As I explain in this video, this is what I have been trying to accomplish for the last 2+ years. It wasn’t possible until the most recent models and harnesses. But with this skill and the API into my NLRB Research database, you really can just ask any labor law question and get a high-quality answer with citations and links to case law.

The way it works is not by hoping Claude has the information in its training data, but rather by instructing Claude and the Claude Cowork/Code harness how to iteratively query my database, analyze the results, query some more, analyze some more, and so on until it can construct an answer. So, in other words, getting it to use the database the way a real lawyer would use a legal database, and indeed the exact way I use the NLRB Research database to research labor law questions.

Check out the video. I do intend to sell this for real to firms and unions in a SaaS-type model. Email matt@nlrbedge.com if interested.

I produced this YouTube video tonight about a topic that has irritated me for a while, which is the discourse around so-called “AI agents.” I think the way people talk about this is extremely confusing, mystical-sounding, and unhelpful.

In the video, I write my own LLM harness (or coding agent?) called Bruenig Code to show you what exactly people are talking about. I hope you leave the video with the understanding that, ultimately, LLMs are just like other APIs: you send text to them, they send text back. Strictly speaking, they don’t do anything other than that. They don’t control your computer, run your web browser, do your shopping, or whatever. You send text to them. They send text back. That’s it.

It just turns out that, if you cleverly construct other tools around this exchanging of text through an API, you can accomplish a lot.

I have never seen someone write a reference book using this method and, if you were to try to do so manually, it would be a monumental effort requiring a massive amount of human labor. I did not do this labor, but instead mostly used Google Gemini and Claude Sonnet to generate the book.

I have been talking about producing this book for several months on X and a number of people have asked for more of the technical details of how I did it. That is what this post is about.

Constructing the Database

A couple of years ago I created the NLRB Law Database at NLRBResearch.com. This database tracks fourteen different types of documents that essentially define the practical meaning of the National Labor Relations Act. The database is updated once a day using bespoke web scrapers that I wrote in Python. At the time, I used ChatGPT to help me a little with the coding, but not that much, as this was in the early LLM days before the code-writing ability of LLMs got really good.

This database is the centerpiece of the book-generation process. It contains all NLRB decisions issued from 1935 to present and every Supreme Court case that has ever used the phrase “National Labor Relations” along with tens of thousands of other sorts of documents that cite to those decisions. All of this material is stored in plain text with metadata like the name, date, and citation for the case.

Finding the Most-Cited Cases

Once the database was constructed, it was fairly trivial to find the most-cited cases. Using code, I made a list of every NLRB and Supreme Court decision in the database along with each decision’s citation. NLRB cases have citations that look like “90 NLRB 289” and Supreme Court cases have citations that look like “395 U.S. 575.” I conducted a search in the database for each of the cases on the list, e.g., this search for “90 NLRB 289.” Every case that contains that text shows up in the search results as a match and so you can add up the number of matches to determine how many times it has been cited.

Once you do this for every single NLRB and Supreme Court decision, you can simply sort the cases from most citations to least citations to find the 100 most-cited cases. Once you have the 100 most-cited cases, you can go back into the database, search for the citation of each of those cases again, and then pull down the 100 most-recent cases that cite to each target case in a JSON file.

So the result of this process is you end up with 100 JSON files, one for each of the 100 most-cited cases. Those JSON files contain the names, dates, citations, and full text of the 100 most-recent cases citing to the target case.

Producing the Summaries

It would be nice if you could just upload each of these JSON files to an LLM and ask it to produce a summary. But you cannot do this because the files are too big. For instance, the JSON file containing the 100 most-recent cases citing to “90 NLRB 289” is over 3.3 million tokens in length. The most tokens you can give a high-quality frontier LLM at a time is 1 million.

So what I did instead was write a Python script that walks through a JSON file one case at a time and sends each case’s information to Google Gemini Flash while prompting it to provide a 100-word summary of that case that focuses on how that case applied the target case. The prompt also instructed Gemini to send its response back as JSON so that I could compile all of the summaries into a new JSON file for the next step of the process.

