiss Financial Budget Act and Financial Budget Ordinance require administrative units to collaborate on recurring budget estimates, providing a framework for data sharing. These units also collaborate with the cash management team by routing all payments through a centralised system, allowing real-time updates in cash planning tools.

However, in other countries without legal obligations, such as Iceland, Luxembourg and Portugal, there are no formal obligations, instead data sharing relies on informal channels and well-established relationships between the cash management team and other government entities. In Iceland, cash managers have access to historical data, but future inflows and outflows are often unreported by line ministries, with no legal consequences for non-compliance. In Luxembourg and Portugal, the budget cycle is key in providing information for cash planning, supported by line ministries on an ad hoc basis in case of significant expected divergences from the budget.

It is worth noting that data collection from the DMO by the cash management team is generally seamless, as both often operate within the same institution. The need for formal institutional frameworks primarily arises when coordinating with other bodies, such as the budget office and line ministries.

2.3.4. Interaction with entities outside the central government

Although the TSA typically covers only central government entities, it may also extend to other public bodies. When the TSA has broader coverage, it often includes SOEs, subnational governments and social security funds. In such cases, the TSA typically remunerates these deposits at or near market rates, and these entities often can voluntarily place their funds in the TSA. Countries with variations of this practice include Germany, Hungary, Italy, Ireland, Portugal, Switzerland, Türkiye and the United Kingdom.

A TSA with extended coverage to other public entities can impact data collection activities for cash flow forecasting. In these countries, government entities outside the TSA may provide forecasts to the cash management team, though the extent and integration of this data can vary. Examples of countries where this information is integrated into the TSA's cash planning include:

• New Zealand, where several Crown Entities and SOEs outside the TSA provide cash flow forecasts, which are monitored by the treasury and integrated into the central cash management process as they might require central funding. These entities manage their own cash independently, but their cash flows are monitored since they can request funding from the central government.

• Switzerland, where decentralised federal units can invest cash surpluses with the central treasury or obtain loans from the treasury. As some of these units have high financing volumes, they provide quarterly information on the expected development of funds, based on an agreement, which is an important input for the TSA’s cash flow planning.

• In Türkiye, the Unemployment Insurance Fund and Saving Deposit Insurance Fund are outside the scope of the TSA. These entities share their cash flow information with the treasury on a daily basis, and these estimates are integrated into the treasury’s cash projection.

In other countries, such as Austria and Belgium, while some entities outside the TSA share financial information, this data are not fully integrated into the cash forecasting process. Austria receives financing needs from subnational governments, which are considered separately in the funding plan.3 Belgium relies on long-standing relationships with entities like social security funds and subnational governments to gather information useful for estimating cash flows, such as payroll withholding tax or social security contributions.

2.4. Cash flow planning overview

2.4.1. Short-term, medium-term and long-term forecasting

Typically, the cash management team produces short-term, medium-term and long-term forecasts with varying levels of detail to achieve different objectives (Table 2.1). Short-term forecasts, which can range from one week to six months (and in some cases up to one year), are more granular, with daily or weekly breakdowns. These are primarily used for operational cash management and liquidity planning. Medium-term forecasts, which can span from 3 to 12 months, are often aggregated monthly. They offer greater cash planning visibility in the medium term, especially for countries that typically project daily flows over shorter periods. Long-term forecasts can cover periods from 1 to 10 years and are usually aggregated annually or biannually. These forecasts provide a strategic overview of financing requirements and support the development of annual issuance plans.

Most countries update cash flow forecasts daily or weekly, but some do so only monthly or quarterly, with the frequency of updates depending on the level of granularity. Short-term forecasts are updated more frequently than medium-term or long-term forecasts due to their role in managing immediate cash flows. More specifically, short-term forecasts are typically revised intraday, daily or weekly, to capture day-to-day changes in cash flows. In contrast, medium-term forecasts are generally updated on a monthly or quarterly basis, while long-term forecasts are revised less frequently (biannually, annually or ad hoc) depending on relevant events, such as budget revisions or significant changes in economic conditions.

