Long Live the Letter of Recommendation

In my last Vantage Point column, I announced that AI was killing the cover letter.

Indeed, it has now been declared dead.

AI killed the cover letter because of the new technology’s ability to quickly, convincingly, and effortlessly replicate what might have otherwise taken an applicant hours to produce.

By making the process too easy, AI eliminated the cover letter’s signaling value to employers.

In the old days, a detailed thoughtful cover letter — the kind that demonstrated a knowledge of the firm and clearly articulated how a particular candidate’s skills and experience could fill a needed role — was difficult enough to write that applicants might only do it for jobs that were truly a great fit. Employers only saw strong letters from candidates who thought the role was right for them. Those candidates were worth giving a careful look.

When ChatGPT or Claude can spit out a great cover letter in seconds, however, having one becomes a prerequisite rather than a differentiator. It is no longer a useful signal of anything (except, perhaps, that a candidate knows how to prompt a chatbot).

But a well-functioning labor market still needs to help match candidates to jobs; without the cover letter, it needs another way to show firms that certain candidates are good fits for their open roles and worth investigating further.

A Signal AI Can’t Create

Thankfully, the answer is not far. It’s the letter of recommendation.

The key to the value of the letter of recommendation is that its signaling power never came from it being hard to write. (They aren’t. Many letter of recommendation writers use a template or ask the subject to put together a first draft themselves.)

Instead, its signaling power came from the fact that the sender was willing to write it at all. When a former employer, supervisor, or professor (like me) writes a letter, sends an email, or picks up the phone to vouch for a job candidate, we are trading with a currency that is in limited supply and costly to waste: our reputation.

I want all my students to succeed, but I am only going to recommend one for a job opportunity or academic program if I truly think that they will be a good fit.

Even if I care more about my students than the firms that might hire them, I wouldn’t burn my reputation by recommending someone mediocre for a role. If I recommended an unsuitable candidate for a job this year, the firm would quickly learn that I could not be trusted and ignore my recommendations in the future. As a result, I only recommend students when I think the match will be successful, and those reviewing my reference can be confident in my assessment.

And because even letters of recommendation that AI helps write have this same signaling value, they will continue to hold sway, even in the AI era.

When a former employer, supervisor, or professor (like me) writes a letter, sends an email, or picks up the phone to vouch for a job candidate, we are trading with a currency that is in limited supply and costly to waste: our reputation.

Evidence From the Labor Market

Indeed, as other signals lose their power, the value of a letter of recommendation is likely to rise. The transition is already starting to happen. A recent working paper found that when candidates started using AI to write cover letters, employers started ignoring them and started relying more heavily on recommendations.

A current limitation with letters of recommendation, however, is that not all job candidates appreciate their value.

I saw this firsthand in research I conducted with University of Michigan’s Sara Heller. We focused on youth workers (aged 24 years old and younger) who had participated in the New York City Summer Youth Employment Program.

We were interested in how these workers transitioned from their summer employment — funded by the city program — to other labor market opportunities. In particular, we wondered whether they were taking full advantage of the power of recommendations from former employers.

We suspected they were not, and we thought that was a mistake.

For our test, we randomly selected youth to be eligible to receive letters of recommendation provided by their supervisors from the summer program. We invited their supervisors to indicate the youth who they thought were deserving of a letter, asked them to complete a short survey about each youth, and turned positive survey responses into letters. (Our analysis of the supervisor survey responses found that they captured labor market readiness; for example, youth with prior labor market experience received substantially higher ratings from supervisors.)

We produced the letters, sent them to the youth, and then checked back in — four years later — to compare the labor market outcomes of those who had been randomly eligible to receive a letter to those who had been randomly selected to be our control group.

We estimated that receiving a letter of recommendation increased earnings for a worker by 4.9% over four years.

We knew that if youth were already optimizing (e.g., by asking supervisors to write letters or serve as references whenever it would be helpful for them to do so), then our letters would have done nothing.

Instead, the impacts were striking. Those randomly selected to receive letters made better matches on the labor market. They found jobs faster, stayed in them longer, and earned significantly more. We estimated that receiving a letter of recommendation increased earnings for a worker by 4.9% over four years.

We learned that the youth benefited dearly from having these letters of recommendation. But despite the benefit of having a letter, the youth were not asking for letters themselves.

Worse still, they did not always use the letters, even when we provided them. Even after we sent youth letters (five hard copies and one soft copy) and encouraged them to include them in their job applications, only 41% bothered to do so.

To establish this, we invited a subset of the youth in our study to apply for a short-term job working for me at the University of Pennsylvania. We included an opportunity to upload materials in support of their application, and a minority of our letter recipients included their letter of recommendation.

This research suggests that letters of recommendation — and related signals like someone serving as a reference, sending an email, or making a call on your behalf — can be a powerful indication of a candidate’s ability and fit at a firm.

But applicants must be aware of the value of these signals, seek them out from people willing to provide them, and then use them when they can. In a hiring market reshaped by AI, the most valuable signals might be ones that others help you send.

For more on labor market dynamics, see Kessler’s book Lucky by Design: The Hidden Economics You Need to Get More of What You Want.

Judd Kessler is a Professor of Business Economics and Public Policy at Wharton. In his research, Judd uses a combination of laboratory and field experiments to answer questions in Behavioral Economics, Public Economics, Labor Economics, and Market Design.

Source: View the original article at Knowledge@Wharton.