So, for example, the most-cited case in NLRB history is F.W. Woolworth Co., 90 NLRB 289 (1950), which is a case about how to calculate backpay remedies. I grabbed a JSON containing the 100 most-recent cases citing to F.W. Woolworth in my NLRB Law Database. Then I went through that JSON file one case at a time and had Gemini provide a 100-word summary that described what legal rule F.W. Woolworth established and how F.W. Woolworth was applied in that case. This 100-word summary was sent back in JSON format with other metadata, allowing me to construct a new JSON file that contained 100-word summaries for each of the cases in the original JSON file.

Once I had this new JSON file — the one with the 100-word summaries — I sent that file to Claude Sonnet and had it produce a summary of the target case using the 100 Gemini Flash summaries. This text is what made it into the book.

To be clear, all of what I am describing above is being done with Python code that was written with Claude Code, mostly with the Claude Sonnet model though occasionally with the Claude Opus model. I didn’t do any of this manually.

Compiling the Book

Once you have all of these summaries, you still have to compile them into a book. People who buy The National Labor Relations Book get sent an email with a PDF, EPUB, and HTML version of the book. To produce those three things, I had to compile all of the summaries, with my edits and other contributions, into some kind of intermediary document that could then produce these files.

The way I did that was to have the final summaries from Claude Sonnet sent to me as markdown files. Markdown is a simple markup language that can create headings, subheadings, hyperlinked text, and so on. These markdown files were all placed in a directory with numerical prefixes to make sure they all lined up in the order desired for the book (so the first part of the book is 000-Introduction.md, the next part is 001-Method.md, and so on). Once the markdown files are all lined up in the directory like this, it is easy to move cases around in the book by changing their numerical prefixes and also easy to edit the files. Because markdown is just plain text, it was also possible to prompt Claude Code to move things around or to edit the text.

To compile the markdown into the PDF, EPUB, and HTML, I used pandoc. I actually created a Makefile that contained the pandoc command for generating each of the three files. And so whenever I wanted to compile the book to see what it looked like, I could just type “make pdf” or “make epub” or “make html” or, to do all of them, “make all.”

The book also contains a couple of graphical illustrations and a nicely-designed table in it. These were written in LaTeX, which was placed in the markdown files.

As with the prior section, basically all of the coding aspects of what I am describing here were done by Claude Code using the Sonnet model. It wrote the Makefile. It wrote the LaTeX markup. It installed pandoc, LaTeX, and all the other packages needed to do all of this.

Selling the Book

When it came time to sell the book, I wanted to see if I could do so without using an e-commerce platform that takes a cut of your sales. There are a number of such platforms, including Amazon, for self-published books. But, what’s the point of a coding agent if not to allow me to set all that up myself without having to pay a third party? I keep hearing that LLMs are going to kill SaaS. So let’s kill it.

I SSH’d into my NLRBResearch.com server, which also has Claude Code installed. I explained to Claude what I was wanting to do — take card payments and automatically email the PDF, EPUB, and HTML to buyers — and asked it to come up with a plan for doing it in the simplest way possible.

Claude concluded that the best way to do it would be to set up a small Flask app with Stripe (payment processing) and Resend (emailing) integrations. I already have a Stripe account for NLRB Research, so Claude walked me through how to create a new product on Stripe and how to get the relevant Stripe secrets/keys. It also walked me through how to set up a Resend account and get a Resend API key. It instructed me not to share those keys with Claude, which was reassuring, and then proceeded to write the Flask app, edit my nginx (server) configuration, and create a systemd entry that launches everything. The systemd entry had three empty variables where I was told to paste my Resend/Stripe keys, which I did.

And it worked. I am selling my book without paying a platform fee.

Conclusion

As I noted in my prior post about AI, I find all of this stuff incredibly useful. In some of the more practically minded LLM discourse (not the philosophical discourse), I keep seeing this painful phrase “solopreneur” and I guess that is what I am and have been for the last 10 years or so, with all of my projects. I’ve never hired an employee and only occasionally hire outside contractors, like writers, editors, and designers for People’s Policy Project.

For someone in my situation, LLMs enable me (1) to do different stuff than I have ever done before because it has abilities I do not, (2) to do more stuff than I have previously been able to do because it works faster, and (3) to in-house things I would have previously had to outsource to a particular person (like an editor) or to a service (like an e-commerce platform).