Table 2.1. Characteristics of short-, medium- and long-term cash flow forecasting

Type	Period	Timeframe aggregation	Objective	Frequency of update	Aggregation method
Short-term forecast	1 week to 6 months (commonly), but can reach up to one year	Daily or weekly	Supports operational cash management and liquidity planning	Intraday, daily, weekly or monthly	Daily flows for fiscal variables are estimated using (adjusted) historical values provided annually or monthly (top-down approach), while financial variables are collected daily
Medium-term forecast	3 to 12 months	Monthly	Provides greater cash planning visibility for the medium term, especially for countries that typically project daily flows over shorter periods	In between long-term and short-term forecasts	Combines top-down approach for projecting daily flows with bottom-up aggregation for annual flows
Long-term forecast	1 to 10 years	Annually or biannually	Provides an overview of financing requirements and supports the development of annual issuance plans as well as more strategic multi-annual plans	Quarterly, biannually, annually or ad hoc, depending on relevant events (e.g. budget revision)	Fiscal variables are typically provided annually (e.g. in the budget), while financial variables are collected daily and aggregated into annual terms (bottom-up approach)

Source: Based on the OECD Ad Hoc Group on Cash Management Survey (2024).

Short-term forecasts require more analytical work from the cash management team, while long-term forecasts are more straightforward. Short-term forecasts involve detailed analysis and are typically refined as key dates approach, relying on historical patterns and due date calendars. Long-term forecasts rely more directly on inputs from the budget and financial data, which are generally aggregated without the need for significant additional analysis. Medium-term forecasts combine elements of both short- and long-term forecasts, providing more granular information than long-term forecasts but without the daily adjustments of short-term forecasts.

Countries differ in their use of bottom-up or top-down approaches to cash flow forecasting, depending on the level of detail in their initial data or the volatility of that data, in particular with regard to tax receipts. In the bottom-up approach, detailed daily or weekly data are collected and then aggregated into monthly or annual figures. This method is more often used for medium- and long-term forecasts, where less frequent updates are required. Conversely, the top-down approach starts with aggregate-level data, such as annual budget figures, which are then broken down into more detailed forecasts using historical trends, key cash flow dates and statistical models. This method is typically used for short-term forecasts, where frequent updates are needed.

Cash flow planning is inherently dynamic, with forecasts continuously updated under a higher level of resolution as new information becomes available. As the forecast period approaches, the level of detail increases; for example, monthly or weekly forecasts are progressively refined into daily figures. This rolling method also allows cash managers to adjust projections according to the most recent information. Automated forecasting systems enhance the effectiveness of this approach by facilitating frequent updates over various timeframes, ranging from intraday to monthly or quarterly intervals.

2.4.2. Budget and banking reconciliation

The projected daily cash position is often reconciled with actual transactions recorded in intraday bank statements from the central bank or commercial banks. Cash flow forecasts are created by aggregating expected fiscal and financial cash flows with the most up-to-date cash balance, usually from the previous day (T-1). Any discrepancies between these forecasted flows and actual flows are adjusted in the final cash position for the day (T), which then becomes the starting point for future cash forecasts. This process ensures continuous accuracy in cash flow management.

Another critical reconciliation involves aligning budget estimates with cash flows, which can be challenging. The reconciliation process with the central budget is typically aligned with official budgetary revisions, which can occur from one to four times a year, depending on fiscal volatility. Cash management teams often incorporate additional information, such as past revenue and expenditure trends, macroeconomic factors, and real cash flow data from previous cycles, into their forecasts to help address these inconsistencies. These trends may be used as benchmarks, growth rates, or other quotients to refine the forecasts (e.g. the ratio between total cash flow and non-budget cash flow to account for discrepancies that arise from non-budget transactions).

Discrepancies between budget estimates and cash flows may vary depending on the spending unit or type of expenditure. The cash management team takes note of these variations to refine future cash flow forecasting and reconciliation activities. In general, cash flow forecasting figures will be consistent with annual budget projections when they reflect the same update/version.

2.5. Modelling, uncertainty and performance

2.5.1. Modelling

In about half of the countries, the cash management team does not use any statistical models, while in the other half, models are used to forecast future cash flows and to disaggregate fiscal flows provided to a longer time frame (e.g. monthly, annually) into a shorter time frame (e.g. daily, weekly). When the cash management team does not use models, some forecasts are based on assumptions for growth rates, while others are provided by external bodies (e.g. line ministries, fiscal authority and tax authority). When models are used, they are mostly internally developed (e.g. in Excel, Stata, or R) and employ regression and time series methods based on past data, considering the seasonality of flows (see Box 2.3 on Brazil’s case).

