The finding
SBA disaster loan statistics for fiscal years 2015 to 2022 come down to one ratio: for every dollar of home damage that SBA loss verifiers recorded, the agency approved about 48 cents in disaster home loans. Across 26,731 rows of published SBA Disaster Loan Data, verified home loss totals $31.3bn and approved home loan dollars total $14.9bn, or 47.7%. For businesses the ratio is 21.7%. The gap is not unpaid damage: SBA lends only for losses not covered by insurance or other sources.
Headline figures
In this report
- What one row of SBA data is
- Verified loss against approved loans
- Structure versus contents
- The disasters that dominate the record
- By state
- By fiscal year
- ZIP-level concentration
- Places with verified loss and no approved loan
- Business loans and economic injury
- The eight files behind the series
- Method
- Reproduce this
- What this does not show
- Questions this report answers
- Sources
Key findings
- SBA approved $14.9bn in disaster home loan dollars against $31.3bn of verified home loss, 47.7%. The difference of $16.4bn is loss that insurance, grants or the household covered, or that went uncovered; the data cannot say which.
- Contents (furniture, clothing, appliances, cars) make up 24.8% of verified home loss, $7.8bn. Approved contents dollars equal 45.1% of verified contents loss, against 48.5% for real estate.
- Fire disasters show the lowest approved share of the major hazard types: 23.5% of verified home loss, against 50.0% for floods (FEMA incident type joined to SBA data).
- The 10 largest SBA disasters hold 67.8% of all verified home loss. Hurricane Harvey alone accounts for $5.8bn, 18.4%.
- Just 175 of 9,633 ZIP codes with verified loss, 1.8%, hold half of all verified home loss in the SBA data.
- Only 1.0% of verified home loss sits in city and ZIP rows where SBA approved nothing at all. The gap between loss and lending opens inside places that did receive loans, not in places that were passed over.
- Business borrowers show a much wider gap: $5.3bn approved against $24.5bn verified, 21.7%. Economic injury (EIDL) working capital is 18.1% of approved business dollars.
- One published row, Houma, Louisiana (Ida, FY2021), reports $2.4bn of verified contents loss, 12 times the next largest contents value in the file. It is excluded here as a probable entry error, and the SBA data itself warns of such errors.
What one row of SBA data is
Read this before any figure. The SBA Disaster Loan Data files do not list loans. Each row is an SBA aggregate for one disaster declaration and one damaged-property city and ZIP code. A row can stand for one household or for hundreds, and the files publish no count of applications, borrowers or loans. So this report never states an average loan or a typical borrower. Every figure is a sum of dollars, or a ratio of two sums.
SBA’s own instructions sheet, included in every workbook, sets the terms. It describes the dataset as containing “raw, unedited data from SBA’s Disaster Credit Management System (DCMS) which may have been entered directly by disaster survivors and as such is subject to human error”, adds that “the dollar values in the data set may not reflect subsequent changes to verified losses or approved loan amounts”, and states that the dataset “is not intended to be an official Federal report, and should not be considered as such.”
| Column | SBA’s definition (Data Menu sheet) |
|---|---|
| Total Verified Loss (Original) | Total $ amount of losses verified by SBA |
| Verified Loss Real Estate (Original) | $ amount of Real Estate losses verified by SBA |
| Verified Loss Content (Original), home | $ amount of the damaged property Content (personal property, auto, etc.) verified by SBA |
| Total Approved Loan Amount (Original) | Total $ amount of approved SBA disaster home loan |
| Approved Amount EIDL (Original), business | $ amount of approved Economic Injury (working capital) loan proceeds |
The same sheet adds that verified losses “are damage estimates only and are calculated by SBA loss verifiers (inspectors)”, and that loan amounts can rise if new damage is found or fall if the survivor later receives insurance or grant money that duplicates SBA’s assistance. Every value here is the original figure, not a final one.
Why approved dollars sit below verified loss: SBA states on its physical damage loans page that these loans “cover disaster losses not fully covered by insurance or other sources” and that insurance proceeds “may be deducted from the eligible loan amount”. Its disaster assistance page says SBA “will determine whether an applicant can obtain credit elsewhere” and sets terms “depending on ability to repay the loan”. A loan is debt, approved against credit and repayment tests, and declined or withdrawn applications still carry verified loss in these rows. The 26,731 home rows and 16,755 business rows are read with that in mind.
