A deep dive into EPA’s Lead Service Line Replacement Funding Allotment Formula
By Erica Galante-Johnson and Marc Santos
In May 2024, the White House and EPA announced the third round of allotments through the Bipartisan Infrastructure Law (BIL) - also known as Infrastructure Investment and Jobs Act (IIJA) - for lead service line replacement (LSLR). This constituted the third of five years of the $15 billion appropriated for BIL LSLR funding.
EPA undertakes a Drinking Water Infrastructure Needs Survey and Assessment (DWINSA) every four years to allocate Drinking Water State Revolving Funds (DWSRF), which includes BIL funds for Federal Fiscal Years (FFY) 2022-25. To allocate BIL LSLR funds in accordance with state needs, EPA added survey questions related to service line material [1] starting in the 7th DWINSA. Using survey results, EPA allocates BIL LSLR funding to states through the DWSRFs based on their projected proportional share of national estimated lead service lines (LSLs) [2]. EPA initiated a one-time update to survey responses in late 2023 to improve alignment between funding with states’ needs. None of the DWINSA survey data or the corresponding DWSRF allotment formulas are directly linked to lead service line inventory data which is due to EPA, according to the Lead and Copper Rule Revisions, by the October 16, 2024 deadline.
Despite this update, EPIC found that these LSL-specific allotments continue to be misaligned with state needs, and that high-burdened (high number of LSLs) states are currently woefully underfunded while funding for low-burdened states far exceeds their needs. We believe that this misalignment is largely driven by two factors:
Minimum one percent DWSRF allotment: The Safe Drinking Water Act (SDWA) requires that each state be allotted a minimum one percent of appropriated DWSRF funds. This makes sense when it comes to general water infrastructure needs, but does not make sense when it comes to funding that is specifically designated for lead service line replacement. As a result, this requirement inevitably provides a baseline of funding that simply isn’t needed by low-burdened states. Though surplus funding can be realloted to high-burdened states over time, this bureaucratic process only delays funding from reaching the communities who need it most. In FFY24 the one percent can be over $28 million, and in FFY24, 28 states (including Washington D.C.) received the one percent minimum allotment for LSLR.
EPA’s methodology: EPA’s methodology to estimate states’ LSL numbers that form the basis on which allotments depend assumes that a portion of the unreported lines or unknown material lines are expected to be lead and incorporates those predictions in the state-level LSL estimates. By doing so, states with high numbers of unreported or unknown lines can end up receiving disproportionately large allotments. This may end up being the case in Florida, for example, which received the largest allotment in FFY23 and the second highest allotment in FFY24.
Eliminating the minimum one percent allotment would require changes to SDWA and therefore an act of Congress – which may be a lengthy and time consuming process though one EPIC would support. In this blog, however, we focus primarily on examining EPA’s current allotment methodology and recommend potential adjustments for consideration to ensure LSLR funds are distributed quickly and equitably.
EPA’s current methodology favors states with incomplete inventories
Funding is greatly impacted by unknown materials and unreported lines
EPA’s service line material categorization includes service lines for which the pipe material is not known (unknown) and lines for which no information is available because the system did not provide a survey response (unreported) [3]. Because these lines must ultimately belong to a specific material category (e.g. lead, galvanized, non-lead), EPA assumes that the proportion of unknown and unreported service lines that are expected to be made of lead is the same as the proportion of known lead service lines.
This extrapolation of the known proportion of lead is objectively reasonable. In practice, however, assigning the same weighting to “unreported” or “unknown” lines expected to be lead as “reported lead” lines in state LSL estimates can disproportionately award larger allotments to states with incomplete inventories.