Box 2.3. Brazil’s cash flow forecasting framework

The budgetary forecast is the primary source of information for Brazil’s TSA cash flow model, a newly developed model aimed at providing a comprehensive view of the cash for the whole single account. This forecast is re-evaluated every two months using inputs from various government bodies. Seasonality across months is addressed through these projections. To estimate daily payments, Brazil uses a regression model that calculates historical daily flow ratios (daily flows over total monthly flows) against calendar factors such as business day number, weekday, and other time-based variables. This allows the incorporation of seasonality within each month. The projected daily ratio is multiplied by the monthly budgetary forecast and adjusted based on historical forecasting errors. Non-budgetary cash flows are estimated as a percentage of budgetary flows, also based on historical data.

In addition to the TSA cash flow model explained above, the DMO uses its own model, which has been well developed over the years, to estimate the cash flow of the resources under its management (or debt liquidity reserve) to ensure debt payments. This DMO model is detailed below.

Debt cash flow projections incorporate both domestic and external debt, as well as payments of guarantees:

• Domestic Debt: The Integrated Public Debt System records the dates and values of each bond issued in the domestic market. The system calculates the projected amounts to be paid based on a macroeconomic scenario and previously selected issuance strategies, allowing daily control over debt-related expenses.

• External Debt: This system also tracks the dates and values of bonds issued in external markets and the expected payments for external credit operations. Since these payments involve advanced exchange contracts, only instalments without contracted exchange allocations are considered, allowing precise daily cash flow control.

• Guarantee Payments: This relates to credit operations contracted by subnational entities with a federal guarantee. Payments from entities with a history of defaults are incorporated into the cash flow, with well-defined dates and totals for daily control.

Primary expenses, including personnel and pension payments, are projected separately based on the updated budget allocation for the year, deducting amounts already paid. The remaining value is distributed evenly over the period under consideration.

On the revenue side:

• Revenues from bond issuances are built daily, and the issuance forecast is incorporated into the cash flow for precise daily control.

• Non-issuance revenues are reported by units responsible for its collection and incorporated according to their schedule. These revenues are evenly spread across the relevant days if reported monthly or annually.

In conclusion, Brazil’s approach illustrates common challenges in revenue forecasting due to the lack of detailed data on the timing of revenue inflows into government accounts. Currently, revenue is estimated as a singular block, but efforts are underway to develop a more disaggregated dataset. On the other hand, reliable data are available for significant expenditure categories, such as pensions, personnel and transfers to sub-federal units, which are scheduled and modelled separately. Other expenses are modelled independently due to their variability.

Adjustments to both past data before predicting future data, and to projected data based on fiscal knowledge, are crucial. In many circumstances, past data is weighted and adjusted for any atypical deviations (e.g. one-off events or trends that have ended). Projected data can also be adjusted to account for new measures or trends.

The cash management team's expertise is key to interpreting and applying statistical outputs when other bodies do not provide projected data with a sufficient level of detail. Cash flows have distinct features, seasonality patterns, and adjustment needs, such as tax due dates and filing deadlines, end-of-year bonuses on withheld taxes, and payment dates for public sector employees and social security programmes.

Tax revenues generally are more difficult to forecast precisely, especially if a country has a dependency on a particular tax source. Tax authorities often provide monthly tax estimates, which must be broken down into daily figures using historical patterns and due date calendars. This process is complicated because, for example, taxpayers do not always pay on specific days but in advance of certain deadlines, leading to uncertainty in the timing of cash inflows. In contrast, financial revenues from debt issuances are more evenly spread throughout the month and predictable as they follow a pre-determined schedule (e.g. issuance calendars).

Expenditures, albeit often more numerous and varied, tend to be easier to forecast accurately. Major expenditure categories, such as staff costs, pensions and transfers to subnational governments, are often modelled separately as they are more predictable and follow regular patterns. Certain countries, however, group all expenditure items under a single liquidity position due to the complexity and diversity of the spending categories. One of the main challenges concerning expenditure forecasting pertains to off-budget cash flows, which can be substantial and not easily anticipated by the TSA or other information-gathering mechanisms.