But one thing I think we can see with this book is that the LLMs are not just labor-replacing and productivity-expanding, but can, in some circumstances at least, enable the production of totally new things. Obviously there are other labor law introductions and reference books. But this particular method of deriving the law from 10,000 carefully selected cases is not something that would have been feasible without LLMs. In the absence of LLMs, the only option for producing a book like this is to use expert judgment. This can also be useful (indeed I used some of that in writing this book), but it is not the same product.

I do think this book will be useful for certain kinds of people who I hope will buy it. But I think the process of making it was also very useful as far as skill-building goes. LLMs look poised to become a standard aspect of most white-collar professions in the future. So better to get ahead of such things than get left behind.

Politically, I align with the socialist left and have since I was a teenager. On the internet, I am most well-known as a left-wing economic policy guy who runs the People’s Policy Project think tank and helps left-wing politicians — like Bernie Sanders and Zohran Mamdani — with their policy agendas. I also practice union-side labor law, maintain the NLRB Research database, and publish the NLRB Edge newsletter, which I’m told has been helpful in a small way to the labor movement.

But I am also really into coding, mostly as a hobby and as something I can use for my various other projects, but not as something I have ever wanted to make a specific career out of. I used to be a lot better at it. When I was a teenager, I got really into Linux and Bash scripting and even contributed some system scripts to a couple of Linux distributions (my most lasting contribution was this rankmirrors script which is still used by the Archlinux package manager). Later on, I got very interested in Python, especially when I wanted to learn statistical programming. I can still do both reasonably well, but, as the saying goes, if you don’t use it, you lose it, and that has happened to some extent with me as I spend most of my work time on legal work and public writing and most of my free time on my kids.

For various reasons — some understandable, others not — there is a lot of reflexive opposition to AI/LLMs on the left. But even though I am of the left, this has not been my reflexive response to the technology, perhaps because of my hobbyist coding background. Instead, when it was first introduced a few years ago, I was very intrigued by it. As it has gotten better, that intrigue has only grown. What generative AI does and how it does it is genuinely amazing, and I have found so many ways to make use of it, both to speed up work I already do and to enable me to do work I didn’t previously have the time or ability to do.

What I Do

For example, I have not written any statistical programming code in many months and I don’t see why I would ever write any such code ever again. I know how to do it, but I don’t need to. Instead, I can take my extensive knowledge about the various public microdata sets and simply prompt an LLM harness, like Claude Code, to do it for me, after which I can eyeball it to make sure it looks right. This is not complicated coding by any means, but it reduces an hour-long task to a few seconds.

For another example, my NLRB Edge publication tracks all new documents put out by the agency using scrapers and the NLRB Research database that I used AI coding assistants to help construct (I did a lot of this before those assistants got as good as they are now, but even in their more error-prone days, they were helpful). Those new documents are then summarized by an LLM and posted to the NLRB Edge newsletter, which has 13,000 subscribers.

More recently, I spent a lot of time with Claude Code and the Opus 4.5 model working through how to reliably grab down election, case, and docket data from the NLRB website, which is not straightforward because the NLRB website is a mess and there are a bunch of frustrating technical problems with the parts of the website that serve up this kind of information. Claude Code also struggled to figure out solutions to these problems. At one point, it even sharply concluded that the NLRB had done “sloppy database work,” which is the meanest I have ever seen it be. But by bouncing ideas back and forth with Opus 4.5 and having it quickly write code to test the ideas, eventually I was able to figure out how to produce the clean, comprehensive, self-updating datasets I wanted.

For legal research, the LLMs have proven less helpful than I once thought they would be because they all have limited “context windows,” which refers to how much material you can feed it to help it answer your prompt. But even there, I have figured out ways to make it useful by narrowing the legal documents I want to get information from through a more conventional search first and then feeding only those matches to the LLM (this is called RAG). Another, even more useful path has been to take a large chunk of cases where I think the information might be and have an LLM go through them one at a time and summarize them, and then feed all of the summaries back to the LLM and prompt it from there (this is called Map/Reduce). I’ve used this approach to generate a legal reference book about the NLRB that summarizes the most-cited points of NLRB case law (to be released soon).

Even in my free time with my kids, it has provided amusement in the form of making simple javascript games with them, which we have posted at GamesAxolotl.com.