To illustrate this point, we created four hypothetical service line material reporting scenarios by varying the proportions of each material for a state with one million service line connections. For each scenario, we compared the estimated number of LSLs before (initial) and after (projected) accounting for unknown and unreported lines. For simplicity, the unknown materials and unreported lines are collectively referred to as “unknown” in our scenarios. Lastly, we compared how expected allotments vary across scenarios before and after accounting for these unknown and unreported lines.
| Scenario | Description | Example |
|---|---|---|
| High Unknown - High Lead | Large proportion of reported unknown and unreported lines, and a large proportion of reported lines are classified as lead. | A state with several well-inventoried cities developed prior to 1950 but mostly underreported suburbs/rural areas built post-1970. |
| High Unknown - Low Lead | Large proportion of unknown and unreported lines, and a low proportion of reported lines are classified as lead. | A state with several well-inventoried cities developed post-1970 but mostly unreported suburbs/rural areas. |
| Low Unknown - High Lead | Low proportion of reported unknown and unreported lines, and a large proportion of reported lines are classified as lead. | A state with a significant number of communities developed prior to 1950, all with well-reported inventories. |
| Low Unknown - Low Lead | Low proportion of reported unknown and unreported lines, and a low proportion of reported lines are classified as lead. | A state with a significant number of communities developed post-1970, all with well-reported inventories. |
Table 1. Description of the four hypothetical service line material reporting scenarios used to compare BIL LSLR allotments after accounting for unknown and unreported lines expected to be lead.
The results of the four scenarios are shown below in Figure 1. In both scenarios with a high proportion of known lead lines (~20%), the initial allotment amounts are identical (approximately $51.6 million). However, once unknown and unreported lines are incorporated into the final LSL estimate, the High Unknown - High Lead scenario received $14.3 million more in estimated funding than than the Low Unknown - High Lead scenario.
This considerable difference in funding is directly attributable to the additional projected LSLs derived from unknown and unreported lines in the high unknown case. Assuming 20% of all unknown and unreported lines are lead, adds 40,000 more LSLs to the High Unknown - High Lead scenario LSL estimate, thus increasing the state’s projected percentage of national LSLs from 2.2% to 3.1%. However, applying the same assumption to the Low Unknown - High Lead scenario only increased the estimated LSLs by 6,000 (from 2.2% to 2.3%).
Figure 1. Estimated BIL LSLR allotment amounts for four service line reporting scenarios in a hypothetical state with 1 million connections before (blue) and after (green) incorporating LSLs derived from unknown and unreported service lines.
Conversely, in both scenarios with low proportions of unreported and unknown lines, the projected LSL funding remained unchanged after adjusting LSL estimates. In both of these scenarios, the hypothetical state would still receive the minimum 1% allotment directed by SWDA ($28.7 million) because the added LSLs are not sufficiently substantial to place said state above the minimum allotment threshold (1% of national LSLs).
These results underscore how EPA’s current method to account for unreported and/or unknown lines in its LSLs estimates, in practice, disproportionately benefits states with largely incomplete inventories. This is especially concerning in cases where a state has high proportions of reported lead lines due to a few water systems that have been very well inventoried but are not representative of the rest of the state (High Unknown - High Lead scenario).
For example, this would be the case of a hypothetical state with several cities developed prior to the 1950s which contain a substantial number of LSLs, and the water systems serving these cities report these LSLs. However, a large portion of the population in the rest of the state is served by water systems in suburbs and other newer cities developed after the 1970s which have little to no lead in their distribution systems. Because lead lines are of lesser concern in these newer communities, they would place less priority on updating their inventories.
We note, however, that for our analysis, we used predetermined service line proportions in each scenario, whereas EPA calculates these proportions weighted by a sample weight [4]. Because sample weight data is not available we could not replicate EPA’s calculations exactly.
We recommend the EPA provide more detailed information to the public on sampling weights in its upcoming report to Congress to increase transparency on the allotment process.
LSL estimates alone may not be sufficient to align state funding with needs
A presumably simple solution to address misalignments stemming from considering lead lines derived from unknown and unreported services equal to known lead lines is to apply a penalization scheme. Such penalization schemes could be used to ensure known lead lines weigh more heavily in state-level LSL estimates than those derived from unknown and unreported lines.
