## Cash Flow Forecasting in Fragile and Low-Income States: Guidance Note (INTRODUCTION and OVERVIEW)

## Source details

**Canonical URL:** [Cash Flow Forecasting in Fragile and Low-Income States: Guidance Note (INTRODUCTION and OVERVIEW)](https://www.imf.org/-/media/files/capacity-developement/home/eng-cash-flow-forecasting-in-fragile-and-low-income-states-guidance-note.pdf)

## Other formats

- [Markdown version](/-/media/files/capacity-developement/home/eng-cash-flow-forecasting-in-fragile-and-low-income-states-guidance-note.pdf.md)
- [Structured JSON version](/-/media/files/capacity-developement/home/eng-cash-flow-forecasting-in-fragile-and-low-income-states-guidance-note.pdf.json)

---

### Purposes of cash flow forecasting (IMF guidance note)
- Facilitate and support orderly execution of the budget: ensure budgeted expenditure is smoothly financed, avoid delays, and give early warning of cash flow shortages that could force expenditure delays or cash rationing.
- Devise strategies for active cash management: plan ahead and adjust financing plans where possible so cash is available when required while avoiding unnecessarily idle cash.

### Context and supporting materials
- The IMF prepared linked materials to support cash forecasting capacity, including:
  - A “How-To-Note” (Sailendra Pattanayak, Racheeda Boukezia, Yasemin Hurcan, and Ramon Hurtado “How to Build Cash Management Capacity in Fragile States” (IMF 2022)).
  - A questionnaire-based diagnostic tool (Racheeda Boukezia, Jean-Baptiste Gros and Sailendra Pattanayak “Questionnaire Based Diagnostic Tool to Assess Cash Management Capacity” (IMF 2022)).
  - An Excel-based tool for preparing cash forecasts; the guidance note explains that tool.

### Intended users and approach
- Tool intended for treasury or officials preparing cash flow forecasts; designed primarily for the practitioner and for presenting forecasts and policy advice to senior decision makers.
- Users should focus on sections relevant to country circumstances; the note reflects a range of experiences and capabilities.

### CFAT design and deployment
- Cash Forecasting and Analysis Tool (CFAT) is primarily top-down using past patterns, but designed to incorporate bottom-up inputs from MDAs and RAs.
- Phased deployment: start with simplifying assumptions and progress to monthly then weekly forecasts as capacity grows; daily forecasts are a medium-term objective.
- Practical requirements:
  - Excel-based spreadsheet; deploy on Excel 2013 or later.
  - Users should enable Macros after opening the file.

*Source: IMF Cash Flow Forecasting guidance note (INTRODUCTION and OVERVIEW sections).*

### Functional capabilities and link to budget exercises
- Tool supports:
  - Constructing from scratch a cash plan and cash forecast for the upcoming year.
  - Adjusting forecasts throughout the year as outturn data become available.
  - Presenting policy proposals to decision makers in response to forecast signals.
- IMF distinguishes three related exercises:
  - Budget profile across the year (usually monthly), coordinated by BD or Treasury.
  - Cash plan (cash program), usually monthly, starting from the approved budget and adjusted for non-cash expenditures, arrears, and financing not in the budget.
  - Cash flow forecast: “an estimate of future government cash inflows and outflows, with a view to taking action necessary to ensure that sufficient funds are always available to meet any net government cash requirements; and, in any period where there is a net cash surplus, to ensure that it is used to best advantage.”
  - At the beginning of the year plan and forecast are similar; forecast should be unbiased and may diverge from plan as year unfolds; toward end of budget year forecast must look to following year.

*Source: IMF Cash Flow Forecasting guidance note (INTRODUCTION and OVERVIEW sections).*

### Institutional Responsibility and Usage

### Where responsibility may sit
- Cash flow forecasting may lie in the Treasury department of the MoF, the budget unit, or a unit combining cash and debt management; guidance note uses “the Treasury” as shorthand.

### Role of MDAs/RAs and weekly forecasts
- Tool designed to incorporate bottom-up flows from MDAs and RAs as inputs become available and important for high-quality forecasts.
- Weekly forecasts rely more heavily on MDA/RA inputs as they develop.

### Decisions About Forecast Characteristics

### Key choices practitioners must make
- Coverage and detail of the forecast.
- Forecast horizon.
- Time granularity (monthly, weekly).
- Refresh rate and frequency of circulation.

### Considerations for fragile and low-income countries
- Capacity constraints may limit long horizons, high granularity, and frequent refresh rates.
- Initial focus on the period immediately ahead (especially to move away from cash rationing).
- Tool is flexible and allows projections to become richer as capacity, data, and experience grow.

*Source: IMF Cash Flow Forecasting guidance note (INTRODUCTION and OVERVIEW sections).*

### Box 2: Key Forecast Characteristics

### Coverage and detail
- Cover at a minimum all cash flows that have a material impact on central government bank balances; avoid unnecessary detail.
- Identify largest budget lines and main revenue sources; forecast is for summary data for decision makers, not monitoring MDA/RA budget performance.
- Tool focuses on monthly forecasts and illustrates weekly development; daily forecasts are a medium-term objective.

### Forecast horizon and refresh
- Link tool to the annual budget (normally formally approved budget).
- Maintain particular focus on months immediately ahead with a rolling forecast at least three months ahead and ideally more; potentially roll across the end of the budget year.
- Refresh at least every month with new outturn data and updated information—at least from large MDAs/RAs.
- Distinguish full and partial refreshes depending on quality, extent, and frequency of new data.
- A refresh in the third week of the month often best balances availability of outturns and need to prepare updated forecast before the next month.
- Circulate revised forecast normally every month, or more often if frequently refreshed; additionally report cash outturns weekly or daily and flag unexpected divergence.

### Time granularity
- Initially identify monthly cash flows; increase granularity over time.
- Weekly forecasts feasible soon in many fragile states; daily forecasts follow later.
- Many large cash flows are known with precision (debt servicing, monthly salaries, some transfers and revenue flows); weekly/daily forecasts capture cumulative within-month movements.

### Target variable (TSA and controlled cash)
- Bottom line should be cash balance in bank accounts MoF/Treasury controls—probably the TSA where established.
- TSA should include all central government cash balances (and in countries including SNG balances, all general government cash balances), though developing a comprehensive TSA is challenging.
- Cash defined as government assets perfectly liquid and usable immediately; adjust budget figures and MDA/RA data for non-cash (accrual) elements and for cash flows outside the budget (e.g., to honor arrears).
- Governments should try to maintain a cash buffer; example implicit buffer target in illustrative chart is 25 million.

### Data format and classification
- Use a format flowing down the page from:
  - Revenue
  - Current Expenditure
  - Interest Payments
  - Operating Cash Surplus
  - Capital Expenditure
  - Net surplus/deficit
  - Financing, net
  - Net change in Cash
- Default classification based on IMF GFSM format; emphasis on economic classification, but mixed classification acceptable where necessary.
- Identify all major cash flows and simplify forecasting for lesser or stable flows.

### Tool structure (CFAT) — key worksheets and functions
- Single spreadsheet with Main worksheet and inter-linked supporting worksheets.
- Settings worksheet: input target budget year; first month of cash forecast; last full month before forecast period; availability of outturn data; opening TSA balance for first day of target budget year (assumed January 1 by default).
- Main worksheet: Budget Outturns, Target Year Budget, Adjusted Budget Numbers, Summary, Outturns & Adjustments, Revised plans, Policy Options.
- Back-end worksheets: Historical Charts, Weekly v Monthly, Auxiliary tables, Script, Parameters, Tool Structure, step-by-step guide.
- Templates for RA and major MDA rolling forecasts and a bridge between calculated TSA and central bank statement.
- User inputs go into light-blue shaded cells; charts compare forecasts with outturns and act as visual aids.