Forecasting models are typically developed and operated in Excel, although some countries use in-house systems or statistical software (e.g. R). The forecast generally starts with the closing cash balances from the previous day and includes inputs for expected inflows (e.g. tax revenues, actual and planned borrowing) and outflows (e.g. government spending, debt service costs). The net result of these flows provides a projected end-of-day balance, which countries usually focus on for the next 30, 60, or 90 days. When countries manage multiple currencies, these flows are disaggregated by currency.

Automation levels in cash flow forecasting differ across countries: two countries reported fully automated processes, six have partially automated processes and four rely entirely on manual processes. Fully automated systems use in-house models integrated with databases. Partially automated processes combine in-house models and databases (e.g. SAP) but may include manual inputs, such as data and analysis from Excel Workbooks. Manual processes depend heavily on Excel. Thus, cash flow forecasting often involves a combination of automated calculations and manual data inputs.

2.5.2. Forecast accuracy and uncertainty

In terms of the accuracy of forecasts, daily cash flows fluctuate widely due to the timing of revenue collections and expenditure payments, while monthly flows are more stable, ranging from 3% to 6% of GDP in many countries. Tax revenues, especially those tied to specific deadlines, tend to cluster at the end of the month, resulting in large variances in daily and weekly flows. Expenditures like payroll and social security payments are generally more predictable, though they can spike at specific points in the month, such as the beginning or end. To reduce volatility in cash balances, some countries use flexibility in scheduling large payments, such as bond redemptions or transfers to subnational governments, to align with significant inflows like bond issuances or tax receipts.

Short-term forecasts are reasonably accurate, with errors typically around 5% under normal conditions, mostly driven by revenue projections. Since some cash management teams use conservative assumptions, the error is often positive (i.e. realised net cash flows tend to be higher than projected). Revenue estimates, whether prepared by the cash management team or the tax authority, are generally more challenging to forecast accurately. In contrast, expenditure flows, though more numerous due to the higher number of line items, are generally more regular and predictable, leading to fewer inaccuracies. However, this can occasionally be offset by unpredictable off-budget cash flows.

The further into the future the forecast extends, the more likely it is to be inaccurate, as more assumptions and approximations must be made. However, given the “rolling” nature of cash flow forecasting in many countries, these forecasts are updated over time with more accurate assumptions and information, increasing their reliability. Ultimately, accuracy is more crucial in the short term, where it is used for liquidity management, than in the long term, where it guides borrowing needs and annual issuance plans, which can be adjusted more easily and gradually.

Uncertainty in cash flows is most commonly managed through liquidity policies (e.g. cash buffers), but some countries use more advanced methods to account for forecasting uncertainty, including conservative forecasting assumptions, stress tests, scenario analysis, and confidence intervals. Conservative assumptions about revenue collection, economic activity, and deep market access can also serve as buffers. Stress testing, often biased toward downside risks, helps mitigate potential disruptions by considering scenarios such as reduced revenues or lack of market access. Scenario analysis may also be employed, with probabilities assigned to potential events (e.g. new legislation). When models are used, confidence intervals account for the intrinsic volatility of cash flows.

Forecasting accuracy is particularly critical for countries with small liquidity buffers and limited market access, and vice-versa. Countries with larger liquidity buffers can absorb significant forecasting errors more easily. However, those aiming to minimise buffers to reduce carry costs may need to rely on deep markets to raise funds if a large forecasting error occurs. Despite these challenges, OECD countries have generally managed to avoid payment failures due to cash forecasting errors for an extended period. This is due to a combination of the accuracy of forecasts, liquidity buffers and the ease with which many OECD countries can access funding markets.

2.5.3. Performance assessments

Most countries prioritise ex-ante performance assessments of cash flow forecasts, while fewer than half also conduct backwards-looking ex-post evaluations.

Ex-ante analysis is generally a built-in part of the forecasting process and is often done automatically when creating models. It involves evaluating the forecast's accuracy before the actual cash flows occur, using the model's assumptions, confidence intervals, and historical data. While it may require manual adjustments or judgment for certain assumptions, the analysis is inherent in the modelling process to ensure that the forecast is as accurate as possible before it is used.