Though these discrepancies can be dramatically reduced, states similar to the High Unknown - High Lead scenario are bound to receive larger allotments even when severe penalization weights are applied to lead lines derived from unknown/unreported lines.
This suggests that although LSL estimates are a determining factor, these estimates alone may not be enough to accurately reflect state needs. In fact, the use of lead plumbing was federally banned in 1986 [5] though lead piping had been largely supplanted by copper by the mid-1950s. For this reason, lead service lines are unlikely to represent a high-priority issue in newer communities. EPA and other federal agencies use additional information such as age of housing stock and sociodemographic factors to identify areas of high exposure to other lead hazards (e.g. paint, soil). EPA’s EJ Screen calculates its “Lead Paint EJ Index” based on the percentage of homes built prior to 1960 along with other socioeconomic and demographic factors. We recommend that EPA generate a similar index for lead service lines and use it as the basis to determine state needs.
Figure 2. Trends for estimated BIL LSLR allotment amounts across a series of penalization weights (0.05 - 0.95) applied to lead lines derived from unknown and unreported lines in a hypothetical state under two reporting scenarios. High proportion of reported lead lead and of unknown and unreported lines (blue). High proportion of reported lead and low proportion of unknown and unreported lines (green).
What should we keep in mind moving forward?
Matching funding with state needs when LSL data is rapidly evolving
Both the Lead and Copper Rule Revisions (LCRR) and proposed Lead and Copper Rule Improvements (LCRI) require all states to submit their initial service line material inventories by October 16, 2024 to be updated annually thereafter. Additionally, many systems across the country have been actively replacing their lead lines. As inventories and replacement efforts ramp up, distributing LSLR funding in accordance with rapidly changing state and water system needs will prove increasingly challenging.
This raises questions as to whether DWINSA, which is distributed every four years, will be able to keep pace with rapidly evolving LSL data to accurately reflect state needs. All of this represents another reason to re-evaluate the formula for the allotment of LSLR funds. Temporal mismatches in the LSL data could be potentially minimized by incorporating additional data that is not subject to rapid changes to estimate state needs, such as age of housing stock.
The Lead and Copper Rule Improvement will accelerate LSLR
With the proposed Lead and Copper Rule Improvement (LCRI) on track to be finalized by this October, the possibility of a ten-year timeline for 100 percent lead service line replacement looms on the horizon. Once the rule goes into effect, all of the concerns regarding matching funding with rapidly shifting needs outlined above will be magnified. It is widely acknowledged that BIL funding alone will not be enough to replace the nation’s estimated 9 million lead service lines which can cost between $72 billion [6] to over $90 billion.
As we approach the end of BIL, with only two funding cycles remaining, it is time to recognize the need for additional LSLR funding. It is equally important to consider how future funding should be channeled, if and how they should flow through the DWSRF, and if additional policies need to be in place to ensure the focus is on an equitable and efficient distribution of funds.
Methodology
To demonstrate the impact of incorporating lead lines derived from unknown and unreported services, we created four hypothetical service line material reporting scenarios varying the proportions of each material for a state with one million service line connections.
Creating service line reporting scenarios
Scenario 1: Large proportion of reported unknown and unreported, and a large proportion reported lead (High Unknown - High Lead).
Scenario 2: Large proportion of unknown and unreported, and a low proportion reported lead lines (High Unknown - Low Lead).
Scenario 3: Low proportion of reported unknown and unreported, and a large proportion reported lead (Low Unknown - Low Lead).
Scenario 4: Low proportion of reported unknown and unreported, and a low proportion reported lead (Low Unknown - Low Lead).