### Initial user choices and conventions
- Decide weights for each previous year’s outturns.
- Decide forecast detail extent (fewer lines preferred) and rounding convention: budget totals rounded to four significant figures, spreadsheet cells typically 2-3 significant figures.
- Adjust if January is not first month of budget year.
- Use domestic currency (DX) throughout, including DX equivalents of foreign currency (FX) flows.
- Use a new spreadsheet each month and always back it up.
- Enter up to three years of historical data normally; facility exists to enter five years. One year is required to calculate initial monthly distributions.
- Prefer fewer rows; when adding rows use “add row” menu and “adjust formulas” button to maintain consistency.

### Forecast maintenance and learning
- Prepare monthly rolling forecasts focused on the next three months; at start of year base month input as “none.”
- Maintain one spreadsheet per month to retain record for forecast-error analysis and learning.

*Source: IMF guidance note — Box 2: Key Forecast Characteristics*

### Annex B: Techniques for Cash-Flow Forecasting

### Data protection and spreadsheet management
- Back up and password protect spreadsheets; add password when spreadsheet is first prepared.
- Spreadsheet autosaves every 30 minutes.
- Passwords help prevent unauthorized changes that might imply misleading policy advice.

### Adjustments — scope and practice
- Adjustments may be technical or forecasters’ judgements to ensure flows refer to cash and affect the TSA.
- Make only adjustments likely to have a material impact.
- Enter the whole of the newly-adjusted row in Adjusted Budget section when an adjustment is made.
- Distinguish adjustments that convert budget numbers to cash from adjustments related to TSA coverage.

### Revenue flow adjustments
- Leakage: non-tax receipts retained by an MDA in an MDA-controlled bank account are not available to the Treasury; options are general assumption about leakage or identify main culprits and open dialogue.
- Recognition of receipts: adjust monthly flows by average time lag between receipt by banks and transfer to TSA where necessary.
- End-of-month timing effects: if large flows fall on weekends, they may be pushed into following month; adjust using past profiles.

### Expenditure adjustments
- Payments for commitments outstanding at end of year may be paid in following year; explore previous years and reflect impacts.
- Unspent provision rolled over may not be reflected in new year budget; payment of arrears should be distinguished from complementary period payments.
- Recognition of payments: cash moves when checks presented and cleared; central or commercial bank data can supply averages.
- Contractual deferrals, extra payments (e.g., “13th month” bonus), linked flows, donor timing, imprests and advances, netting and non-cash treatments require attention.

### Banking and financing flows — coverage and accounting issues
- TSA defined as accounts whose balances can be controlled or accessed by government; short-term control/access may be constrained.
- Coverage divergences: TSA may be less than, equal to, or larger than budget coverage; EBF inclusion affects forecasts.
- Pre-funding of banks (“seed money”) may be required; often stable month-to-month and may be ignored for monthly forecasts.
- Accounting treatment of securities may diverge from cash flows; identify rare adjustments with accountants/statisticians.

### Constraints accessing cash within the TSA — types and handling
- Types include earmarked flows, dispersed accounts, legacy sub-accounts, accounts held outside the TSA at the central bank.
- Interim workarounds: aggregate low-transaction debt-related accounts in spreadsheet; rely on monitoring central bank/commercial bank statements to trigger transfers.
- Reporting: government should receive regular reports on balances outside the TSA (EBF accounts, donor project accounts, MDA-controlled accounts) and incorporate transferable flows into forecasts.

### Placement of adjustments in the tool
- Place adjustments in “Adjusted Budget Numbers” or “Outturns and Adjustments.”
- Place specific, monitorable adjustments against respective flows; place judgemental adjustments below surplus/deficit line.

### End of budget year procedures and forward practice
- Forecaster should look months ahead into following budget year.
- For new budget year:
  - Save file as new version and clear old-year data.
  - Shift historical data: year T-1 becomes T-2, etc.
  - Copy outturn data for nearly completed year to historical T-1.
- Practical steps for inserting short-term columns: insert columns after December, copy values for January/February outturns, link TSA opening balance for January to December closing, adjust copied data as necessary.

### Detailed flows and previous year outturns
- General approach: identify monthly expenditures over previous three years; calculate percentage falling in each month; average the three percentages for each month; apply averaged percentage to current year budget figures.
- Use approach with care and make adjustments as needed.

*Source: Annex B of IMF Guidance Note.*

### Box 4: Possible Adjustments to Outturn Data

### When to apply adjustments
- Adjust outturns recorded in budget accounting rather than cash terms.
- Overwrite outlier historical data when one-off issues affect series and do not affect many lines.
- Ignore historical profiles when procedural or classification changes render earlier profiles unhelpful.

### Use of reported profiles and missing historical data
- Profiles from BD or MDAs may be unreliable; RAs’ profiles may cover largest taxes only.
- If no historical data exist, construct a prior year budget outturn and enter it; tool assumes equal split across 12 months if nothing entered.

### Consistency and weighting historical outturns
- Ensure consistency when current-year outturns become prior-year data; adjust for differences in cash vs budget classification.
- Spreadsheet can use last 5 years; user can select number of years (1 to 5).
- Example weightings:
  - Three-year weights example: 0.45, 0.35, 0.20.
  - Five-year weights example: 0.35, 0.30, 0.20, 0.10, 0.05.
- Default is equal weights determined by user-input number of years.
- Note: “The responses to Covid-19 might suggest that 2020 or 2021 should always be given a lower weight.”

### In-year outturns, reconciling TSA balances
- Monthly or weekly outturns must be entered by hand unless linked to IFMIS; must be in cash or adjusted.
- Central bank statements are an additional source; if calculated TSA balance differs from actual bank statement and gap is large, roll it into future forecasts and account for it in policy options.

### Handling incomplete outturn data and alternative sources
- Use assumptions for allocation to rows or temporarily record unspecified flows in “other” categories.
- Template Bank Account-GFS Bridge provided to allocate bank statement data lacking full classifications; not linked to other worksheets.
- Obtain local bank statements or daily reports where payments dispersed across local banks.

### Recording adjustments and revised plans
- Enter fully adjusted monthly forecasts (not just changes to previous forecasts). New rows will be picked up in Revised Plans section.
- Fundamental adjustments to future years belong in Adjusted Budget Numbers and Revised Plans.

### Inputs from MDAs/RAs and focus on major flows
- Develop contacts in key MDAs/RAs and require forecasts at least monthly for three months ahead.
- Forecasters should focus on major flows and aggregate inputs; assume remainder as flat or in line with plans.
- Initially request information from budgetary units responsible for about 75-80 percent of government expenditure (often 5-7 MDAs).

*From Cash Flow Forecasting in Fragile and Low-Income States: Guidance Note (Box 4).*

### Box 6: Revenue and Expenditure Templates

### Revenue template
- Forecasts cover three main tax groups; weekly forecasts preferred but monthly shown as consistency check.
- Comparisons with initial monthly and weekly plans included to identify trends such as persistent shortfall.
- Exercise framed as identifying what will happen, not to hold MDAs/RAs to account.
- Chart optional but recommended.

### Expenditure templates
- Two templates: current expenditure and capital expenditure; design similar to revenue template.
- Individual MDA contributions to total expenditure typically small relative to RAs’ contribution to revenue; may require extra decimal places.
- Aggregate contributions from selected large MDAs and allow for proportion of total expenditure covered.