Verified loss against approved loans
| Line | Home loans | Business loans |
|---|---|---|
| Rows (city and ZIP aggregates) | 26,731 | 16,755 |
| Verified loss, real estate | $23.5bn | $22.1bn |
| Verified loss, contents | $7.8bn | $2.4bn |
| Verified loss, total | $31.3bn | $24.5bn |
| Approved, real estate | $11.4bn | $3.7bn |
| Approved, contents | $3.5bn | $0.6bn |
| Approved, economic injury (EIDL) | not applicable | $1.0bn |
| Approved, total | $14.9bn | $5.3bn |
| Approved as a share of verified loss | 47.7% | 21.7% |
Of $31.3bn in verified home loss, $16.4bn has no matching approved loan dollars. That figure is easy to misquote. It is not a shortfall SBA failed to pay: it is the part of verified damage that insurance settled, that grants covered, that applicants declined or withdrew from, or that SBA did not lend against on credit grounds. The data separates none of those. For the parallel picture on the grant side, see the disaster grant ledger, and for federal spending by state, FEMA spending by state.
Business lending runs at 0.46 times the home ratio: 21.7% against 47.7%. The files do not explain the difference. Businesses are more likely to hold commercial property insurance and face different credit tests, but neither factor is recorded in the SBA data.
Structure versus contents
For restoration firms this is the split that matters. SBA verifies home loss in two parts: real estate (the structure) and contents, which SBA’s Data Menu defines as “personal property, auto, etc.”. Contents make up 24.8% of verified home loss, $7.8bn of $31.3bn. Approved contents dollars equal 45.1% of verified contents loss, while approved real estate dollars equal 48.5% of verified structure loss.
| FEMA incident type | SBA disasters | Verified home loss | Approved home loans | Approved share | Contents share |
|---|---|---|---|---|---|
| Hurricane | 42 | $20,399m | $10,874m | 53.3% | 24.4% |
| Fire | 14 | $3,774m | $886m | 23.5% | 19.9% |
| Flood | 34 | $3,612m | $1,807m | 50.0% | 35.4% |
| No FEMA number | 378 | $1,122m | $354m | 31.6% | 18.0% |
| Severe Storm | 28 | $960m | $412m | 42.9% | 32.1% |
| Earthquake | 4 | $439m | $154m | 35.0% | 4.3% |
| Severe Ice Storm | 4 | $325m | $161m | 49.5% | 25.8% |
| Tornado | 10 | $323m | $121m | 37.7% | 23.3% |
| Typhoon | 3 | $151m | $91m | 60.2% | 19.7% |
Among hazard types with at least $1.0bn of verified home loss, flood disasters carry the highest contents share, 35.4%. Floods stand at 35.4% against 19.9% for fire. One plausible reading is that water ruins furniture, clothing and appliances in homes whose frame survives, while fire more often takes the structure itself; the data records the split, not the cause. For a contractor sizing pack-out and contents cleaning after a flood declaration, that is the relevant proportion. The depth side of the same story is in our water depth damage curve.
The approved share splits the hazards more sharply. The lowest among the large types is fire. Fire disasters show 23.5% of verified home loss matched by approved loans, against 50.0% for floods and 53.3% for hurricanes. That pattern fits the point made by FEMA FloodSmart: “Most homeowners insurance does not cover flood damage”, so more flood loss may be left for SBA to lend against. The SBA files record no insurance status, so this is a consistent reading, not a measured cause. The insured side is set out in the insured flood loss gap report.