For simplicity, unreported and unknown are referred to collectively as “unknown” in our scenario labels. The proportions assigned to each service line material were modeled after real DWINSA data to make the scenarios more realistic.
| Scenario | Lead | Non-lead | Unknown | Unreported |
|---|---|---|---|---|
| High Unknown - High Lead | 20% | 40% | 20% | 20% |
| High Unknown - Low Lead | 5% | 40% | 50% | 5% |
| Low Unknown - High Lead | 20% | 75% | 3% | 2% |
| Low Unknown - Low Lead | 1% | 95% | 2% | 2% |
Table 2. Detailed breakdown of service line material proportions in accordance to each of the hypothetical reporting scenarios. For simplicity, unknown and unreported lines are collectively referred to as “unknown” in the scenario label.
Estimating allotment amounts across scenarios
For each scenario, we calculated the following metrics while holding the total number of service lines constant (one million):
A) Initial lead estimate: To obtain initial LSL estimates under each scenario, we multiplied the proportion of reported lead lines by the total number of service connections (one million), without accounting for unknowns and unreported lines.
B) Allotment formula approximation: To determine the allotment amounts, the hypothetical state would receive under each scenario, we followed EPA’s methodology as reported in the 7th Report to Congress [7]. However, given that EPA’s allotment formula is not publicly available, we approximated the relationship between EPA’s projected percentages of LSLs [8] and BIL LSLR Federal Fiscal Year 2024 allotments via a linear regression.
Because the Safe Drinking Water Act (SDWA) requires each state to receive a minimum allotment of 1% (and 1.5% across all territories) [9], regardless of their number of LSLs, we only included states receiving more than the minimum allotment in the regression analysis as follows:
(1) Additional allotment percentage ~ Percentage of national LSLs
Formula 1. Linear regression formula describing the relationship between allotment percentages above the minimum one percent received by states and their estimated percentage of LSLs relative to the national total as calculated by EPA.
We found a perfect linear relationship (R² = 1) between the state percentage of estimated LSLs and the additional allotment percentages states received.
Figure 3. Relationship between estimated state lead service lines (LSL) percentages and the additional percent of Federal Fiscal Year 2024 allotment above the minimum one percent. State LSL percentages were obtained from EPA’s 7th Report to Congress. Only states receiving more than the minimum one percent allotment were considered.
| Slope | Intercept |
|---|---|
| 68.3 | -0.69 |
| R² | Standard Error |
| 1 | 0 |
| F-statistic | Degrees of Freedom |
| 12181936 | 21 |
| Regression SS | Residual SS |
| 103.1 | 0 |
Table 3. Results from the linear regression estimating the relationship between the state LSL percentages and additional Federal Fiscal Year 2024 allotment above the minimum one percent received by states.
Based on this regression were were then able to approximate EPA’s allotment formula as follows:
(2) Additional allotment % = -0.69 + 68.3 * State % LSLs
Where:
Additional allotment % is the additional Federal Fiscal Year 2024 allotment percentage beyond the minimum one percent allocated to a state.
State % LSLs is the state’s percentage of estimated lead service lines relative to the national total as reported in the 7th Report to Congress.
C) Initial allotment calculations: To calculate initial allotments under each scenario, we then applied the formula described above to the respective estimated LSL percentages based solely on reported lead lines. These percentages were calculated based on the national total LSLs reported by EPA in the 7th DWINSA update. Keeping in line with EPA’s methodology, we assigned the minimum allotment scenarios where the state’s percentage of LSLs fell below one percent of the national estimate. Otherwise, we applied formula (2) to calculate initial allotments.
| Scenario | Lead | Unknown | Unreported | LSL Percent National | Initial allotment (%) |
|---|---|---|---|---|---|
| High Unknown - High Lead | 200,000 | 200,000 | 200,000 | 2.2% | $51,570,000 (1.8%) |
| High Unknown - Low Lead | 50,000 | 500,000 | 50,000 | 0.6% | $28,650,000 (1%) |
| Low Unknown - High Lead | 200,000 | 30,000 | 20,000 | 2.2% | $51,570,000 (1.8%) |
| Low Unknown - Low Lead | 10,000 | 20,000 | 20,000 | 0.1% | $28,650,000 (1%) |
Supplemental Table 3. Initial LSL estimates and allotment amounts under each reporting scenario for a hypothetical state with one million lines based solely on reported lead lines.