Ex-post analysis, when employed, aims mainly to inform improvements to future forecasts. These retrospective evaluations compare forecasted cash flows with actual outcomes and are particularly useful for finding biases and recurrent errors in forecasting models or frameworks. Ex-post analysis is especially important for countries that aim to minimise cash or liquidity buffers, where forecasting accuracy is critical to avoid cash shortages. An example is Hungary, where ex-post analysis is used to help calibrate buffer targets and improve forecasting accuracy.

Practices related to reporting, automation, and the scope of ex-post analysis differ significantly among countries that implement them. Some countries prepare detailed reports on ex-post analysis to present to a committee (e.g. Belgium) or for internal use (e.g. the United States). Others focus the analysis on specific items with large discrepancies (e.g. Japan). In some cases, ex-post analysis is not formally conducted, but realised daily cash flows are always presented alongside forecasted values (e.g. Luxembourg). In other instances, the process is fully automated, with reports that generate statistical data on forecasting errors, such as maximum and minimum errors and standard deviation, across different forecast horizons (e.g. Switzerland).

Although cash management teams frequently use Key Performance Indicators (KPIs), these often focus on broader liquidity indicators (explored in the next chapter) rather than forecasting accuracy. Examples of cash flow forecasting related KPIs are the difference between daily discrepancies between total projected liquidity need and final real liquidity need (e.g. Belgium) and deviations from the daily projected cash balance and deviations from monthly revenue and expenditure flows (e.g. Türkiye).

2.6. Governance structures for decision-making

Countries approach decision-making in cash management through three main types of governance structures: 1) integrated into existing debt management structures, 2) committees or structures with a broader mandate that also includes cash management in their scope, and 3) dedicated cash management committees.

Typically, participants in these arrangements include the front office and cash management team for countries with integrated structures in the DMO, senior officials from the fiscal authority or central bank in broader financial decision-making committees, and budget offices and fiscal authorities in dedicated cash management structures. In the latter case, cash management decisions may only be occasionally submitted to the authorities, with routine operational decisions handled by the cash management team and/or the DMO.

Cash management decisions are integrated into existing debt management structures, especially in the euro area, and in particular in Austria, Belgium, Iceland, Ireland and Portugal. For example, in Belgium, the daily Morning Committee and the Front Office team of the DMO are responsible for cash management decisions alongside broader debt-related matters. Similarly, decisions are made within the debt management units in Austria, Ireland and Portugal without needing a standalone cash management forum, as the cash management unit operates within the approved strategic and risk framework. Iceland also follows this model, where cash flow forecasts are presented to the debt management committee for deliberation, which also handles broader fiscal operations.

In contrast, Brazil, Switzerland, and the United States have established dedicated committees that meet regularly for cash management decisions. Brazil’s Cash Analyses and Monitoring Subcommittee, a subcommittee of the Financial Programming Committee, meets monthly to discuss cash flow projections and relevant issues. In Switzerland, the Asset-Liability Committee assesses liquidity risks on a quarterly basis. In the United States, senior officials make decisions based on cash flow forecasts provided by the Office of Fiscal Projections.

In other countries, there is no dedicated cash management committee, but cash management decisions can be made at senior and multi-body meetings or committees. In Hungary and Türkiye, for example, cash flow forecasts are presented at broader financing or risk management committees, where senior officials from ministries of finance, DMOs or central banks review cash and debt management operations for deliberation. This happens at the request of the cash management team in Hungary for extraordinary decisions and regularly (i.e. monthly) in Türkiye. In Japan, relevant divisions within the Finance Bureau of the MoF, such as the Treasury Division and Debt Management Policy Division, meet monthly to review income and expenditure projections and affirm a common understanding of cash flow forecasts.

Committees, when they exist and regardless of whether they are dedicated to cash management or not, often focus on strategic cash management decisions, leaving operational details to the cash management team and DMO. For instance, the Asset-Liability Committee sets liquidity targets in Switzerland, but the front office manages operations to meet them. Similarly, in Hungary, the DMO proposes liquidity buffer benchmarks approved by the MoF while retaining responsibility for operations like buybacks and credit lines. In Belgium, the DMOs’ Morning Committee provides high-level guidance, such as indicative auction sizes, but the front office has flexibility in managing daily cash operations.