The disasters that dominate the record
| SBA disaster | Event | State | Fiscal year | Verified home loss | Approved home loans | Approved share | Contents share |
|---|---|---|---|---|---|---|---|
| TX-00487 | Harvey | Texas | FY2017 | $5,753m | $2,901m | 50.4% | 28.6% |
| FL-00178 | Ian | Florida | FY2022 | $2,621m | $1,479m | 56.4% | 27.4% |
| PR-00031 | Maria | Puerto Rico | FY2017 | $2,514m | $1,261m | 50.2% | 20.5% |
| FL-00130 | Irma | Florida | FY2017 | $2,092m | $1,078m | 51.5% | 12.9% |
| LA-00115 | Ida | Louisiana | FY2021 | $1,970m | $1,161m | 58.9% | 24.0% |
| LA-00065 | Severe Storms and Flooding | Louisiana | FY2016 | $1,947m | $1,017m | 52.2% | 40.7% |
| CA-00295 | Wildfires | California | FY2019 | $1,478m | $412m | 27.9% | 19.8% |
| FL-00140 | Michael | Florida | FY2019 | $990m | $522m | 52.8% | 20.9% |
| LA-00103 | Laura | Louisiana | FY2020 | $983m | $564m | 57.3% | 22.8% |
| CA-00279 | Wildfires | California | FY2018 | $864m | $140m | 16.2% | 16.8% |
| NC-00099 | Florence | North Carolina | FY2018 | $603m | $333m | 55.2% | 23.7% |
| VI-00009 | Irma | U.S. Virgin Islands | FY2017 | $382m | $204m | 53.3% | 22.0% |
Hurricanes hold 65.2% of verified home loss in the SBA data. Hurricane Harvey alone holds 18.4% of verified home loss in the eight years, and the 10 largest SBA disasters together hold 67.8%. The billion-dollar events behind that concentration are catalogued in our billion-dollar disaster record.
Across the 112 SBA disasters with at least $10m of verified home loss, the median disaster has 46.1% of its verified loss matched by approved home loans. The fire declarations in the top 12, CA-00295 (27.9%) and CA-00279 (16.2%), sit well below that, the hazard pattern from the previous section seen at event level.
The join to FEMA works well. Of rows carrying a FEMA disaster number, 99.99% match a declaration in FEMA DisasterDeclarationsSummaries v2. 6,932 home rows carry no FEMA number at all; these appear to be SBA declarations made without a FEMA disaster, and they hold 3.3% of verified home loss. For 19,548 of 19,798 matched rows the file’s fiscal year equals the fiscal year of the FEMA declaration date, so the fiscal year here tracks when a disaster was declared, not when the loan was made.
By state
| State | Rows | Verified home loss | Approved home loans | Approved share | Contents share |
|---|---|---|---|---|---|
| Texas | 4,382 | $6,939m | $3,428m | 49.4% | 28.8% |
| Florida | 3,696 | $6,001m | $3,241m | 54.0% | 21.0% |
| Louisiana | 2,255 | $5,478m | $2,939m | 53.7% | 29.5% |
| California | 2,467 | $3,296m | $781m | 23.7% | 18.8% |
| Puerto Rico | 1,212 | $2,923m | $1,383m | 47.3% | 18.9% |
| North Carolina | 987 | $779m | $420m | 53.8% | 24.4% |
| New Jersey | 695 | $481m | $242m | 50.4% | 35.6% |
| New York | 863 | $445m | $221m | 49.7% | 37.1% |
| Michigan | 472 | $420m | $187m | 44.4% | 41.6% |
| South Carolina | 747 | $415m | $225m | 54.3% | 20.3% |
| Colorado | 98 | $379m | $103m | 27.2% | 24.5% |
| Alabama | 627 | $301m | $162m | 54.0% | 21.3% |
| Kentucky | 547 | $284m | $123m | 43.3% | 28.3% |
| Tennessee | 363 | $262m | $52m | 19.9% | 15.4% |
| Oregon | 297 | $200m | $60m | 29.8% | 27.9% |
Texas, Florida and Louisiana together hold 58.9% of verified home loss. Of the three, Louisiana has the highest contents share, 29.5%.
| State | Verified home loss | Approved share |
|---|---|---|
| Alaska | $110m | 69.6% |
| Wisconsin | $34m | 58.7% |
| Indiana | $28m | 57.1% |
| Iowa | $109m | 56.7% |
| Pennsylvania | $126m | 54.6% |
| Oregon | $200m | 29.8% |
| Hawaii | $121m | 29.3% |
| Colorado | $379m | 27.2% |
| California | $3,296m | 23.7% |
| Tennessee | $262m | 19.9% |
The approved share runs from 69.6% in Alaska to 19.9% in Tennessee. A state figure pools every declaration in eight years, so it reflects the hazard mix it happened to draw as much as anything about the state. No ranking of state administration follows from it.