Skepticism

The skepticism I’ve read about AI/LLMs comes in various forms and it’s important not to lump it all together. These forms are listed below:

1. Skepticism of the technology itself. This is the most indefensible type of skepticism, but also the least concerning. If something like this actually works, it will prove itself because people will make use of it. Of course, they already are making use of it, but it’s early days, and there are many sectors that are unaffected. Sometimes, this skepticism comes along with examples of it not working. But everything has an error rate and things only work if you use them correctly. On this point, I think it’s useful to distinguish between people who just prompt general-knowledge chatbots without providing any of their own material (“context”) to work with and people who prompt the LLMs along with context. A lot of regular people have only ever experienced the former use of LLMs, but that is the most error prone. Professional applications of LLMs all use context. Indeed much of the design of those applications is dominated by context management questions.
   
2. Skepticism of the valuation of the technology. This is what most of the “AI bubble” discourse is about. This particular skepticism can be fueled by (1), but is logically distinct from it. After all, it could be that the technology works and is really useful, but also that financial markets think it is 10x more valuable than it actually is or will be. This seems like very reasonable skepticism to me, but also not that interesting and hard to really know. Overvaluation of companies and sectors happens all the time. If you can spot it and also time the correction, you can make a lot of money shorting it. But that’s easier said than done.
   
3. Skepticism of the distributive effects of the technology. Necessarily a labor-saving technology like this will result in the reallocation of labor factors (i.e. people having to change their jobs, which also involves unemployment spells). The companies that win the scramble for market share will end up minting many billionaires and the creation of extremely rich industrialists can have hugely negative impacts on American society and politics, as we’ve seen. These concerns also seem quite reasonable to me, but I see them as valid critiques of the capitalist system, not of LLMs. It’s a great case for socialism, but not really a case against any particular labor-saving technology. I’ve written extensively about the fact that capitalism distributes the gains from economic production and innovation in a completely insane and indefensible way, but that’s a problem with capitalism not with economic production and innovation per se.

One of the ironies of the left-tinted skepticism of AI is that it seems in part to be related to dislike of the tech sector, which also is very reasonable. But the sector that is going to be most shook up by this is the tech sector. The labor that is being replaced is tech labor, coding labor. Whether this results in a total decimation of tech head counts (same output but now done by LLMs) or in an expansion of tech output (more output enabled by LLMs) is hard to say at this point. But if you don’t like the tech sector, a new technology that makes it so that coding is trivially easy seems like exactly the thing you might invent to stick it to them!

For me, ultimately what I like is the technology, not any particular story I have about what its ultimate distributive results will be. It reminds me of what it was like when I first figured out that you can use a computer, not just to run other people’s software, but to write your own, to have it do whatever you want to do. It reminds me of what it was like when I first got into open source software and the free software movement where you could read any code you wanted and therefore, with enough time, figure out how to do whatever you wanted. This is all of that on steroids and it’s very nice to have a tool that enables you to do a lot more of what you want to do than you could before.

Pointing out this amusing contradiction piqued Geoff who is apparently the managing editor of Compact and has a bunch of unrelated grievances about higher education. Here is a sampling of the incredibly stupid exchange that followed:

From an argument perspective, this should be a very humiliating exchange for Geoff.

He initially tries to argue that exporting 600,000 units of higher education only creates administrative jobs, which is both obviously untrue but also beside the point: even if it only created those kinds of jobs, it would still be the case that these education exports balance trade, promote industry, and create middle class jobs, which is what Mottl says MAGA should focus on.

Then he complains that America is price-gouging the Chinese by charging them high tuition only to later claim he never said that. He did say that (the tweet remains undeleted), but, more importantly, as with the first argument, he seems totally unaware of the irrelevance of this point. Charging foreigners high prices for our exports promotes the “America First” principle of “balancing trade,” which is what Mottl says we should focus on.

Throughout the exchange, Geoff clearly lacks an understanding of what industrial policy is as well as the mental precision to distinguish between the claim “American governments provide subsidies to American universities” and the claim “American governments subsidize the tuition of foreign students.” This puts him in the hilarious position of somehow arguing that we both gouge Chinese students and subsidize them.