### Weekly forecasts — construction techniques
- Combine techniques: allocate known-date flows to relevant week; model regular patterns; spread some monthly figures equally across 4 or 5 weeks and overwrite for lumpy transactions.
- Early practice: assume revenue at end of month and payments at beginning until in-month profiles are understood.
- Weekly forecasts should be broadly consistent with monthly equivalents; monitor cumulative weekly vs monthly totals.
- Weekly outturns entered directly into Outturns & Adjustments and recorded for performance monitoring.

### Weekly calendar issues
- Handle months with 4 or 5 weeks by identifying months with 4 or 5-week periods (preferred) or keep strict monthly calendar with stub periods of 1–3 days.
- Be careful with weekends and lumpy flows that may shift into adjacent month-week mapping.

### Policy Options section
- Three segments:
  - Central forecast: recent outturns, projections, TSA projections, discretionary vs non‑discretionary expenditures.
  - Sensitivities/what‑ifs: show impacts if risks materialize and attached probabilities.
  - Policy responses: possible responses to emerging cash positions; temporary changes may shift cash timing.
- Policy Options table picks up data from Revised Plans section; emphasis on next 3 months.

### TSA projections, buffer, and reconciliation
- Bottom-right chart shows TSA under central base, “worst” case (plausible not extreme), and impact of policy options.
- Example includes implicit cash buffer target of 25 million; action identified to prevent balance falling below this level.
- Outturns & Adjustments includes reconciliation row for gap between top-down calculated TSA and actual TSA balance; Policy Options adds row for this difference and treats it as a risk factor.
- Large discrepancy treatment: carry drain forward in future forecasts unless reversal likely; ignore drain when cumulating flow effects of other risks but add back when calculating total impact on TSA.

### Practical notes
- Three templates included as worksheets for convenience; not linked to other worksheets and typically copied and adapted before sending to MDAs/RAs.
- Policy Options chart usable dynamically in CCC meetings.
- Spreadsheet not designed to monitor budget execution performance; focus is forecasting cash and informing short-term policy decisions.

*Source: IMF Guidance Note — Box 6: Revenue and Expenditure Templates.*

### Annex B: Analyzing Forecast Errors

### Purpose and initial decisions
- Two purposes: plan responses to forecast deviations and learn lessons for future forecasts.
- Decide how far to drill down and the relevant period to analyze (often start with forecast as made at start of month).
- Identify constrained variables (e.g., lines constrained by cash rationing) and policy variables (e.g., domestic securities issuance removed from study where policy choice dominates).

### Independence, timing, and serial correlation
- Identify whether errors are independent across time and components; test for serial correlation by plotting errors against previous period errors.
- Timing changes cause serial correlation; within-year net error may be much smaller than monthly errors.

### Distribution and one-off events
- Forecast errors unlikely to be normally distributed due to serial correlation and many one-off events.
- Identify triggers of large errors and assess risk of repetition.

### Practical examples and implications
- Revenue profiling issues: corporate income tax affected by end-of-month weekends; dividends require direct information.
- Expenditure examples:
  - Payments outstanding from previous year of DX 3 million assumed paid DX 2 million in January and DX I million in February.
  - Estimated DX 1 million of budgeted payments likely outstanding at end of target year.
  - SNGs assumed entitled to 25 percent of all tax receipts collected in previous month.
- Capital and financing examples:
  - Local counterpart funds assumed to be 8 percent of externally-financed expenditure in absence of other information.
  - Financing examples: issuance proceeds 97 this month, redemption 100 in 3 months’ time.
  - Borrowing from central bank assumed zero in initial plan; advances/overdrafts repaid within same year and entered with negative sign in ‘new borrowing’ line.
- Project-related donor flows linked to project timing; donor flows may lag expenditure unless pre-financing agreed.

*Source: ANNEX B of IMF Guidance Note.*

### Annex D: The Bridge Table

### Purpose and usage
- Bridge table (worksheet “Bank Account-GFS Bridge”) illustrated for mapping central bank statement transactions to forecast categories.
- Assumes central bank statement in Excel with sufficient descriptive/coded information to identify majority of transactions.

### Practical workflow
- Use same table for each of 5 days of the week; sum identified flows over the week and compare with an “average” week (annual totals divided by 52).
- Add notes column for reasons flows differ from average or forecast.
- Sub-total identifiable flows by plan/forecast categories and transfer sub-totals to bridge table; treat smaller unidentified flows as residuals.

### Key assumptions and recommendations
- Bank statements unlikely to allow identification of all flows.
- Net out intra-TSA transfers.
- Focus on identifying large flows; smaller flows in residual category.
- Use coding linking to IFMIS where possible.
- Work closely with central bank and RAs to ensure statement provides manipulable detail.
- Treat provisional classifications as subject to follow-up if amounts are material.

### Illustrative assumptions and mappings
- Some income tax flows identifiable; large residuals likely personal income tax in an “unspecified” line.
- Salary payments identified; many current expenditures and smaller capex not categorisable and allocated to goods and services.
- Financing transactions often individually large and identifiable.

*Source: IMF Guidance Note — ANNEX D: THE BRIDGE TABLE.*

*Source: IMF Cash Flow Forecasting guidance note (INTRODUCTION, OVERVIEW, Box 2, Annex B, Box 4, Box 6, Annex D).*

### INTRODUCTION ...........................................................................................................

### INTRODUCTION

### Purposes of cash flow forecasting
- The ability to forecast flows across government bank accounts has two broad purposes, according to the IMF guidance note:
  - To facilitate and support orderly execution of the budget; to ensure that budgeted expenditure is smoothly financed avoiding delays; and to give early warning of any cash flow shortages that might force the government to delay expenditures or otherwise ration cash if it is unable to increase borrowing.
  - To devise strategies for active cash management, i.e., planning ahead, and where it can, adjusting financing plans, to ensure that cash is available when required while avoiding any unnecessarily idle cash that is not earning a satisfactory return.

### Context and supporting materials
- The IMF prepared a set of linked documents to support cash forecasting capacity in fragile and low-income states:
  - A “How-To-Note” (Sailendra Pattanayak, Racheeda Boukezia, Yasemin Hurcan, and Ramon Hurtado “How to Build Cash Management Capacity in Fragile States” (IMF 2022)).
  - A questionnaire-based diagnostic tool (Racheeda Boukezia, Jean-Baptiste Gros and Sailendra Pattanayak “Questionnaire Based Diagnostic Tool to Assess Cash Management Capacity” (IMF 2022)).
  - An Excel-based tool for preparing cash forecasts; this guidance note explains that tool.

### Intended users and approach
- The tool is intended for treasury or other officials responsible for preparing cash flow forecasts; it is designed primarily for the practitioner but also provides recommendations for presenting forecasts and policy advice to senior decision makers.
- Users are advised to focus on sections relevant to their country circumstances; the note reflects a range of experiences and capabilities.

### OVERVIEW OF THE TOOL

### Top-down primary design with bottom-up inputs
- The Cash Forecasting and Analysis Tool (CFAT) is primarily a top-down mechanism making use of past patterns, while being designed to incorporate bottom-up information from Ministries, Departments and Agencies (MDAs) and Revenue Authorities (RAs) as that information becomes available.

### Phased deployment and granularity
- Full deployment can grow over time:
  - Users can start with simplifying assumptions and progressively use more capabilities as capacity develops.
  - Users will probably want to start with monthly forecasts, but move to weekly forecasts as capacity and confidence grows. The spreadsheet facilitates this move.
  - Users can incorporate existing templates, charts or categories by inputting data in the relevant lines in the tool.