By fiscal year
| Fiscal year | Rows | Verified home loss | Approved home loans | Approved share | Contents share | Largest SBA disaster |
|---|---|---|---|---|---|---|
| FY2015 | 1,239 | $523m | $201m | 38.5% | 24.7% | Severe Storms, Tornadoes, Straight-Line Winds and Flooding (Texas) |
| FY2016 | 2,481 | $3,112m | $1,518m | 48.8% | 35.5% | Severe Storms and Flooding (Louisiana) |
| FY2017 | 5,040 | $11,737m | $5,937m | 50.6% | 23.3% | Harvey (Texas) |
| FY2018 | 1,056 | $2,078m | $681m | 32.8% | 19.0% | Wildfires (California) |
| FY2019 | 2,465 | $3,587m | $1,409m | 39.3% | 18.5% | Wildfires (California) |
| FY2020 | 4,899 | $2,523m | $1,088m | 43.1% | 21.3% | Laura (Louisiana) |
| FY2021 | 6,480 | $4,071m | $2,201m | 54.1% | 29.3% | Ida (Louisiana) |
| FY2022 | 3,071 | $3,651m | $1,889m | 51.8% | 27.3% | Ian (Florida) |
FY2015 to FY2022 is not a trend line. The largest year, FY2017, holds 37.5% of the eight-year total because Harvey, Maria and Irma were all declared in it. A year’s figure is set by its largest storm, the same pattern our inflation-adjusted flood damage report finds in NFIP claims. The series ends at FY2022 (30 September 2022): no FY2023 or later file is listed on the SBA dataset page as of 6 October 2026.
ZIP-level concentration
Summing every home row by five-digit ZIP code across all eight years gives 9,633 ZIP codes with verified loss. The top 96, one percent of them, hold 39.3% of verified home loss, and 175 ZIP codes (1.8%) reach half. 20 rows without a valid five-digit ZIP are left out of this section only.
| ZIP | Place | State | Verified home loss | Approved home loans | Approved share | Contents share |
|---|---|---|---|---|---|---|
| 95969 | Paradise | California | $991m | $284m | 28.7% | 19.9% |
| 70726 | Denham Springs | Louisiana | $395m | $202m | 51.2% | 41.9% |
| 95403 | Santa Rosa | California | $374m | $64m | 17.3% | 18.4% |
| 95404 | Santa Rosa | California | $333m | $46m | 13.7% | 14.7% |
| 33908 | Fort Myers | Florida | $331m | $165m | 49.8% | 37.8% |
| 80027 | Louisville | Colorado | $310m | $91m | 29.3% | 25.1% |
| 00802 | St Thomas | U.S. Virgin Islands | $288m | $155m | 53.9% | 23.2% |
| 77450 | Katy | Texas | $260m | $156m | 60.2% | 33.1% |
| 77662 | Vidor | Texas | $226m | $114m | 50.6% | 29.4% |
| 32404 | Panama City | Florida | $214m | $110m | 51.5% | 23.0% |
For a restoration business the useful reading is that half of verified disaster home loss lands in 175 ZIP codes, not across whole states. A declared county is a wide net; the verified damage inside it is narrow.
Places with verified loss and no approved loan
4,223 home rows carry verified loss but no approved loan dollars. Together they hold $313m, 1.0% of verified home loss. The rest of the gap between loss and lending sits in rows where SBA approved at least some loan dollars.
Two other row types are worth knowing about. 6,744 home rows (25.2%) show neither verified loss nor an approved loan; the files do not say why. And 2,491 rows show approved dollars above verified loss. SBA’s physical damage loans page states that loans can be increased by “up to 20% of the verified loss” for mitigation and that SBA “can refinance all or part of a previous mortgage in some cases”. Either could lift a row above its verified loss, but neither is itemised in the data.
Business loans and economic injury
Business rows add one field: Approved Amount EIDL, the economic injury (working capital) part of a business disaster loan. Of $5.3bn in approved business loan dollars, $1.0bn (18.1%) is EIDL, $3.7bn is real estate and $0.6bn is contents. 793 business rows record EIDL approvals with no verified physical loss at all.
| Fiscal year | Business rows | Verified business loss | Approved business loans | Of which EIDL | Approved share |
|---|---|---|---|---|---|
| FY2015 | 530 | $324m | $82m | $11m | 25.3% |
| FY2016 | 1,336 | $2,736m | $553m | $77m | 20.2% |
| FY2017 | 4,076 | $13,337m | $2,864m | $495m | 21.5% |
| FY2018 | 825 | $1,431m | $249m | $71m | 17.4% |
| FY2019 | 1,337 | $3,570m | $816m | $98m | 22.9% |
| FY2020 | 2,912 | $569m | $134m | $38m | 23.6% |
| FY2021 | 3,503 | $1,139m | $250m | $44m | 22.0% |
| FY2022 | 2,236 | $1,405m | $378m | $131m | 26.9% |
Business contents (inventory, machinery, equipment and leasehold improvements, in SBA’s definition) are only 9.9% of verified business loss. The largest business totals are in Florida ($6,253m verified) and Texas ($6,076m).