2.7. Cash flow forecasting as an input for decision-making

Cash flow forecasts provide predictability and guide proactive decisions aimed at maintaining cash balances at an adequate level. When forecasts indicate potential liquidity challenges, DMOs adjust their financing operations accordingly. This creates a feedback loop where forecasts influence financial decisions, leading to updated projections. Adjustments often involve increasing short-term borrowing, such as issuing more T-bills, if cash balances fall below a lower bound. If cash balances are projected to exceed an upper bound, financing needs from money market operations are adjusted downwards.

Typically, cash flow forecasts influence decisions related to money market operations and, in some cases, fine-tuning the timing and amounts in debt issuance plans. Adjustments to bond issuances are usually minor, especially for large issuers, to maintain predictability and market liquidity. In Japan, Iceland, Switzerland and the United States, cash flow forecasting is crucial for short-term financing, managed through money market operations. In other countries, such as Luxembourg (Box 2.4) and in New Zealand, cash flow forecasting plays a more prominent role in adjusting the debt issuance plan. Additionally, these forecasts can also serve other purposes, such as informing the distribution of the treasury's financial resources across public banks (e.g. Türkiye) or supporting monetary policy decisions (e.g. Iceland).

Box 2.4. The decision-making process for new debt issuance by Luxembourg’s government

In Luxembourg, the DMO, comprising cash and debt management specialists, meets regularly with the Director of the treasury to determine the timing and structure of new bond issuances. These decisions are based on the current liquidity buffer and forecasts, ensuring a well-timed approach to meet funding needs. The DMO team also assesses different issuance options, including the type of instrument, maturity, and size, and regularly discusses with selected banks, often part of the syndicate, to refine these recommendations.

These banks provide critical updates on capital market conditions and bond pricing expectations across various maturities (from 7 to 30 years). Given the limited liquidity in Luxembourg’s secondary bond market, such pricing information is essential for accurate assessment ahead of new issuances, as the bond yield curve is incomplete (missing several maturity points). Due to Luxembourg's relatively small financing needs, which typically requires only one or two issuances per year (in Q2 and Q4), the publication of an annual issuance calendar is not considered necessary. This flexibility allows the DMO to select the optimal market window, balancing visibility for investors and managing competition with significantly larger and more regular issuances from other countries. Ultimately, the DMO team’s recommendations are submitted to the Minister of Finance for final approval, ensuring all issuance decisions align with the broader fiscal strategy.

Cash flow forecasting has a more limited impact on decision-making when governments run substantial cash surpluses. For example, in Ireland, cash flow forecasting currently has limited influence on the timing and amount of issuance. The cash balance is still relevant for determining the annual funding plan. Similarly, in Portugal, the cash management team may recommend early debt repayments to manage excess cash balances.

The frequency and granularity of the cash flow forecasting information required by decision-making also depend on the size of the cash or liquidity buffer. Frequent updates of daily cash balances are less necessary when the buffer is large enough to cover most intra-month funding needs. If the liquidity framework relies on smaller cash balances and makes greater use of short-term funding, such as repos and credit lines, to manage unexpected cash flows, more detailed and timely forecasts are essential to inform these decisions.

2.8. Conclusion

This chapter has highlighted the importance and challenge of cash flow forecasting in government debt management. Over the years, economic growth and increased borrowing have significantly increased the volumes of cashflows to be managed. A robust forecasting process enables governments to take proactive measures to ensure sufficient liquidity for future expenses, serving as a cornerstone for effective fiscal and debt management.

Key features of cash flow forecasting systems include integrating historical data with current fiscal projections, combining automated and manual data inputs, relying on short-term rolling forecasts to manage daily cash flows, maintaining close collaboration between government departments, and having decision-making mechanisms (e.g. committees) to act on forecast outcomes.

However, producing accurate and timely cash flow forecasts poses several challenges. These include dependence on third-party information, often without binding mechanisms, which can be unreliable or delayed, and the inherent uncertainty in fiscal projections. Risk aversion also plays a role, with some countries adopting conservative approaches to insure against potential inaccuracies. Policy limits and requirements add further complexity, requiring a balance between precision and flexibility.

Looking ahead, advancements in regulation and technology offer opportunities to improve cash flow forecasting. Regulatory changes may promote greater standardisation and transparency in forecasting practices, contributing to stronger and more reliable cash management strategies. Enhanced data analytics, machine learning, and real-time data integration could lead to more dynamic and accurate forecasting models.