### Practical requirements
- The CFAT is an Excel-based spreadsheet that should be deployed on Excel 2013 or later.
- Users should enable Macros after they have opened the file.

### KEY CHARACTERISTICS

### Functional capabilities
- The tool supports:
  - Constructing from scratch a cash plan and cash forecast for the upcoming year.
  - Adjusting forecasts throughout the year as outturn data become available.
  - Presenting policy proposals to decision makers in response to forecast signals.

### Relationship to other budget exercises (Box 1: Cash Plans and Cash Forecasts)
- The IMF distinguishes three related cash planning and forecasting exercises in the budget year:
  - Preparation of a budget profile across the year (usually monthly), coordinated by the Budget Department (BD) or the Treasury within the Ministry of Finance (MoF). The profile is often based on submissions from spending and revenue departments and may be the basis for the release of spending authority.
  - Preparation of a cash plan (cash program), usually monthly and starting from the approved budget, adjusted for non-cash expenditures, arrears, and potentially including financing transactions not in the budget. The cash plan checks consistency of the budget and planned financing and is typically constrained by approved budget targets.
  - Preparation of the cash flow forecast, defined as “an estimate of future government cash inflows and outflows, with a view to taking action necessary to ensure that sufficient funds are always available to meet any net government cash requirements; and, in any period where there is a net cash surplus, to ensure that it is used to best advantage.” At the beginning of the budget year, the cash plan and cash forecast are likely to be very similar, but the forecast should be an unbiased and unconstrained best estimate and will often diverge from the plan as the year unfolds. Toward the end of the budget year the cash forecast must also look to the following year even if there is no formal budget.

### INSTITUTIONAL RESPONSIBILITY AND USAGE

### Where responsibility may sit
- The tool makes no precise assumption about institutional responsibility; cash flow forecasting may lie in the Treasury department of the MoF, the budget unit, or a unit combining cash and debt management.
- The guidance note uses “the Treasury” as shorthand for these options.

### Role of MDAs/RAs and weekly forecasts
- The tool is designed to incorporate bottom-up flows from MDAs and RAs as these inputs become available and prove important to high-quality forecasts.
- Weekly forecasts are likely to rely more heavily on inputs from MDAs/RAs as they develop.

### DECISIONS ABOUT FORECAST CHARACTERISTICS

### Key choices practitioners must make
- Coverage and detail of the forecast.
- Forecast horizon.
- Time granularity (monthly, weekly).
- Refresh rate and frequency of circulation.

### Considerations for fragile and low-income countries
- Capacity constraints in fragile states or low-income countries may limit long horizons, high granularity, and frequent refresh rates.
- Initially, forecasters will focus on the period immediately ahead (especially when seeking to move away from cash rationing).
- The tool is flexible and allows projections to become richer as capacity, data, and experience grow.

*Source: IMF Cash Flow Forecasting guidance note (INTRODUCTION and OVERVIEW sections).*

### Box 2: Key Forecast Characteristics

### Box 2: Key Forecast Characteristics

### Coverage and detail of the forecast
- Cover at a minimum all the cash flows that have a material impact on the central government’s bank balances, while avoiding extensive unnecessary detail.
- Identify largest budget lines and main revenue sources; forecasts are not a tool to monitor budget performance of MDAs/RAs—decision makers will need only summary data.
- The tool focuses primarily on building monthly forecasts, but illustrates how weekly forecasts might be developed in tandem; subsequently daily forecasts can be developed from the weekly forecasts. Daily forecasts are a more medium-term objective for fragile or low-income countries.

### Forecast horizon and refresh
- Link the tool to the annual budget (normally the formally approved budget), accommodating whatever is input.
- Maintain particular focus on the months immediately ahead, with a rolling forecast at least three months ahead and ideally more (and potentially rolling across the end of the budget year).
- Refresh the forecast at least every month with new outturn data and updated information—at least from large MDAs/RAs.
- Distinguish between full and partial refreshes depending on quality, extent and frequency of new data.
- A refresh in the third week of the month is often the best trade-off between availability of outturn information for the previous month and the need to prepare an updated forecast before the start of the following month.
- Circulate the revised forecast normally every month, or more often if frequently refreshed; additionally report cash outturns on a weekly or daily basis and flag unexpected divergence.

### Time granularity of forecast
- Initially identify monthly cash flows.
- Increase granularity over time: weekly forecasts are feasible soon in many fragile states; daily forecasts follow later.
- Many large cash flows are known with precision (e.g., debt servicing payments, monthly salaries, some transfers and some revenue flows), so moving to weekly and then daily forecasts captures substantial cumulative cash movements within a month.

### The target variable (TSA and controlled cash)
- The bottom line of the forecast should be the cash balance in the bank accounts over which the MoF or Treasury has control—probably the balance in the Treasury Single Account (TSA) where established.
- The TSA should include all central government cash balances (and in countries that include SNG balances, all general government cash balances), but developing a comprehensive TSA is challenging.
- What matters is the cash over which the government at the center has control and which is available to meet its commitments, whether in a single account or a network of accounts that gives a consolidated view of government cash resources.
- Emphasize cash defined as government assets that are perfectly liquid and can be used immediately; adjust budget figures and MDA/RA data for non-cash (accrual) elements and for cash flows outside the budget (e.g., to honor arrears).
- Governments should try to maintain a cash buffer to protect the budget from unexpected outflows or financial market disruption; one output of the forecast is the decisions needed to sustain the required buffer or reduce the likelihood it is exhausted, pushing the MoF to ration cash.

### Data format and classification
- Avoid a simple inflows/outflows split; use a format that flows down the page from:
  - Revenue
  - Current Expenditure
  - Interest Payments
  - Operating Cash Surplus
  - Capital Expenditure
  - Net surplus/deficit
  - Financing, net
  - Net change in Cash
- This format emphasizes financing choices as the main policy variable.
- Default classification is based on the IMF GFSM format; emphasis is on economic classification of cash flows, but a mixed classification can be used where necessary.
- Where economic classification is unavailable or lagged, adopt a mixed approach (e.g., economic classification for major MDAs and administrative classification for the rest).
- Forecasters must identify all major cash flows—those that have a material impact on the TSA balance—and may simplify forecasting for lesser or more stable flows.

### Tool structure (CFAT) — overview and worksheets
- The tool is a single spreadsheet with a Main worksheet and inter-linked supporting worksheets; it helps prepare an annual cash plan and generates a short-term forecast as outturns and revisions are incorporated.
- The spreadsheet:
  - Is designed to support user choices.
  - Includes charts as visual aids to decision making.
  - Allows users to input sensitivity scenarios and policy responses.
- Key worksheets and functions:
  - Settings: input target budget year; first month of cash forecast; last full month before forecast period; availability of outturn data; opening TSA balance for first day of target budget year (assumed January 1 by default).
  - Main: Budget Outturns (historical monthly revenue and expenditure data), Target Year Budget (total yearly allocations), Adjusted Budget Numbers (cash adjustments and comments), Summary (monthly/weekly aggregation, finalization of forecast, policy options), Outturns & Adjustments, Revised plans, Policy Options.
  - Back-end worksheets: Historical Charts, Weekly v Monthly, Auxiliary tables (consistency tests, weeks per month calculation), Script (labels), Parameters (menu ranges), Tool Structure, step-by-step guide.
  - Templates for RA and major MDA rolling forecasts and a bridge between the calculated TSA and the central bank bank statement.
- User inputs go into light-blue shaded cells on the Main worksheet; charts compare forecasts with outturns and act as visual aids.

