Columns, presets, and units

The tossd-reader package processes 53 columns from each official TOSSD activity-level file, standardised into snake_case names and typed according to schema.csv. The columns parameter of get_tossd accepts a preset name ("minimal", "analysis", or "all") or a list of specific snake_case column names. Five columns always appear in query results regardless of selection, exposed as the FORCED_COLUMNS tuple (year, tossd_pillar, tossd_subpillar, is_aggregate, and unit). The is_aggregate flag (provider_code == 0) and unit indicator are derived columns added by the package, bringing the "all" preset to 55 total columns.

Presets

Preset	Total Columns	Analytical Scope
`minimal`	19	Core identifiers, provider and recipient names, pillar categories, and financial amounts.
`analysis`	44	Adds sector classifications, channel codes, raw SDG tags, keyword fields, and financing modalities.
`all`	55	All packaged schema columns, derived fields, and any unmodelled publisher passthrough columns.

Memory footprint

Preset (`columns=`)	Columns	2024 (Single Year)	2019 to 2024 (All Six Years)
`"minimal"`	19	55 MB	278 MB
`"analysis"`	44	102 MB	505 MB
`"all"` (default)	55	377 MB	2.1 GB

Memory figures represent pandas memory_usage(deep=True) across the official dataset.

Schema columns

Columns are listed in source file order from schema.csv.

Column	Dtype	Minimal	Analysis	Description
`year`	`Int16`	Yes	Yes	Reporting year (2019 to 2024).
`provider_code`	`Int16`	Yes	Yes	Numeric provider code. Code `0` indicates an aggregate summary row.
`provider_name`	`category`	Yes	Yes	Standard English name of the provider country or multilateral institution.
`provider_agency_name`	`category`			Extending agency or department within the provider.
`tossd_id`	`string`	Yes	Yes	Unique TOSSD activity identifier.
`project_number`	`string`	Yes	Yes	Provider internal project or commitment identifier.
`recipient_code`	`Int16`	Yes	Yes	Numeric recipient country or regional code.
`recipient_name`	`category`	Yes	Yes	Standard English name of the recipient country, territory, or region.
`region_name`	`category`		Yes	Geographic region of the recipient.
`channel_raw_text`	`string`			Unstructured channel text as reported by the provider.
`channel_code`	`Int32`		Yes	Numeric code for the implementing channel organisation.
`parent_channel_code`	`Int32`			Numeric code for the parent organisation of the implementing channel.
`parent_channel_name`	`string`			Decoded English name of the parent channel organisation from the channel codelist.
`other_partners`	`string`			Names of co-financing entities or implementing partners.
`channel_name`	`category`		Yes	Standard category of the implementing channel.
`finance_instrument_code`	`Int16`		Yes	Numeric code for the financial instrument (grants, debt, equity).
`finance_instrument_name`	`category`		Yes	Name of the financial instrument.
`financing_arrangement_code`	`category`		Yes	Code for the financing arrangement (blended finance, Islamic finance, recipient-counterpart co-financing, officially supported export credits, or SDR transactions).
`financing_arrangement_name`	`category`		Yes	Name of the financing arrangement.
`framework_of_collaboration_code`	`category`		Yes	Code for South-South or triangular collaboration frameworks.
`framework_of_collaboration_name`	`category`		Yes	Name of the collaboration framework.
`modality_code`	`category`		Yes	Code for the aid modality (e.g. project aid, core contributions).
`modality_name`	`category`		Yes	Name of the aid modality.
`project_title`	`string`			Short title of the activity.
`external_link`	`string`			URL pointing to external project documentation.
`sdg_codes_raw`	`string`		Yes	Semicolon-separated SDG goals and targets tagged for the activity.
`sdg_goal_level_explanation`	`string`			Narrative rationale for the assigned SDG classifications.
`keywords_raw`	`string`		Yes	Pipe-separated thematic keywords (e.g. climate, biodiversity, gender).
`purpose_code`	`Int32`		Yes	5-digit DAC purpose code identifying the sector sub-discipline.
`purpose_name`	`category`		Yes	Description of the 5-digit DAC purpose code.
`sector_name`	`category`		Yes	Broad sector classification name (3-digit DAC sector).
`sector_code`	`Int16`		Yes	3-digit DAC sector code.
`isic_code`	`category`		Yes	International Standard Industrial Classification (ISIC) industry code.
`isic_description`	`category`		Yes	Description of the ISIC industry code.
`isic_section_letter`	`category`		Yes	High-level ISIC section letter (e.g. A for Agriculture).
`isic_section_description`	`category`		Yes	Description of the ISIC section.
`project_description`	`string`			Narrative description of project objectives and activities.
`tossd_pillar`	`Int8`	Yes	Yes	Pillar classification (`1` for Pillar I, `2` for Pillar II, `0` for legacy publisher placeholder rows).
`tossd_subpillar`	`category`	Yes	Yes	Sub-pillar category. `NA` unless the row carries a `"21"` (Pillar II.A) or `"22"` (Pillar II.B) tag.
`usd_commitment`*	`float64`	Yes	Yes	Total gross financial commitment in current USD thousands.
`usd_commitment_deflated`*	`float64`	Yes	Yes	Total gross financial commitment in constant 2024 USD thousands.
`usd_disbursement`*	`float64`	Yes	Yes	Gross financial disbursement in current USD thousands.
`usd_disbursement_deflated`*	`float64`	Yes	Yes	Gross financial disbursement in constant 2024 USD thousands.
`usd_reflow`*	`float64`	Yes	Yes	Financial reflows and repayments in current USD thousands.
`usd_reflow_deflated`*	`float64`	Yes	Yes	Financial reflows and repayments in constant 2024 USD thousands.
`salary_cost`	`float64`			Administrative salary costs reported for technical cooperation.
`concessionality_flag`	`Int8`		Yes	Concessionality status (`1` for concessional, `0` for non-concessional).
`maturity`	`Int16`		Yes	Maturity duration for debt instruments as reported by the provider.
`mobilisation_instrument`	`category`		Yes	Instrument used to mobilise private commercial finance.
`usd_amount_mobilised`*	`float64`	Yes	Yes	Private capital mobilised by official interventions in current USD thousands.
`usd_amount_mobilised_deflated`*	`float64`	Yes	Yes	Private capital mobilised in constant 2024 USD thousands.
`mobilisation_origin`	`string`			Geographic origin of mobilised private investment.
`source_name`	`category`		Yes	Data source identifier for the reported activity.