By addressing these challenges and embracing future developments, governments can enhance their ability to produce precise and timely cash flow forecasts, ultimately improving financial stability and efficiency.

References

[1] Foxall, S. and L. Policino (2025), Sovereign retail debt programmes and instruments: A review of country practices, OECD Publishing, Paris, https://doi.org/10.1787/e2a782d0-en.

Annex 2.A.  Different ends of the forecasting spectrum: From Iceland to the United States

The approach to treasury cash flow forecasting can vary significantly between large countries like the United States and much smaller countries like Iceland, reflecting differences in economic scale, complexity, and available resources. The US economy is almost a thousand times larger than Iceland’s and the US Treasury spends approximately 630 times more in nominal terms.

While cash flow forecasting is done by 13 US Treasury Office of Fiscal Projections employees, in Iceland, it is done part-time by a Ministry of Finance and Economic Affairs staff member. In the United States, this Office is also supported by approximately 25 employees of the Federal Reserve Bank of St. Louis, serving as the treasury’s fiscal agent, who provide forecasting system operations, maintenance and development services, and modelling and infrastructure support.

There are numerous differences in the treasury cash flow forecasting methods used in large and small countries. Greater deviations from the forecast are typically more common for smaller countries. Companies tend to be proportionally larger and fewer, leading to larger tax receipts that can be difficult to time and create significant deviations from the forecasts. By the same token, large investment projects can lead to greater deviations in smaller economies. Iceland’s smaller scale offers greater agility in adjusting forecasts and implementing fiscal measures.

Iceland’s cash flow projections are generated with numerous programmes. Historical data is extracted from a Structured Query Language (SQL) database and loaded into a Power BI model so that it may be defined and projected into a desired form. The database is stored at a local private IT company but managed by the Financial Management Authority along with the Power BI model. The forecasting part of the model is presently done in the Ministry of Finance and Economic Affairs using Excel but will soon also be moved into Power BI. Daily fluctuations are based on previous years, but change is in line with various economic indicators such as inflation, wage growth and private consumption.

Due to its size and complexity, the United States requires a more structured and frequent update mechanism to help maintain forecast accuracy. The US cash flow projections consist of approximately 250-line items or categories of cash flows and a similar number of line items for categories of debt covering a 12-to-15-month time horizon. It includes the investments and redemptions of nearly 200 trust and investment funds, although many smaller funds are grouped and forecast in bulk. A new quarter is added to the forecast on a rolling basis every quarter, and the quarterly benchmark forecast forms the basis of the US Treasury’s Marketable Financing estimates announced to the public. The initial forecasts are created through statistical models maintained outside of the financial planning system (FPS) and monthly spreading and trend analysis functionality that are part of the FPS. The forecast is updated continuously to account for the daily variance between estimated cash flows and actual results, new information reported from other government agencies, or changes in underlying assumptions, including economic projections.

Reports summarising the daily cash balance, debt, daily flows and variance from the benchmark and other reference periods are run at least twice daily, once in the morning and then again towards the end of the business day. The end-of-day forecast is stored for historical variance analysis, allowing analysts to assess this through various parameters, including individual or cumulative days or months. This statistical data are then used to project a confidence interval around the forecast. Additionally, the forecasting system provides periodic intra-day reports used to monitor the variance between the current day’s cash flow projections and actual results.

Notes

← 1.  Fiscal authority refers to the Ministry of Finance or treasury, depending on the country.

← 2.  QT refers to the process through which central banks reduce the size of their balance sheets by allowing bonds they hold to mature without reinvesting the proceeds or by actively selling them, effectively reversing the effects of earlier QE. After the global financial crisis, central banks in major economies implemented QE to stimulate growth, which involved large-scale asset purchases, especially of government bonds. Over the last decade, these purchases expanded central bank balance sheets significantly, lowering long-term interest rates and fostering liquidity. However, as inflation surged globally following the COVID-19 pandemic and war-related shocks, central banks initiated QT, starting in 2022-23. This marked a shift towards tighter financial conditions, with central banks such as the Federal Reserve, BoE, and the ECB actively reducing bond holdings by either selling or not fully reinvesting maturing securities.

← 3.  Austria’s federal government is an important source of funding for subnational governments; it issues debt and transfers proceeds to them upon request of the Federal Minister of Finance.