### Using the tool — initial choices and conventions
- Initial user choices:
  - Decide weights for each of the previous years’ outturns.
  - Decide extent of forecast detail (fewer lines preferred).
  - Decide rounding convention: budget totals rounded to four significant figures, with spreadsheet cells typically being 2-3 significant figures.
  - Adjust the illustrative spreadsheet if January is not the first month of the budget year.
  - Use domestic currency (DX) throughout, including DX equivalents of foreign currency (FX) flows.
  - Use a new spreadsheet each month and always back it up.
- Historical data entry:
  - Normally enter up to three years of historical data; facility exists to enter five years.
  - If fewer than three years are available, enter whatever data are available (one year is required to calculate initial monthly distributions).
  - Weights to previous years are entered in “Budget outturns adjustments” in Settings.
- Practical guidance:
  - Prefer fewer rows; rows can be left blank rather than deleted.
  - When adding rows, use the “add row” menu and the “adjust formulas” button to maintain consistency across sections.
  - Distinguish capital expenditure between externally financed projects and domestically financed projects.
  - All data are in DX; DX should be used throughout even if FX accounts exist.
  - Use a copy of the previous month’s spreadsheet as the base for the next month to preserve historical forecasts for forecast-error analysis.

### Forecast maintenance and learning
- Prepare monthly rolling forecasts focused on the next three months; at the start of the year preparing the 3-month forecast, the base month should be input as “none.”
- As the year proceeds, the base month will be that before the start of the 3-month forecast period; the spreadsheet picks up available outturn data.
- Maintain one spreadsheet per month to retain a record for analysis of forecast errors and to learn lessons to improve future forecasts.

*Source: IMF guidance note — Box 2: Key Forecast Characteristics*

### Annex B outlines some techniques.

### Annex B: Techniques for Cash-Flow Forecasting in Fragile and Low-Income States

### Data protection and spreadsheet management
- All spreadsheets should be backed up and password protected appropriately.
- The password should be added when the spreadsheet is first prepared.
- The spreadsheet autosaves every 30 minutes.
- Excel files are not always stable and can be changed accidentally; passwords help prevent unauthorized changes that might imply misleading policy advice.

### Adjustments — scope and practice
- Adjustments may be technical or forecasters’ judgements; they are made to ensure flows refer to cash and affect the TSA.
- The user need only make adjustments likely to have a material impact on the forecast.
- If an adjustment is made, the whole of the newly-adjusted row is entered in the Adjusted Budget section (not just pluses or minuses on the Target Year Budget rows). The same applies to adjustments following input of outturns.
- Many adjustments are demonstrated in the Adjusted Budget Numbers section of the Main worksheet; they could, however, be made to any series.
- It is useful to distinguish between:
  - adjustments that convert budget numbers to cash; and
  - adjustments related to TSA coverage.

### Revenue flows — potential adjustments
- Potentially important: Leakage — non-tax receipts retained by an MDA in a bank account controlled by the MDA will not be available to the Treasury. Options:
  - make a general assumption about such “leakage”; or
  - identify main culprits and open a dialogue with them.
- Recognition of receipts: Outturn data may be defined in budget terms but not in cash terms (e.g., accrual components and timing differences). Example: tax paid into a RA’s account at a commercial bank may be recorded as revenue but is usually not under Treasury control until it reaches the TSA.
  - In absence of other information, forecasters can adjust monthly flows by the average time lag between receipt by banks and transfer to the TSA.
- End-of-month timing effects: If a large flow is expected at month-end (e.g., a corporate tax payment) and month-end falls on a weekend, the payment may be pushed into the following month. Adjust by identifying weekend impacts in past profiles.

### Expenditure flows — potential adjustments
- Potentially important: Payments for expenditure commitments outstanding at the end of year may be made in the following year; practice depends on annuality/accounting rules and complementary payment periods. Forecasters should explore previous years and reflect impacts in cash forecasts.
- Potentially important: Unspent provision rolled over to the following year may not always be reflected in the new year’s budget. Decisions to pay off arrears may not be part of the budget and can require large adjustments to the cash forecast once a repayment program is agreed.
  - Payment of arrears should be distinguished from payments during the complementary period; establish amounts for forecasting and control reasons.
  - If an agreed arrears payment program is included in the budget, include it in the initial forecast.
- Recognition of payments: Example — budget records expenditure on issuance of checks, but cash only moves when presented and cleared. Central or commercial bank data can supply averages for adjustments.
- Contractual deferrals: Some contracts defer payments beyond establishment of obligation; salaries may be recorded as paid but actually deferred in cash-strapped environments.
- Extra payments: Example allowance for a “13th month” salary bonus in December.
- Linked flows: Some grants/transfers may be linked to flows in a previous period (e.g., sub-national governments receiving a share of tax receipts only after a lag). Forecasters can use the government-controlled lag as an assumption.
- Donor timing: There may be a lag between donor-financed expenditures and receipt of funds; projects may be financed in advance by the government even without budget provision. If donor funds are paid in advance, they usually cannot be used by the government except drawn against project expenditure.
- Imprests and advances: Cash advanced to facilitate expenditure (e.g., imprests, advances to overseas missions) can bring forward cash flows relative to budget provision.
- Netting and non-cash treatments: Some revenue/expenditure flows may be netted (e.g., a subsidy payment netted off a tax liability) and thus have no net cash flow; check internal budgetary consistency. Dividend payments from an SoE may be netted against a promised grant or capitalization payment.

### Banking and financing flows — coverage and accounting issues
- The TSA is defined as those bank accounts whose balances can be controlled or accessed by the government; in some circumstances short-term control/access may be constrained.
- Coverage divergences:
  - The TSA coverage may be less than, equal to, or larger than the budget coverage (e.g., if TSA includes balances of an EBF). If the TSA includes an EBF, the EBF will need to be included within the forecast processes.
  - Leakage to commercial banks can be significant.
- Pre-funding (“seed money”) of banks may be required if core banking systems are not fully integrated; typically stable month-to-month and may be ignored for monthly forecasts (within-month variation more important for weekly forecasts).
- Accounting treatment of securities may diverge from cash flows (e.g., amortization of bond discounts affecting recorded interest; issue discounts or premiums affecting cash proceeds). Such adjustments are rare but should be identified with accountants or statisticians.

Box: Constraints accessing cash balances within the TSA — common types and handling
- Types of constraints:
  - Earmarked flows to different bank accounts; legally protected balances may be effectively outside the TSA (EBF), though periodic or arms-length transactions with the TSA may be possible.
  - Dispersed accounts in peripheral areas when banking sector integration is incomplete — only net flows reach the TSA with delay.
  - Legacy sub-accounts requiring Treasury intervention to move cash between accounts.
  - Accounts accessible to the Treasury but held outside the TSA at the central bank (examples exist in some Francophone countries); Treasury intervention required to move cash into the TSA.
- Interim workarounds:
  - For low-transaction, debt-related accounts, amend the spreadsheet to aggregate and monitor transactions and balances, giving early warning of required transfers.
  - If many separate accounts exist, rely more on monitoring bank statements from the central bank or commercial banks to trigger transfers as required.
  - Central banks have sometimes considered total net balances across government accounts, ignoring technical negative balances, to aid liquidity management.
- Reporting: Government should receive regular reports on balances in government bank accounts outside the TSA (central bank and ideally commercial banks), including EBF accounts, donor project accounts and MDA-controlled accounts. Such monitoring and control procedures lie outside forecasts but flows that can be transferred into the TSA should be incorporated in forecasts.

