NFIP Claims Data Summary
2023-12-07
Data Cleaning:
- Joined Claims with Census Data
- Filtered to years with Census Data, 1990 - 2012 (1990, 2000
Decennial, 2008-2012 ACS)
- Claims data associated with 2010/2012 Census Tract boundaries
- Historical data associated with 2012 boundaries acquired from
Longitudinal Tract Database (Brown University)
- Filtered to years with Census Data, 1990 - 2012 (1990, 2000
Decennial, 2008-2012 ACS)
- Census Data Cleaning:
- Calculate simplified percents of census estimates, calculated based on LTDB codebook
- Suppress tract estimates with high coefficient of variation (margin of error above accuracy threshold)
- Filter down to key variables (see selected variables in
Claims12_fields)
- All monetary data converted to 2012$
- Calculated monetary ratios:
- Claim Payment/Property Value
- Claim Payment/Assessed Damage
- Assessed Damage/Property Value
- Claim Payment + Deductible / Property Value
- Claim Payment + Deductibel / Assessed Damage
- Claim Payment/Property Value
- Calculated time frames:
- loss to claim: time between date of loss and date of claim submitted
- claim to close: time between date of claim submitted and date of claim closed
- loss to close: total time between date of loss and date of claim closed
Data Checks:
| fields | record_count | record_percent |
|---|---|---|
| Loss Date | 1521430 | 100.0% |
| Open Date | 1521430 | 100.0% |
| Close Date | 1401215 | 92.1% |
| Claim Payment for Building | 1089136 | 71.6% |
| Building Property Value | 1189503 | 78.2% |
| Assessed Damage | 1102013 | 72.4% |
| Closed w/o payment | 320823 | 21.1% |
| Closed w/o payment; w/ Reported Damage | 62296 | 4.1% |
| Claims from Catastrophes | 1092892 | 71.8% |
Data Questions
- How long do consumers take to file a claim after a loss, and how
long do claims take to process? How does that vary over time?
- How does claim payment and assessed damage (as a proportion of
the total property value) vary over time? How does the claim payment
ratio vary over time?
- How many claims were closed with no payment over time? What are the common reasons?
Claims Processing Time
How long do consumers take to file a claim after a loss, and how long do claims take to process? How does that vary over time?
NFIP claim processing over time suggests a few trends: the time between loss and claim being filed by a policyholder has decreased over time, the time it takes to process claims increases in years with major natural disasters that likely increase the volume of claims, and the amount of time for a claim to be processed seems to be relatively stable over time, except for big loss years (like Hurricane Katrina). This suggests that cross-year comparisons of claim-to-close time treatment is reasonable even without time fixed effects, if controlling for catastrophe or volume of claims.
The distribution of processing time is long-tailed: there are a relatively few records that appear to take a very long period of time to process. The statistics in the tables below are limited to time periods < 730 days (~ 2 years). That long tail of data could be from a few extreme natural disasters, or the product of record errors.
Average total time between date of loss and claim payment is ~ 2 months. Average time for claim payments to be made is about 7 weeks.
| time_variable | Minimum | Median | Mean | SD |
|---|---|---|---|---|
| claim_to_close | 0 | 49 | 100 | 964 |
| loss_to_claim | 0 | 8 | 23 | 58 |
| loss_to_close | 0 | 66 | 124 | 964 |
NFIP claims produced from “catastrophes” (as designated by NFIP) have a slightly longer claim payment time. NFIP claims that are “closed without payment” are significantly shorter claim-to-close time. This confirms some intuition that decisions to decline payment happen quickly, and decisions for payment take longer to assess and resolve.
| time_variable | Minimum | Median | Mean | SD |
|---|---|---|---|---|
| claim_to_close | 0 | 53 | 111 | 1146 |
| loss_to_claim | 0 | 9 | 22 | 50 |
| loss_to_close | 0 | 71 | 134 | 1146 |
| time_variable | Minimum | Median | Mean | SD |
|---|---|---|---|---|
| claim_to_close | 0 | 30 | 62 | 134 |
| loss_to_claim | 0 | 12 | 35 | 89 |
| loss_to_close | 0 | 53 | 96 | 161 |
Claim Amount Ratios
How does claim payment and assessed damage as a proportion of the total property value vary over time? How does the proportion of the assessed damage that is actually paid via claim payment vary over time?
There is not full data coverage for claim payment, damage estimate, or value estimate. There are also some records with illogical payment records: claims payment that are higher than the assessed damage or the property value, or negative monetary values (resulting in ratios below 0, or 1). For our analysis based on claims payment ratios, we remove those illogical payment records as well. The result is about 60 - 70% of claims with valid ratio estimates.
| payment_ratio | record_count | record_percent |
|---|---|---|
| payment/value | 1033544 | 67.9% |
| payment+ded/value | 1027331 | 67.5% |
| payment/damage | 983216 | 64.6% |
| payment+ded/damage | 931112 | 61.2% |
| damage/value | 1101631 | 72.4% |
| payment_ratio | median | mean | min | max | sd |
|---|---|---|---|---|---|
| payment/value | 0.114 | 0.238 | 0 | 1 | 0.269 |
| payment+ded/value | 0.124 | 0.247 | 0 | 1 | 0.271 |
| payment/damage | 0.891 | 0.822 | 0 | 1 | 0.191 |
| payment+ded/damage | 0.976 | 0.922 | 0 | 1 | 0.130 |
| damage/value | 0.117 | 0.260 | 0 | 1 | 0.302 |
NFIP claim payments and assessed damage as a proportion of the property value has stayed generally stable over time, with the exception of major natural disasters: Katrina and Sandy appeared to have caused damage that was on average a larger proportion of property value, and claim payments seem to have matched that. The proportion of the assessed damage paid in claim payments (represented by the red line) appears to have increased slightly over time.
Claims Closed without Payment
How many claims were closed with no payment over time?
We consider claims closed without payment as any claim labeled as “closed without payment,” or any claim labeled “closed” but with $0 or NA payment. A significant proportion (20-25%) of historical claims submitted are closed without payment. The noticeable decrease in 2012 may be due to many claims being unresolved at the tail end of the dataset.
The reasons for claims being closed without payment vary, and the categories are broad and vague. Of claims denied payment, a total of 28% have “Other” reason listed, or no reason listed at all, leaving much ambiguity as to the denials of those claims. 21% of claims are denied due to “not actual flood”. Another 31% are denied due to inadequate documented damage: either”no demonstrable damage” or “damage less than the deductible.” Note: counts and distributions are almost identical for building and contents claims denied.
| code | n_claims | percent | Description |
|---|---|---|---|
| 06 | 68438 | 21 | Not actual flood |
| 97 | 59614 | 19 | Other |
| 01 | 55109 | 17 | Claim denied that was less than deductible |
| 20 | 45862 | 14 | No demonstrable damage |
| 00 | 28533 | 9 | No Reason Listed |
| 99 | 19535 | 6 | Erroneous assignment |
| 08 | 11339 | 4 | Failure to pursue claim |
| 16 | 10845 | 3 | Not insured, wind damage |
| 98 | 7640 | 2 | Error delete claim (no assignment) |
| 02 | 3751 | 1 | Seepage |