D) LSLs derived from unknown & unreported lines: To calculate the number of unknown and unreported lines expected to be lead we followed EPA’s methodology by multiplying the initial estimates of these material categories by the proportion of reported lead lines for each scenario.
| Scenario | Unknown | Unreported | Total Expected Lead |
|---|---|---|---|
| High Unknown - High Lead | 40,000 | 40,000 | 80,000 |
| High Unknown - Low Lead | 25,000 | 2,500 | 27,500 |
| Low Unknown - High Lead | 6,000 | 4,000 | 10,000 |
| Low Unknown - Low Lead | 200 | 200 | 400 |
Supplemental table 4. Number of unknown and unreported lines whose material is expected to be lead under each reporting scenario given the proportion of reported lead lines.
E) Projected lead estimate: We then calculated the projected (i.e. “final”) lead service line estimates for each scenario by adding the initial LSL estimates (step b) and the LSLs derived from unknown and unreported lines (step d).
F) Projected LSLR allotment: Finally, we applied the same formula (2) to the projected lead percentages and evaluated how accounting for unknown and unreported lines influences the allotment amount the hypothetical state would receive under each scenario.
| Scenario | Lead | LSL Percent National | Projected allotment (%) |
|---|---|---|---|
| High Unknown - High Lead | 280,000 | 3.1% | $68,760,000 (2.4%) |
| High Unknown - Low Lead | 77,500 | 0.9% | $28,650,000 (1%) |
| Low Unknown - High Lead | 210,000 | 2.3% | $54,435,000 (1.9%) |
| Low Unknown - Low Lead | 10,400 | 0.1% | $28,650,000 (1%) |
Supplemental Table 5. Projected LSL estimates and allotment amounts under each reporting scenario for a hypothetical state with one million service lines after accounting for unknown and unreported lines.
Estimating allotments under different penalization schemes
To illustrate how penalization schemes can reduce the allotment disparities stemming from EPA’s methodology to estimate LSLs, we applied a series of penalization weights to both high lead scenarios (High Unknown - High Lead and Low Unknown - High Lead)and compared the resulting allotments.
To do so, we multiplied the number of expected lead lines derived from unknown and unreported for each scenario by a series of penalization weights ranging from 5% to 95% and proceeded to estimate projected LSLs as described above. Meaning that if a penalization factor of 25% was applied, a lead line derived from an unknown/unreported line would weigh three quarters (0.75) of a known lead line in the projected LSL calculations. Finally, we calculated the corresponding allotment amounts after LSL calculations were adjusted by each penalization factor.
[1] US EPA, Office of Water. 2023. Drinking Water Infrastructure Needs Survey and Assessment: 7th Report to Congress p. 23. https://www.epa.gov/system/files/documents/2023-09/Seventh%20DWINSA_September2023_Final.pdf
[2] Ibid
[3] Ibid p. 42
[4] Ibid
[5] 40 CFR §143.13 as revised June 19, 1986. https://www.ecfr.gov/current/title-40/chapter-I/subchapter-D/part-143/subpart-B/section-143.13
[6] US EPA Economic Analysis for the Proposed Lead and Copper Rule Improvements. Office of Water (4607M) EPA 815-R23-005 (Nov. 2023), https://www.regulations.gov/document/EPA-HQ-OW-2022-0801-0712
[7] US EPA, Office of Water. 2023. Drinking Water Infrastructure Needs Survey and Assessment: 7th Report to Congress pp. 23-24. https://www.epa.gov/system/files/documents/2023-09/Seventh%20DWINSA_September2023_Final.pdf
[8] Ibid p. 26
[9] 40 CFR § 35.3515 as revised August 7, 2000. https://www.ecfr.gov/current/title-40/chapter-I/subchapter-B/part-35/subpart-L