Although this was my first introduction to Geoff, I’ve actually met dozens of Geoffs in my life, especially when I used to participate in competitive debate. Geoff is a particular type of person who does not understand how arguments work, does not realize that he lacks this ability, and so winds up spinning his wheels saying things that sort of look like arguments but that are actually just adjacent claims about some of the subjects being argued about. He’s the kind of guy who hangs on at the bottom-rung of academia or the debate circuit who cannot ever figure out why nobody else seems to think his stuff is any good. He can’t see how what he is doing differs from what others are doing because he doesn’t understand arguments have an internal logic, that when you make them you are constructing a system of claims that has to hang together, not just spewing random, disconnected thoughts that express your feelings about something. When he does occasionally write something that works as an argument (I assume this happens), it is like when ChatGPT writes something good: it’s a simulation that happened to hit, not the result of achieving the actual human mental state of understanding.

Bernie Bro Elegy

First impressions can sometimes be misleading. But, unfortunately, my second impression of Geoff — formed after reading a piece called “Bernie Bro Elegy” that was sent to me by several people — is much the same as the first. Geoff is a man who is so unable to understand argument that he can only comprehend people making them as either supporting or opposing the specific players that the argument is about. The idea that someone might have a specific argumentative line they are pursuing, one that hangs together with the available facts and is consistent with a broader theory, is totally lost on him. For Geoff, if you don’t like something or someone, then you will support all negative arguments about it or them, regardless of how flimsy or contradictory they are. And if you don’t do that, then that means you do like it or them or that you have some other kind of nefarious motivation that can be deduced through speculative psychoanalysis.

Geoff thinks everyone is Geoff.

The jumping off point for Geoff’s “Bernie Bro Elegy” is that Jacob Savage wrote a piece at Compact called “The Lost Generation” in which he argued that, starting in 2014, 30-something white male professionals began losing out on a lot of economic opportunities because “DEI became institutionalized across American life.” Savage’s argument is that this spike in affirmative action happened in the media (Section I), academia (Sections II and III), and “Everywhere Else” (Section IV). Savage’s piece also includes a theory as to why aggregate statistics may understate the extent of the change that took place, which is that the DEI-branded affirmative action operated to prevent new white male entrants into PMC jobs but did not clear out the white male incumbents.

As Geoff writes, this piece went “mega-viral” and even got picked up by Ross Douthat at the New York Times. Through these viral mechanisms, it got to me.

When I read the piece, I thought it was very interesting because it makes very clear claims with specific time frames and age ranges that can be checked with Census data. The author attempted to support his theory with a handful of examples from specific employers and then speculated that this was happening more generally. But we don’t need to speculate about whether it was happening more generally because we have data that lets us look at the experience of a representative sample of 30-something white male professionals from 2014 to 2024.

I was curious about what that data said and so I spent thirty minutes or so grabbing it from IPUMS and writing some code to get the answer. I then posted the results to People’s Policy Project with a few paragraphs of analysis. It turns out that the Census data does not really back up Savage’s theory: 30-something white males over this period increased their Bachelor’s and post-Bachelor’s degree attainment and employment, while holding steady in the arts/media sector and in the upper echelons of the personal earnings distribution. Thus it appears that, though affirmative action disfavoring 30-something white males probably ramped up in certain areas (like Hollywood writing rooms, the example Savage starts with), it did not do so across the labor market generally.

When I was sent Geoff’s “Bernie Bro Elegy,” I thought he might have an actual critique of my piece. Maybe he spotted an error in my calculations or in my brief commentary on them. But he did not. Instead he tries to come up with a theory to explain why I wrote it in the first place.

His theory starts by stating that People’s Policy Project and The Bruenigs podcast (I’d also add NLRB Edge and my legal practice, which is actually my main job) exist “outside of mainstream legacy institutions” and have “avoided all institutionalization.” This is why I am “doing fine,” because these things are “independent enterprises that avoided the dynamics that overtook elite left-of-center institutions during the 2010s.”

Given all of this, he says it is a puzzle why I am (A) “devoting [my] efforts to debunking Savage’s criticisms of the legacy institutions.” But he has figured out the solution to this puzzle which is that I am (B) “reliant on subscriber bases that are mostly aligned with the politics of those institutions” and therefore (C) “if [I] took too forceful of a stance against identitarianism, [I] would alienate many of [my] subscribers.”

Every single bit of this is false.

