Every journalist has ideas about what makes a good public records request. But surprisingly few people have actually tried to systematically analyze how requests can be written to improve their chances of success.
To fill this vacuum, we analyzed more than 33,000 Freedom of Information Act requests and identified a few characteristics that were typical of those that were fulfilled.
The requests were made to five federal agencies that publish to FOIAonline.gov: the Environmental Protection Agency, the Department of Commerce, Customs and Border Protection, the Department of the Navy, and the National Archives and Records Administration. All were filed between 2011 and 2016.
We defined success as the receipt of all records requested, as defined by the agency. There was no straightforward relationship between wait time and any of the characteristics we considered, so we factored it out as a measure of success.
For the requests we examined, the full-grant rate across all five agencies was around 23 percent. That’s the same success rate for requests across all federal agencies, according to Max Galka of FOIA Mapper, a project funded by the Knight Foundation that outlines the record systems of federal agencies.
Requesters in our sample typically waited around 142 days, or a little more than four months, to get responses. Less than 39 percent of requests received responses within 28 days, which is the longest amount of time an agency could spend fulfilling a request while still meeting FOIA’s 20-business-day time limit.
That’s a pretty bleak picture. So, how can you improve your chances?
Brevity isn’t always beneficial.
Perhaps surprisingly, few features were consistently associated with full grants across all five agencies. One characteristic that did predict successful requests was simple length. For example, exceptionally short requests (i.e., tweet-sized requests under 24 words long) were given full grants only 17 percent of the time, whereas more substantive requests were far more likely to be fulfilled (29 percent).
Although requests containing fewer than 24 words may seem unusually short and vague, slightly more than half of the requests sampled were even shorter. (The average length of the requests we examined was around 42 words.) This group of shorter requests appears to encompass what we would call throwaways—or requests people throw together and send away without serious effort on the off chance that it gets a useful response.
Another feature associated with a higher full-grant rate was the use of links to website, which requesters used to provide context and clarify the kinds of records they desired. Although only around one percent of requesters used them, links boosted requesters’ chances from 23 percent to 36 percent. The presence of a link was not highly correlated with any of the other variables we considered.
Finally, requests that specifically asked for “data” were given full grants at a markedly higher rate than others—36 percent as compared to 23 percent. That said, the power of this effect varied considerably from agency to agency: Whereas data requests sent to most agencies were at least 12 percentage points more likely to receive full grants, these requests only saw a 4 percentage point bump when sent to the Department of Commerce.
Not all conventional advice was helpful.
Some typical suggestions from FOIA writing guides—like explicitly referring to the Freedom of Information Act or providing one’s email address or phone number—didn’t improve requests’ chances of a full grant. That doesn’t necessarily mean these details don’t help, but our data suggests they’re not key to success.
Additionally, although some FOIA-writing guides advise journalists to make their requests as formal as possible, this advice also doesn’t look helpful. Our analysis provided no evidence for the idea that social formality benefited requesters’ chances. In fact, of the almost 900 requests from our sample that started with formal greetings like “Madam FOIA Officer” or “Dear Ms. FOIA Officer,” none received a full grant.
In these ways, conventional wisdom failed to translate to real results in our sample. But that shouldn’t be too shocking. It’s simply too much to expect a human to be able keep a complete and reliable tally of countless FOIA requests and distill their content and outcomes into accurate advice. (And that’s exactly why we felt this study was important to do.)
That said, requesters to one agency seem to have honed in on a successful formula.
EPA requesters know what they’re doing.
By including identification numbers for facilities or documents of interest, people filing requests to the EPA saw their probability of a full grant jump from 42 percent to 64 percent. When an identification number wasn’t provided in an EPA request, other language related to describing the records requested became more important. In particular, requesters that asked for specific data or narrowed their request with dates achieved full-grant rates of over 60 percent.
Requesters who used knowledge of the EPA’s organizational structure by specifying which EPA Region had the records they sought or which EPA Site was the subject of their request enjoyed success rates of about 62 percent, regardless of whether they gave an identification number.
The EPA had the highest full-grant rate of all the agencies we examined. Are FOIA officers at the EPA more amenable? Maybe. But our analysis also found that fulfilled EPA requests were more sophisticated than requests sent to other agencies, and this detail had an especially large impact on success rates. In this regard, EPA requesters have a valuable lesson to offer the rest of us: Know the agency to which you’re sending your request.
There are ways to unravel agencies’ records systems.
To understand how an agency organizes its available records, the best place to go is often the agency itself.
In the case of the EPA, its website lays out explicit instructions for those who wish to submit a FOIA request. For example, the EPA states FOIA requests should be directed to regional offices. (Our analysis suggests that mentions of a regional office number appear in fulfilled EPA requests at slightly higher rates.) The EPA’s website also has a “Docket Center” where one can explore regulations and associated reports—a valuable starting point for finding identification numbers.
FOIA.gov also provides contact information for offices obscured within the larger federal departments and agencies. Based on the experiences of FOIA-experts like Galka, calling the agency’s officer can sometimes be the most efficient way of understanding an agency’s record system. (Our data didn’t capture any correspondence between the requesters and FOIA officers outside the requests themselves.)
Beyond agency resources, Galka’s FOIA Mapper project provides official documents submitted to the National Archives that lay out the framework of agencies’ record systems. Documents like these can serve as an useful starting point for asking concrete, targeted questions if you plan to reach out to a FOIA officer directly.
Additionally, the growing community of FOIA enthusiasts can be a resource. A previous CJR piece noted that requesters have compiled an incredible amount of information about the nuances of crafting requests to specific agencies. Accessing that budding community of experts can be as simple as following #FOIA on Twitter.
Our nation needs good FOIA requesters.
The Freedom of Information Act has been an unparalleled legal spotlight for exposing the inner workings of the federal government since 1967. Effectively exercising our right to government records offers a path forward at a time the public’s trust in officials and news organizations is shaken. Legally-acquired, official records also serve as a much-needed buffer against claims of media bias.
Additional reporting by Jake Crouse. View our code and analysis on Github.
Nicholas Dias, Rashida Kamal and Laurent Bastien are students at the Columbia Journalism School. Dias and Bastien are pursuing master's of science degrees with a specialty in data, and Kamal is pursuing a dual degree with journalism and computer science.