The eight files behind the series
| File | Reporting period | Report run date (as printed) | Declaration types (as printed) |
|---|---|---|---|
| FY2015 | 10/01/2014 to 09/30/2015 | 01/11/2015 | PRES_IA, ADMINISTRATIVE |
| FY2016 | 10/01/2015 to 09/30/2016 | 11/8/2016 | PRES_IA, ADMINISTRATIVE |
| FY2017 | 10/01/2016 to 09/30/2017 | 4/10/2018 | ALL |
| FY2018 | 10/01/2017 to 09/30/2018 | 3/19/2019 | ALL |
| FY2019 | 10/01/2018 to 09/30/2019 | 4/10/2020 | PRES_IA, PRES_PA, ADMINISTRATIVE |
| FY2020 | 10/01/2019 to 09/30/2020 | 7/19/2022 | PRES_IA, PRES_PA, ADMINISTRATIVE |
| FY2021 | 10/01/2020 to 09/30/2021 | 7/19/2022 | PRES_IA, PRES_PA, ADMINISTRATIVE |
| FY2022 | 10/01/2021 to 09/30/2022 | 3/16/2023 | PRES_IA, PRES_PA, ADMINISTRATIVE |
Two things in that table limit year-to-year comparison. The declaration types change between files: 2 files list PRES_IA, ADMINISTRATIVE; 2 files list ALL; 4 files list PRES_IA, PRES_PA, ADMINISTRATIVE. And each file was run at a different time after its year closed, so files run soon after year end had less time for amounts to be revised. The FY2015 run date is printed as 01/11/2015, before its own reporting period ends; it is reproduced as printed.
Method
How we counted. We read the Home and Business sheets of the eight SBA Disaster Loan Data workbooks, FY2015 to FY2022, retrieved 6 October 2026: 26,732 home rows and 16,755 business rows. Verified loss is defined as Verified Loss Real Estate plus Verified Loss Content, not the Total Verified Loss column, because in 1,595 home rows (FY2019: 1,595) the published total is zero while the two parts are not ($2.4bn). In all but 11 other home rows the parts and the published total agree within $1. Summing the published total column instead gives $28.9bn. One home row (LA-00115, Houma 70364, FY2021) is excluded from every figure: its verified contents loss of $2.4bn is 58 times its own real estate loss and 12 times the next largest contents value in any row, and approved contents dollars cover 0.31% of it. The rule applied is contents above $100m and above 10 times real estate; no other row meets it. With the row included, verified home loss is $33.7bn and the approved share 44.4%. Two business-sheet defects were repaired, not dropped: the header row of FY2016 (1,336 rows) is truncated in the source, so that sheet was read by column position in the order of SBA’s Data Menu; and 315 business rows in FY2015 and FY2016 are missing their FEMA number cell, with every later value one column to the left, so they were moved back. Approved share is summed approved total over summed verified loss. FEMA incident types come from joining the FEMA Disaster Number to DisasterDeclarationsSummaries v2. State tables use the damaged-property state code; $888m of verified home loss in other territories or with a blank code is kept in national totals only. ZIP figures sum rows by five-digit ZIP. All dollars are nominal.
Reproduce this
Reproduce this
- Endpoint
https://data.sba.gov/dataset/disaster-loan-data and https://www.fema.gov/api/open/v2/DisasterDeclarationsSummaries- Query
Download 'Disaster Loan FY15' to 'Disaster Loan FY22' (xlsx); read sheets 'FYnn Home' and 'FYnn Business' from the row starting 'SBA Physical Declaration Number'. FEMA: $select=disasterNumber,incidentType,declarationDate,declarationTitle- Rows retrieved
- 26,732 home rows and 16,755 business rows in eight workbooks
- Retrieved
- 6 October 2026
The SBA files are static downloads with no API. Check the Business sheet header of FY2016 and rows whose FEMA Disaster Number field holds an SBA disaster number before summing; both are described in Method.