My piece was focused on Savage’s claim that widespread affirmative action had significantly hurt the economic prospects of white male millennial professionals not his “criticisms of the legacy institutions.” I wasn’t trying to debunk his claim. I was trying to see if the Census data supported it. It turns out that it doesn’t, but I did not know that when I wrote the code.

Writing responses to popular economic pieces that are wrong is a lot of what I do at People’s Policy Project. The piece I wrote just before this one was a response to Michael Green’s viral article in The Free Press in which he argues that the real poverty line is $140,000. As with the response to Savage, this piece is driven by my analysis of Census microdata. The two pieces I wrote just prior to that one was a response to Jacobin articles that argued that Nordic countries rely mostly on market income compression, not the welfare state, to achieve low inequality. Jacobin is probably the closest thing to “my audience” as there is (I currently have 203 pieces on the Jacobin website and have spoken at multiple Jacobin-sponsored events) and yet here I am criticizing one of their writers, not once, but twice, including with a piece I headlined “Do Predistribution People Know How to Read?” This is what I do.

I actually don’t specifically know who my audience is (they are all anonymized on the various crowdsourcing platforms). But I definitely don’t play to any particular audience. This is not even my main job. I am a lawyer. But if I had to guess, I’d say that my audience likes critiques of identitarianism. The reason I would guess this is because I have been an outspoken critic of it for over a decade and am kind of known for doing it. Here are just a few of my earliest pieces on it:

1. 2012: Identitarianism’s class problem
2. 2013: Class and intersectionality
3. 2013: Identitarianism and the Working Class
4. 2013: Liberals and Class
5. 2013: What does identitarian deference require?

The last piece is actually among the most shared things I’ve ever written and is still frequently shared today. I’ve continued to write about this sort of stuff into the present. For example, five years ago I wrote a piece titled “Identitarian Deference Continues to Roil Liberalism” where I talked about how identitarianism is being cynically used to fire people. I’ve produced piece after piece highlighting the centrality of class over race as the driver of inequality in the US. In January of this year, I even wrote a piece specifically about DEI titled “Corporate DEI Was a Mess” in which I argued that DEI is empty HR nonsense and that only unions can offer real checks on employer abuses. Again, this is what I do.

Sidelining Class Politics

There is not much else to Geoff’s piece, but he does comically end the piece this way:

Writers like Adolph Reed and Jennifer Pan have long argued that the function of the DEI regime is to neutralize any serious broad-based challenge to the current political and economic order. That socialists remain so reluctant to criticize that regime is evidence of its continued success on that front.

I am very familiar with Reed’s arguments on this subject, in part because Reed frequently cites my work when he makes those arguments. Here he is citing me in a piece he wrote with Walter Benn Michaels (of The Trouble With Diversity fame) in 2020 and here he is citing me in an interview he gave to JSTOR in 2021. Reed and I have the same position! Don’t take my word for it. Just ask him!

Ultimately, because of Geoff’s struggle with argumentation, he cannot understand that it is possible for DEI to distract from or neutralize class-based politics while also not actually being implemented in the form of broad-based affirmative action in education and employment. That is a common argument about DEI from the left, that politicians (like Clinton) and institutions (like Goldman Sachs) make a big show about how they care about it, but don’t actually do anything about it. According to this argument, DEI is basically one big land acknowledgement: a performative gesture that does not actually change anything. In this telling, “DEI” often operates like a cynical black hole that swallows a lot of well-meaning political participation.

If this is what you believe, which is what I think the weight of the evidence supports, then Savage’s argument must be wrong. It can’t be that (1) DEI was cynical corporate nonsense and that (2) DEI was a widespread affirmative action program sidelining millions of 30-something white men. Geoff cannot understand this because he just scans (1) and (2) as saying negative things about DEI and thinks that if you don’t like DEI them you must support both (1) and (2) without regard for whether they are contradictory or false. And because he cannot grasp that actually a good argument can’t say both (1) and (2), he cannot understand what someone is doing when they pick one or the other but not both. So, he’s left to try to make sense of the world with the dumbest possible psychoanalysis of socialists while he is in fact engaged in the project of sidelining class politics in favor of DEI gibberish.

I wish I could say he was doing so because he is captured by his audience of Compact-reading midwits. But really it appears that he is himself a midwit.

