The Dataset of Electric School Bus Adoption in the United States is often quoted as the authoritative source for tracking data on electric school bus (ESB) adoption nationwide. It tracks ESBs through four stages of the adoption process. These are:

1. Awarded: A fleet operator is awarded funds, most often from a public entity, to purchase an ESB.
2. Ordered: The fleet operator submits an order with a bus dealer to purchase an ESB.
3. Delivered: The ESB arrives at the fleet operator’s bus depot, yard, or other transportation hub.
4. Operating: The ESB is now used on regular routes transporting students.

Historically, the ESB Initiative indicated the quarter (for instance, 2021 Q4) that the ESB was awarded, ordered, delivered and first operating when the data were publicly available. This enables tracking of the time it takes buses to reach each stage of adoption and tracking the number of buses at different stages by quarter. The most recent known adoption stage is indicated as “Current status of bus.” If there is no known publicly available data, these fields are left blank. 

However, the amount of missing data presents an incomplete picture of ESB adoption nationwide and several limitations for secondary analyses. Information for each stage of ESB adoption is not always released publicly, creating a significant knowledge gap. 

Currently, there is enough public data to, in combination with internal research and a set of assumptions, provide quarterly estimates for ESBs through the four stages of the adoption process. This document explains how the estimations were developed and applied in the dataset for unknown adoption stages. 

How the estimations were calculated

First, for the bus-level data in the adoption dataset (current through November 2023), each bus observation was categorized into a relevant year determined by the year of earliest adoption stage that is observed. Any bus observation without information on when the bus was adopted is excluded from this estimation analysis. 

Second, to reduce bias from larger orders from single school districts in the same quarter, we reduced the dataset to single bus observations per district per time period. This was done by deduplicating the dataset so that each school district and adoption stage observations were represented only once, regardless of the number of buses. 

Third, the dataset was further categorized into three time periods, where years were clustered together based on overall trends in electric school bus adoption. This was also done to reduce the bias of smaller fluctuations year to year. The three time periods were categorized based on increases in overall volume of buses in any adoption stage: 

1. Early, 2014 – 2018;
2. Middle/COVID, 2019 – 2021
3. The Clean School Bus Program and after, 2022 – present

Fourth, using existing data within each of these time periods, the median number of quarters between each pair of adoption stages (awarded to ordered, awarded to delivered, awarded to in operation, ordered to delivered, ordered to in operation, and delivered to in operation) were calculated. Calculations were done separately for each time period. Instances where there were fewer than five observations for the pair of adoption stages were excluded from use in the estimation process.  

The following table summarizes those observed time periods, with those in red excluded as there were fewer than five observations. 

How the estimations were applied

The above data were added to the adoption dataset to fill in missing time periods for every adoption stage for each ESB. The estimations were only applied for unknown adoption stages. For example, if we knew a bus was first awarded in 2020 Q1 but we did not know when it was ordered, we counted forward three quarters and assigned the estimated ordered date to 2020 Q4. If we knew a bus was awarded in 2018 Q1 and ordered in 2018 Q4, we did not adjust the ordered date to align with the estimation formula. The publicly available data superseded all estimations. In this example, we only counted forward seven quarters from 2018 Q1 to estimate a delivered date. 

There were several rules, assumptions, and exceptions applied throughout the estimation process. They are as follows:

• Estimates were not applied if the date of the most recent data source is within the current or previous quarter from the present. The reasoning is that we have up-to-date, publicly available information and should not assume the bus has moved through other stages.
• Estimations were not applied if the estimation placed the bus’s adoption status in the future (2024 Q2 or later).  
• Estimations were extrapolated from the awarded date for consistency across estimations and chronological purposes. For example, if a bus was first awarded in 2019 Q1 and we wanted to estimate when it was ordered and delivered, we counted forward three quarters from 2019 Q1 to assign an ordered date and counted forward five quarters from 2019 Q1 to assign a delivered date. Estimates were never made moving backwards or from other estimates. 
  • If a bus was committed but had no publicly available awarded data, we extrapolated a quarterly estimate for the awarded stage from the year of the earliest data source that mentioned the bus. For example, if the data source year was 2021, then the awarded estimation was assigned as 2021 Q1. 
• Some estimations were not applied to buses from certain funding sources that provide regular data updates, specifically the EPA’s Clean School Bus Program and the School Bus Replacement Program run by the California Energy Commission. 
  • The Clean School Bus Program provides data on ESB awarded, ordered and delivered dates. The unknown data for these three stages were not estimated, and instead will be completed when expected data is obtained. If the delivered date was known for a CSBP funded bus, an operating date was estimated. 
  • The School Bus Replacement Program provides data on ESB awarded and delivered dates. The unknown data for these two stages were not estimated, and instead will be completed when expected data is obtained. If the awarded and delivered dates were known, ordered and operating dates were estimated. 
• Occasionally our methods produced anachronistic estimations, meaning an ESB was estimated as operating before it was delivered, or delivered before it was ordered, and so forth due to a mix of estimated dates and sourced dates throughout the adoption process. This occurred for 67 ESBs (.01% of all committed ESBs) when estimations were first applied in January 2024. Through observation, it was noted that these ESBs moved through the adoption process very quickly, which may account for the incorrect estimations. 
  • For these buses, we corrected the chronological order on a case-by-case basis by applying the most recent publicly available source date to the anachronistic, estimated date. For example, if a bus had estimated adoption dates for awarded through delivered stages of 2013 Q1, 2013 Q2, 2015 Q1, but a publicly sourced operating date of 2014 Q2, then we changed the delivered date from 2015 Q1 to 2014 Q2. For another example, if a bus had publicly sourced awarded and delivered dates of 2021 Q2 and 2021 Q3 and estimated ordered and operating dates of 2022 Q1 and 2021 Q4, then we changed the anachronistic ordered and operating dates to 2021 Q3.
• When a new ESB is discovered to be delivered or operating that we were previously unaware of, the most recent publicly available source date denoting its delivered or operational status is added to the dataset. Absent any earlier adoption stage data in the dataset for this new bus, we count backwards based on the observed time periods in Table 1 to estimate relevant earlier adoption stage dates; awarded, ordered, and/or delivered.