What this does not show
What this does not show
- No loan counts. Rows aggregate a declaration and a city and ZIP. No number of loans, applicants or borrowers can be derived, so no per-loan figure exists here.
- Not final values. SBA publishes ‘original’ verified loss and approved amounts, which can be revised up or down later. The files say they are not an official Federal report.
- Raw entry data. SBA warns the data may have been entered by survivors and contain errors. One implausible row is excluded; smaller errors cannot be detected.
- The gap is not unmet need. Verified loss beyond approved loans includes insured loss, grants, declined and withdrawn applications. The files record none of these separately.
- Fiscal year, not disaster year. Files run 1 October to 30 September and track declaration dates, and the declaration types covered change between files.
- Series ends in FY2022. No later file is listed on the dataset page as of 6 October 2026, so recent disasters are absent.
- Nominal dollars. No inflation adjustment is applied across the eight years.
Most quotable figures
- SBA approved $14.9bn in disaster home loans against $31.3bn of verified home loss, FY2015 to FY2022: 47.7% (SBA Disaster Loan Data, retrieved 6 October 2026).
- Contents are 24.8% of verified disaster home loss in SBA data, and 35.4% in flood disasters (SBA Disaster Loan Data, FEMA).
- In fire disasters SBA approved home loans equal to 23.5% of verified loss; in floods, 50.0% (SBA Disaster Loan Data, FEMA incident types).
- Hurricane Harvey accounts for 18.4% of all verified disaster home loss in SBA data for FY2015 to FY2022 (SBA Disaster Loan Data).
- Half of verified disaster home loss in SBA data sits in 175 ZIP codes (SBA Disaster Loan Data, FY2015 to FY2022).
- SBA business disaster loans equal 21.7% of verified business loss, FY2015 to FY2022 (SBA Disaster Loan Data).
Questions this report answers
What do SBA disaster loan statistics show?
For FY2015 to FY2022, $31.3bn of verified home loss and $14.9bn of approved home loan dollars (47.7%), plus $24.5bn of verified business loss and $5.3bn approved for businesses.
Can these files give an average SBA disaster loan size?
The public SBA files cannot answer that. They aggregate dollars by declaration, city and ZIP and publish no count of loans, so no average can be computed from them.
Why is the approved amount so much lower than the loss?
SBA lends only for losses not covered by insurance or other sources and assesses credit and ability to repay. Insured loss, declined and withdrawn applications all sit in the difference, and the files do not separate them.
How much disaster damage is contents rather than structure?
24.8% of verified home loss is contents, rising to 35.4% in flood disasters and falling to 19.9% in fires.
Is there SBA disaster loan data after 2022?
Not on the public dataset page as of 6 October 2026. The most recent file is FY2022, ending 30 September 2022.
Can I reuse these figures?
Yes, under CC BY 4.0 with attribution to Remediation Insider. For decisions about a specific loan, contact SBA directly.
How to cite
Remediation Insider. SBA Disaster Loan Statistics: Verified Losses Against Loans Approved, FY2015 to FY2022. https://remediationinsider.com/sba-disaster-loan-report/ (retrieved 6 October 2026).
Data and tables in this report are released under CC BY 4.0. Attribution: Remediation Insider.
Sources
- U.S. Small Business Administration. SBA Disaster Loan Data, files FY15 to FY22 (xlsx), including the Instructions and Data Menu sheets. Retrieved 6 October 2026.
- Federal Emergency Management Agency. OpenFEMA DisasterDeclarationsSummaries v2, for incident type, title and declaration date. Retrieved 6 October 2026.
- SBA, Physical damage loans, for insurance deduction, mitigation and refinancing terms.
- SBA, Disaster assistance, for credit elsewhere and ability to repay.
- FEMA FloodSmart, for flood coverage under homeowners policies.
- No survey, interview or inspection was conducted for this report. Every figure is computed from the retrieved files.
This product uses the Federal Emergency Management Agency’s OpenFEMA API, but is not endorsed by FEMA. The Federal Government or FEMA cannot vouch for the data or analyses derived from these data after the data have been retrieved from the Agency’s website(s).