### Treatment and placement of adjustments in the tool
- Adjustments are made in the “Adjusted Budget Numbers” or the “Outturns and Adjustments” sections of the illustrative tool.
- Where adjustments are specifically identified and can be monitored during the year, place them against the respective revenue or expenditure flow.
- Where adjustments are forecasters’ judgments (somewhat arbitrary, drawing on experience), they can be placed below the surplus/deficit line, along with adjustments for differences between TSA coverage and total government cash balances.

### End of the budget year — procedures and forward-looking practice
- The forecaster should always look some months ahead to advise on necessary action, even if the horizon extends into the following budget year.
- When setting up for the new budget year:
  - Save the file as a new version and clear data for the old budget year.
  - Adjust data in the Budget Outturns section so year T-1 becomes year T-2, etc.; copy outturn data for the nearly completed budget year to become historical data for year T-1.
- Forecasts should not be confined to months/weeks of the current budget year; e.g., a forecast prepared in November may cover December to February and must add forecasts for January and February, particularly in the Policy Options section of the Main worksheet.
- Practical steps for inserting short-term columns (monthly or weekly):
  - Insert columns after the December column (one column for a forecast November to January; two columns for a December to February forecast), but before the ‘total’ column.
  - Copy into those columns, as a default, the outturns for January and February of the current year. Copy only values of individual entries (not associated formulae); formulae for sub-totals can remain unchanged.
  - Link the TSA balance so the opening balance for January equals closing balance for December.
  - Adjust copied data as necessary (options: common multiplier related to GDP growth or inflation; copy across data from plans/forecasts for the new year; insert other estimates).
  - Complete rows showing estimated impact of sensitivities and policy options for the forecast months.
  - Amend charts to accommodate extra columns via normal chart series editing.

### Detailed flows and previous year outturns
- General approach for monthly detailed flows:
  - Identify monthly expenditures over the previous three years.
  - Calculate the percentage falling in each month for each year.
  - For each month, take an average of the three percentages applying to that month.
  - Apply each averaged percentage to the budget figures for the current budget year.
- Use this widely used international approach with care; some adjustments may be needed (see previous examples and Box 4 in the Guidance Note).

*Source: eng-cash-flow-forecasting-in-fragile-and-low-income-states-guidance-note — Annex B.*

### Box 4: Possible Adjustments to Outturn Data

### Box 4: Possible Adjustments to Outturn Data

### Types of adjustments and when to apply them
- Outturns recorded in budget accounting rather than cash terms require adjustments (either directly or in the Adjusted Budget Numbers section).
- Identify and analyze substantial differences in monthly profiles across years (often due to major flows falling in different months, or no discernible pattern).
- Overwrite outlier historical data with an estimate of the underlying “normal” flow when one-off issues (e.g., intensified civil conflict, natural disasters, election expenses, responses to Covid-19) affect individual series and do not affect too many lines.
- Ignore historical profiles when wide-ranging changes in procedures or data classification (e.g., tax collection reforms) render earlier profiles unhelpful.

### Use of reported profiles and missing historical data
- Future monthly profiles may be available from the BD or from MDAs; such profiles can be unreliable (e.g., simple divide-by-12 approaches). Profiles from RAs may be more useful but likely cover only the largest taxes.
- Forecasters must judge whether to use reported profiles or the averaging method, and which adjustments to apply.
- If no historical data exist (e.g., first budget after conflict), construct a prior year budget outturn and enter it; in practice this may match the Target Year Budget and later be overwritten by actual outturns for the target year.
- If nothing is entered in Budget Outturns or the Target Year Budget, the tool assumes all revenue and expenditure flows are split evenly across the 12 months.

### Consistency when current-year outturns become prior-year data
- When current-year outturns become year T-1 outturns, ensure consistency with earlier outturn years; current-year data may be on a cash basis while T-2 and T-3 use budget classification including non-cash items.
- Possible adjustments may be needed as current-year data migrate to T-1 (adjustments can be made in the Adjusted Budget Numbers section but that may be more complicated).

### Using multiple years and weighting historical outturns
- With greater fiscal stability, reliance on outturn data stretching back more than three years increases.
- The spreadsheet can use the last 5 years of data to create average profiles; the user can select the number of years in the Budget outturns adjustments section (ranging from 1 to 5 years).
- When data quality varies over time, apply differential weights to past years. Example weightings:
  - Three-year weights example: 0.45, 0.35, 0.20.
  - Five-year weights example: 0.35, 0.30, 0.20, 0.10, 0.05.
- The spreadsheet defaults to equal weights determined by the user-input number of years; users can input other weights provided they add to one.
- Note: “The responses to Covid-19 might suggest that 2020 or 2021 should always be given a lower weight.”

### In-year outturns, weekly/monthly updates, and reconciling TSA balances
- Monthly or weekly outturns must be entered by hand unless linked to IFMIS; they must be in cash (or adjusted).
- Central bank statements are an additional source; actual TSA balance from the bank statement may differ from the calculated balance. If the gap is large, roll it into future forecasts and account for it when framing policy options.
- If outturns diverge from forecasts, adjustments to future plans are usually needed and may be substantial if the budget is revised.

### Handling incomplete or absent outturn data and alternative sources
- Outturns & Adjustments section requires manual input; if outturn data are from other departments, clarify whether they are cash or budget accounting and adjust as necessary.
- Where data are unavailable in required detail, make assumptions about allocation to rows or temporarily record unspecified flows in “other” categories; blank rows must be completed by the user.
- Bank statements can provide early transaction information (especially if transactions are centrally managed); if detailed coding is lacking, focus on larger transactions and net out transfers between sub-accounts to identify major tax receipts, debt-related flows, SNG transfers, and personnel-related expenses.
- The worksheet Template Bank Account-GFS Bridge is provided to allocate bank statement data lacking full classifications; it is not linked to other worksheets and is for forecaster judgement.
- If payments are dispersed across local banks or based on credit limits, obtain local bank statements or agree on generating daily reports of required data.

### Common occurrences illustrated in example outturns
- Domestic interest rates slightly higher than plan.
- External debt repayments slightly different from forecasts due to FX changes.
- Shortfalls in auction proceeds from domestic issuance.
- Government borrowed from the central bank in March to avoid an unexpected cash drain; assumed repayment before the end of the financial year with allowance in December (can be overwritten if repayment month differs).

### Recording adjustments and revised plans
- Include all changes and suggested adjustments in the Outturns & Adjustments section; the section also calculates divergences.
- When adjusting future months or weeks, enter the fully adjusted monthly forecasts (not just changes to previous forecasts); new rows will be picked up in the Revised Plans section.
- More fundamental adjustments to future years may follow revised budgets, changes in macro assumptions, a new government, new legislation, or a supplementary budget; these belong in Adjusted Budget Numbers and Revised Plans and may substantially overwrite previous plans.
- The BD is the main source for revised annual budgets; MDAs/RAs may need to reprofile remaining months.

### Inputs from MDAs/RAs and focus on major flows
- Develop contacts in key MDAs/RAs to obtain real-time information; require RAs and larger MDAs to submit forecasts at least monthly for, say, three months ahead.
- Forecasters should focus on major flows and aggregate inputs (often in a separate worksheet), making assumptions for the remainder (flat, simple pattern, or in line with plans).

### Identifying major expenditure flows (Box 5 summary referenced)
- Initially request information from budgetary units responsible for about 75-80 percent of government expenditure (often 5-7 MDAs).
- Prioritize:
  - Series with substantial nominal volatility or seasonality (e.g., agricultural support, transfers to local government).
  - MDAs with a small number of large capital projects (focus on those projects; other parts of the MDA budget can be residual).
  - Project Management Offices (PMOs) or Implementation Units as sources for key project information.
  - MDAs with large capital budgets; MPI may have an overview of capital spending.
- Recommended practice: require RAs/MDAs to complete monthly Treasury-supplied templates and submit by a fixed timetable (e.g., before the end of the second week of the month) with forecasts for the remaining weeks of the month and the three subsequent months (i.e., ideally 13 weeks).