* Reported in USD thousands as published. units="usd_million" divides these 8 columns by 1,000. units="usd" multiplies them by 1,000.

Amount columns

Column	2024 Non-Null Rows	Analytical Scope
`usd_commitment`	390,190	Official commitments in current prices.
`usd_commitment_deflated`	390,190	Official commitments adjusted to 2024 constant prices.
`usd_disbursement`	441,645	Gross disbursements in current prices.
`usd_disbursement_deflated`	441,645	Gross disbursements adjusted to 2024 constant prices.
`usd_reflow`	215,264	Loan repayments and capital returns in current prices.
`usd_reflow_deflated`	215,264	Loan repayments adjusted to 2024 constant prices.
`usd_amount_mobilised`	1,693	Private finance mobilised in current prices.
`usd_amount_mobilised_deflated`	1,693	Private finance mobilised adjusted to 2024 constant prices.

Each current price column has the identical non-null count as its corresponding deflated column. For conceptual differences between commitments, disbursements, reflows, and mobilised amounts, see The amount columns.

Always present columns

FORCED_COLUMNS names the five columns present in every query output, regardless of columns= requested.

from tossd_reader import FORCED_COLUMNS

FORCED_COLUMNS

('year', 'tossd_pillar', 'tossd_subpillar', 'is_aggregate', 'unit')

Column	Dtype	Analytical Role
`year`	`Int16`	Reporting year (2019 to 2024).
`tossd_pillar`	`Int8`	Pillar classification (`1` for Pillar I cross-border flows, `2` for Pillar II global public goods, `0` for publisher placeholder rows).
`tossd_subpillar`	`category`	Sub-pillar category. `NA` unless the row carries a `"21"` (Pillar II.A) or `"22"` (Pillar II.B) tag. Categories are exactly `{"21", "22"}`.
`is_aggregate`	`bool`	True for summary records (`provider_code == 0`) and False for individual activity records.
`unit`	`category`	Denomination unit (`"usd_thousand"`, `"usd_million"`, or `"usd"`).

tossd_subpillar is NA unless the row carries a real "21"/"22" tag. A Pillar I row, an untagged Pillar II row, and a Pillar 0 row all read NA.

import tossd_reader as tossd

df = tossd.get_tossd(years=2024, columns="minimal")
df["tossd_subpillar"].value_counts(dropna=False)

tossd_subpillar
NaN    319526
22      94754
21      59746

list(df["tossd_subpillar"].cat.categories)

['21', '22']

get_tossd_raw returns the published sentinel values unchanged. This NA contract applies only to get_tossd.

Data quality notes

Empty strings in published files convert to typed null values in get_tossd. get_tossd_raw preserves raw publisher string representations.
Modality code c01 standardises to uppercase C01. Published files contain mixed case across different reporting years.
The maturity column reflects values directly from the publisher without unit conversion.

Schema drift detection

During data loading, tossd_reader compares published columns against schema.csv. A missing expected column, an unconvertible data type, or duplicate normalised column names raises SchemaDriftError identifying the affected column and offending value.

Unmodelled publisher columns

Columns present in source files that do not appear in schema.csv emit a single warning per process and pass through under their original names when using columns="all".