As things go on X, the back-and-forth on the topic has featured the usual mix of confusion, stupidity, and deliberate misreading. Clarification does not ever really help such matters because most posting on X seeks in-group approval not accuracy in any ordinary sense. But I do really like this topic and am happy to have an opportunity to discuss it again, especially as I have developed slightly different ways of talking about it in the last decade.

As you might imagine, there are quite a few competing theories of distributive justice, including:

1. Desert — A just distribution is one that distributes to each person an amount commensurate with their contribution.
2. Utilitarianism — A just distribution is one that maximizes the population’s aggregate utility, i.e. happiness or well-being.
3. Egalitarianism — A just distribution is one that maximizes equality or the condition of the worse off.
4. Voluntarism — A just distribution is one that results from voluntary economic processes.
5. Democracy — A just distribution is one that results from democratically legitimate laws.

There are more and these are simplified groupings and descriptions, but they are good-enough for our purposes here.

In this list, the first three theories focus on the distributive result to determine whether justice has been achieved while the last two theories focus on the process that generated the distributive result.

As pro-capitalist philosopher Chris Frieman noted in his contribution to the X discussion, “virtually all defenses of capitalism made by economists and political philosophers are rooted in efficiency or rights, not desert theory.” By “efficiency,” he is presumably referencing utilitarianism and by “rights,” he is either referencing voluntarism or perhaps natural rights theories.

From my survey of the writings on this topic, Frieman is correct. Virtually no pro-capitalist philosophers base their arguments in appeals to desert, especially not these days. But in my experience, the majority of non-philosophers do so. Indeed, one of the pieces I linked above was in response to Noah Smith doing so and the discussion on X was full of people attempting desert-based justifications for capitalist income distributions.

Desert Theory

To argue that capitalist distributions correspond to desert, an individual needs to do two things:

1. Articulate what kinds of contributions or characteristics make one deserving. This is sometimes called the “desert base.”
2. Show that capitalist distributions are patterned such that each person’s distributive share aligns with their share of the desert base as you define it. People who have more of the desert base (which I will describe as having “more desertils”) should receive a greater distribution than those who have less of the desert base (“less desertils”). Likewise, people who have the same amount of the desert base (“same desertils”) should receive the same distribution.

The most straightforward way to construct such an argument is to start by looking at the capitalist distributive result and then work backwards to find some way of defining a desert base that corresponds with that distributive result. If you can achieve that, others can still critique your position by arguing that you have chosen the wrong desert base or by rejecting desert theory altogether. But you will at least have gotten an argument off the ground. If you can’t even find a desert base that matches the pattern of capitalist distributions, then your position is dead on arrival.

So what could the desert base for capitalist distributions actually be? Can someone articulate one that actually works?

When it comes to the capitalist distribution of labor income among laborers, there is a plausible-enough desert base: personal productivity. More productive workers receive more labor income. Less productive workers receive less labor income. Similarly productive workers receive similar labor income. There are some possible objections to this account of things of course, but the argument at least gets off the ground.

Although personal productivity is a plausible-enough desert base for labor income, it fails entirely for capital income. This is because, as Joan Robinson put best, even if we want to say that capital is itself productive, “owning capital is not a productive activity.” This is the point I was getting at in my X post that set off the recent discussion, where I mused that “blind trusts, where the owner gives money to a fund manager and the fund manager invests it without telling the owner where they invest it, is like a thought experiment you would construct to tease out whether capitalists actually do anything for their capital income.”

Of course, the non-productiveness of owning capital is not exclusive to blind trusts. The interests, dividends, rents, and capital gains that flow to owners have nothing to do with any work or productivity they are contributing. This is why it is possible for a stream of capital income to be received by someone in a coma and even by someone who is dead through an estate. This is why it is possible for capital income to be received by entities that are not even humans, such as foundations and sovereign wealth funds. This is why it is possible for capital income streams to be shifted from one person to another via transferring of assets, such as through inheritance. None of this can be said of labor, laborers, or labor income.

Given that the personal productivity desert base does not fit with capital income, a desertist argument for capitalism has to either ditch personal productivity for some other desert base or articulate a desert base that has additional components.

One way of expanding the desert base in attempt to also justify capital income is to add “undertaking risk” to it. This is what Noah Smith attempted to do in our clash a decade ago and what almost all of the discussion on X was about.