*From Cash Flow Forecasting in Fragile and Low-Income States: Guidance Note (Box 4).*

### Box 6: Revenue and Expenditure Templates

### Box 6: Revenue and Expenditure Templates

### Revenue template
- Forecasts cover the three main tax groups; weekly forecasts are preferred but a monthly forecast is also shown as a consistency check.
- Comparisons with the initial monthly and weekly plans are included to help identify trends such as a persistent shortfall from the budget.
- The exercise is framed as identifying what will happen, not as a tool to hold MDAs/RAs to account for performance.
- A chart is optional but recommended to provide a visual check.

### Expenditure templates
- Two templates: one for current expenditure and one for capital expenditure; their design is essentially similar to the revenue template.
- Individual MDA contributions to total expenditure are typically small relative to RAs’ contribution to revenue; this may require submitting data to an extra place of decimals.
- Contributions from selected large MDAs must be aggregated, and allowance made for the proportion of total expenditure they cover.

### Weekly forecasts — construction techniques
- Weekly forecasts will likely require a combination of techniques because some flows are known, some have regular patterns, and others require assumptions.
- Practical allocation rules:
  - Flows with known precise dates (e.g., debt servicing payments) can be allocated to the relevant week.
  - Known due dates for many tax receipts and expected payment dates for some expenditures (e.g., salaries, transfers to SNG) can be allocated to the relevant week.
  - Some flows have a regular pattern across a month and can be modeled accordingly.
  - Other flows may be spread equally across the weeks of a month; if only monthly forecasts exist, a cautious assumption is that revenue is received at the end of the month and payments are made at the beginning of the month until forecasters gain experience of in‑month profiles.
- The illustrative spreadsheet builds weekly data from monthly data by spreading the monthly figure equally over the 4 or 5 weeks identified for that month; forecasters should overwrite these numbers to reflect lumpy transactions or overlaps between weeks and months.
- Emphasis should shift toward incorporating more bottom‑up information from MDAs/RAs for weekly forecasting, particularly where good weekly outturn data are not available.
- Weekly forecasts should be broadly consistent with monthly equivalents until weekly forecasts are built independently; forecasters should monitor cumulative weekly totals against cumulative monthly totals.
- Weekly (and monthly) outturn data should be entered directly into the Outturns & Adjustments section and recorded in supporting data worksheets for performance monitoring.

### Weekly calendar issues and preferred approach
- A month comprising fractions of a week can be handled by:
  - Identifying months with either 5- or 4-week periods (technique usually preferred and adopted here), or
  - Keeping strictly to the monthly calendar with stub periods of 1–3 days.
- Care is needed because a “month” may have 4 or 5 weeks; lumpy flows may fall in a week allocated to an adjacent month.
- Flows due on a certain date will not materialize on that date if it falls on a weekend; more such adjustments are needed than for monthly projections.
- Example implications: a corporate tax receipt deemed to arrive at the end of June may fall in a week regarded as part of July.

### Policy Options section — structure and purpose
- Designed to be the submission that forecasters might present to the Cash Coordinating Committee (CCC) or similar decision‑making body; two versions exist depending on whether forecasts are monthly or weekly.
- The section comprises three segments:
  - The central forecast: includes recent outturn figures and projections of future months; highlights TSA projections and separates discretionary and non‑discretionary expenditures (example treats personnel expenses and debt interest as non‑discretionary).
  - Sensitivities/what‑ifs: opportunity to show impacts if risks materialize (examples include called guarantees or poor auction performance); intended to give senior management an understanding of the range of potential outcomes and attached probabilities.
  - Policy responses: possible responses to emerging cash positions (examples are illustrative and can be overwritten); changes shown may be temporary (e.g., delay in a program loan implies less cash in one month but made up subsequently).
- The Policy Options table picks up data from the Revised Plans section; some series are subtotaled to avoid oversized tables and to focus decision makers on the TSA or cash over which government has control.
- Emphasis for policy action is likely on the next 3 months; the next three months plus the current month are highlighted to reflect stronger status of those entries.

### TSA projections chart, buffer, and reconciliation treatment
- The bottom‑right chart shows projections of the TSA under:
  - the central base,
  - the “worst” case (intended to be plausible not extreme),
  - the impact of policy options.
- The example includes an implicit cash balance/ cash buffer target of 25 million; action is identified to prevent the balance going below this level over the next few months.
- The Outturns & Adjustments section includes a reconciliation row to capture the gap between top‑down calculated TSA changes and the actual TSA balance; the Policy Options section adds an additional row for this difference and treats it as a risk factor.
- Treatment of a large discrepancy/drain:
  - If large and the drain on the TSA is greater than expected, it is prudent to take the discrepancy into account in future forecasts.
  - Unless a good reason exists to expect reversal, the drain should be carried forward and potentially an additional allowance added for further drain depending on judgment about likely cause.
  - The drain represents a change in the cash balance; it is ignored when cumulating the flow effects of other risks and sensitivities, but added back when calculating the total impact on the TSA.
  - If the drain is expected to worsen, a higher allowance should be included in future months.

### Practical notes on templates and use
- Three templates are included as worksheets in the main spreadsheet for convenience; they are not linked to other worksheets and forecasters will typically copy them into another spreadsheet and make design changes before sending to MDAs/RAs.
- The Policy Options chart can be used dynamically in CCC meetings to show the impact of discussed policy options.
- The spreadsheet is not designed to monitor budget execution performance; the focus is on forecasting cash and informing short‑term policy decisions.

*Source: IMF Guidance Note "Cash Flow Forecasting in Fragile and Low‑Income States" — Box 6: Revenue and Expenditure Templates.*

### ANNEX B: ANALYZING FORECAST ERRORS

### ANNEX B: ANALYZING FORECAST ERRORS

### Purpose of forecast-error analysis
- Two purposes: planning responses to forecast deviations; and learning lessons to apply in future forecasts.
- First question when an error arises: will the divergence be sustained or correct itself; advice may be needed from MDAs/RAs.

### Decisions to make before forensic analysis
- How far to drill down:
  - Every series is made up of other series with specific characteristics; large forecast errors must be identified and they will not always lie with the larger series.
- Relevant period to analyse:
  - Analysis often starts with the forecast for each month as made at the start of that month because that is how data are stored.
  - It is relevant to consider forecasts over shorter periods and forecasts made for the same period at different points (e.g., forecast made at the start of the month vs. 3 months previously vs. start of the year).
- Identify constrained variables:
  - At times of cash rationing, forecast error for a low-priority expenditure line may say little about forecast performance for that line.
  - Cash rationing can constrain expenditure to revenue; the forecast error of net cash flow may be misleadingly small even if revenue and expenditure errors are large.
- Identify policy variables:
  - If the main policy response is to change planned issuance of securities, divergence may reflect policy choice rather than an error in securities issuance forecast.
  - It is often sensible to take domestic securities issuance out of the cash flows being studied.

### Independence of errors and relationships across series
- Identify whether errors are independent across time and across components of inflows and outflows.
- Explore independence over time (e.g., whether an error in one month is offset in the next month) and across components (e.g., whether an error in commodity-related tax revenues is positively or negatively associated with an error in expenditure on goods and services).
- Some understanding of direction of causality is needed to interpret associations.

### Timing changes and serial correlation
- Timing changes give rise to serial correlation: errors over time are not independent and errors in one period may be offset in subsequent periods.
- Many countries experience the error over a year being much less than the errors for individual months.
- Formal statistical tests exist for serial correlation, but simple plotting of the error in any one period against the error in the previous period can reveal clustering suggestive of serial correlation.
- Even if monthly forecast error for net cash flows is acceptable, there may be substantial within-month fluctuations that require a higher cash buffer.

### Distribution of forecast errors and one-off events
- Forecast errors are unlikely to be normally distributed due to serial correlation and a large number of one-off or unexpected events affecting revenue and expenditure each year.
- It is often impractical to fully “clean” data series.
- Analysts should identify factors that triggered large errors and assess risks of repetition.

### Practical implications and examples from the Guidance Note
- Revenue profiling and adjustments:
  - Corporate income tax receipts can be affected by whether end-of-month due dates fall on weekends.
  - Dividends show no regular pattern and require direct information from SoEs or the central bank.
  - Forecasters may apply historical average monthly profiles to annual budget totals or adjust profiles based on domain knowledge and updated information; the forecast must not be constrained to the budget if new information is available.
  - VAT refunds, if significant and uncertain in timing, should be treated separately from gross VAT receipts.
- Expenditure adjustments and example figures:
  - Payments outstanding from the previous year of DX 3 million are assumed to be paid DX 2 million in January and DX I million in February.
  - An estimated DX 1 million of budgeted payments are likely to be outstanding at the end of the target year.
  - SNGs are assumed to be entitled to 25 percent of all tax receipts collected in the previous month; forecasts for SNG grants are aligned with direct and indirect tax forecasts.
- Capital expenditure and project-related flows:
  - Local counterpart funds are assumed, in the absence of other information, to be a constant proportion of externally-financed expenditure; example proportion shown is 8 percent.
  - In the absence of specific timing information, externally-financed projects may be spread across the year and slightly back-end loaded, with adjustments for expected under-execution.
  - No allowance is made for fixed asset sales unless plans are clear.
- Financing and debt mechanics:
  - Deficit financing can include donor grants, external loans and credits, domestic issuance of T-bills or T-bonds, or reduction in cash balances.
  - Interest and debt repayment lines are mostly determined by past borrowing and should be calculated accordingly.
  - All T-bills issued must be entered into defined categories; changes in issuance (e.g., additional 3-month T-bills) alter T-bill repayment projections for the months ahead.
  - Example simplification: proceeds from issuance might be 97 this month, with a redemption of 100 in 3 months’ time.
  - Borrowing from the central bank is assumed in the initial plan to be zero; advances/overdrafts should normally be repaid within the same year and entered with a negative sign in the same ‘new borrowing’ line.
  - Project-related donor flows are assumed tightly linked to timing of work on projects; the relevant estimates are equal to the adjusted line for investment on externally financed projects (row 37 in the Main worksheet), but donor flows might lag expenditure unless pre-financing is agreed.

*Source: ANNEX B of "Cash Flow Forecasting in Fragile and Low-Income States: Guidance Note" (IMF).*

### ANNEX D: THE BRIDGE TABLE

### ANNEX D: THE BRIDGE TABLE

### Overview
- The example of the possible use of the bridge table included in the worksheet “Bank Account-GFS Bridge” is intended only as an illustration. The guidance note states there are many different ways the table might be used; different users will find different solutions.
- The worksheet example assumes the availability of a central bank statement in Excel with sufficient descriptive or coded information to permit identification of the majority of transactions.

### Use in the worksheet
- The same table is used for each of the 5 days of the week. The entries in the column for each day show the flows identified from the bank statement.
- The sum of identified flows over the week is compared with an “average” week of the year (i.e. simply dividing the annual totals in the Monthly Summary section by 52).
- A column is added for notes – as an aide memoire for the user to record why flows may be very different from the average, or different than forecast.
- The user is expected to sub-total flows by plan/forecast categories on the statement (e.g. by using the filter tool) and transfer sub-totals to the bridge table. If that is not possible, individual flows may need to be transferred and then summed, which is more time consuming.

### Key assumptions and practical considerations
- There is unlikely to be fully detailed information in the bank statement to allow all flows to be identified.
- Some entries do not reflect cash flows into and out of the TSA and must be netted out (e.g. transfers between accounts within the TSA).
- Users will likely focus on identifying all large flows and include smaller flows in a residual category, balancing time spent against the benefit of further disaggregation.
- Adjustments in the forecast for non-cash flows or other TSA-related flows are mostly not included in the bank statement or, where they are, are negligible or cannot be specifically identified.
- An unidentified category will typically reconcile identified flows with the balance on the statement; this category is likely to shrink with experience.
- A coding that links with the IFMIS codes is recommended as likely the best approach.
- Users should work closely with the central bank and the RAs to ensure the statement provides the necessary detail and is manipulable.

### Assumptions used in the worksheet example (explicit)
- Although most income tax flows can be identified, there is a large residual, likely related to personal income tax rather than corporate income tax, included in an “unspecified” line.
- The same department of the revenue authority is collecting sales tax and import duties; while the department is identified on the statement, individual taxes are not and have been combined in one line.
- Salary payments are identified, but most other current expenditures (apart from debt interest) and many smaller capital expenditures are not categorisable on the statement; these have been allocated to goods and services.
- Financing transactions, many individually large, are individually identified.

### Examples drawn from central bank statements (illustrative)
- Example 1 (sub-Saharan English-speaking country): statement descriptions (with identifying details removed) are mapped to bridge-table classifications such as:
  - “PAYMENT FOR xxx AND RETENTION FEE xxx” → Misc personal income tax
  - “xxx COLLECTIONS xxx GST” → Sales etc tax
  - “xxxTBONDxxx” → Debt repayment
  - “INT ON 364 DAYS TB ISD xxx” → Debt Interest
  - “CASH TRF” → Transfer with TSA - ignore
  - “USDxxx IFO xx” → Project related loan
  - “GOSL CHQ NO XXX” → Domestic capex [as identified]
  - “xxx-GST REFUND xxx” → Sales etc tax (refund, negative)
  - Several items map to “Misc goods and services,” “Corporation tax,” “Misc personal income tax,” etc.
- Example 2 (sub-Saharan French-speaking country): entries are grouped under headings and each line is categorised as definitive or provisional (i.e. subject to further investigation if the size warrants it). Headings and mappings include:
  - Tax Receipts → Other tax revenues (multiple incoming-credit entries)
  - Customs Receipts → Customs duties, excluding oil
  - Treasury Receipts → Other non-tax revenue
  - Treasury Payroll → Reduction of payment arrears; Staff costs
  - Treasury Expenses → Transfer expenses; Debt repayment
  - Treasury Deposits → Deposit of third parties
  - Classification status: “Provisional classification pending further information” or “Definitive classification”

### Practical recommendations and workflow guidance
- Ensure the central bank statement is available in Excel and contains sufficient descriptive or coded transaction information.
- Sub-total large identifiable flows on the statement and transfer sub-totals to the bridge table; treat smaller unidentified flows as residuals.
- Net out non-cash or intra-TSA items (e.g. transfers between accounts within the TSA).
- Use the weekly comparison against an “average” week (annual totals divided by 52) as a diagnostic, acknowledging any single week may differ substantially from the average.
- Coordinate closely with the central bank and RAs; consider linking statement coding to IFMIS codes.
- Treat provisional classifications as subject to follow-up investigation if amounts are material.

---


_Source: https://www.imf.org/-/media/files/capacity-developement/home/eng-cash-flow-forecasting-in-fragile-and-low-income-states-guidance-note.pdf_