Although “undertaking risk” is something you could describe capital owners as doing in general, it is not the case that the capital income distribution matches this desert base. I explained this well in my prior piece, but for novelty sake, I will give a different presentation of this point here.

Imagine three people — Persons A, B, and C — all of whom are identical in all relevant ways except that Persons A and B have undertaken risk by purchasing and owning the same amount of equally risky financial assets while Person C opted to put the same amount of money under the mattress. Using the language of “desertils” discussed above, we could say something like Persons A and B both have 100 desertils owing to their undertaking of risk while Person C has 0 desertils.

If capitalist distributions were being done according to this “undertaking risk” desert base, then Persons A and B should receive the same distribution because they both have 100 desertils. Also, Persons A and B should receive a greater distribution than Person C who has 0 desertils. Instead, this happens:

1. Person A’s investment works out and they receive $1,000.
2. Person B’s investment fails and they lose $100.
3. Person C neither gains nor loses any money.

This outcome fails to comply with the requirements of desert. Person A got a greater distribution than Person B despite each having 100 desertils. Even Person C got a greater distribution than Person B despite the fact that Person C had 0 desertils and Person B had 100 desertils.

In the X discussion about this analysis, most of the negative reactions made one or both of the following points:

1. Risk, by definition, results in people who undertake identical risk receiving different rewards.
2. If we were to make it so that people who undertook identical risk received identical rewards, capital markets and capitalism would not work.

To this I respond:

1. Exactly. The way compensating for risk works is inherently incompatible with any desert theory.
2. Exactly. The way capital markets and capitalism works is incompatible with any desert theory.

What is happening with a lot of these responses is that those making them are responding to excerpts of the piece without understanding how those excerpts function in my overall argument about capitalism and desert. In other cases, what’s likely going on is that people are simply too stupid to follow any kind of philosophical argument and therefore cannot distinguish between the claims “a desert theory that uses undertaking-risk as the desert base would have to compensate all risk-takers equally” and “Matt Bruenig thinks our economic system should compensate all risk-takers equally.” For these latter people, I am not sure if informing them that I do not believe in desert theory would make it easier for them to grasp the difference between these claims or just make their heads spin further.

The other substantive reactions to the argument typically just shifted into justifying capital income using one of the other distributive justice theories, most commonly utilitarianism and voluntarism. I have things to say about the compatibility of capitalism with those other theories of distributive justice as well, but for our purposes here, it suffices to say that having a utilitarian or voluntarist justification for capital income is not the same thing as having a desertist justification for it. I am merely pointing out that that capitalism is incompatible with desert theory.

I’ve been thinking about this in the context of The Argument, a new publication that debuted last week that intends to, among other things, have real people argue real things directly against one another. How will people consume these arguments or make sense of them in a discursive environment devoid of generally-accepted notions of what makes an argument good or successful?

The overwhelming response to the first head-to-head clash — between myself and Kelsey Piper over the importance of cash welfare — provided a good opportunity to answer this question. This response proved the theory that this is a potentially very entertaining and engaging form of content. But it has also confirmed that there is no consensus at all on how to actually evaluate the argument that everyone just read.

Rich Yeselson read the exchange and concluded that it amounted to “Matt Bruenig eviscerat[ing] a hapless lib out of the gate” while Logan Bowers read it and found that he “agreed more and more with Piper after each successive [Bruenig] paragraph.” These are just two tweets, but based on my scan, they are reasonably representative of the thousands of other tweets reacting to the debate.

If you think there is something to evaluating argument itself, not just agreeing or disagreeing with conclusions or people, there is something really strange about the inability to reach some kind of consensus about the exchange. It should be possible to say this or that person got the best of it in a way that is mostly detached from your own views. But to do this well requires the ability and willingness to technically dissect an argument, not just impressionistically consume it.

When I looked at Kelsey’s first piece from a pure argumentation perspective, what I saw were two moves that go as follows:

1. She contends that cash welfare does not really help much. She presents a few recent studies showing null results for cognitive and health outcomes. She doesn’t present an explicit framework for evaluating whether a particular welfare policy is good, but implicitly adopts an evaluative framework that says welfare programs can be deemed good or bad by looking at the extent to which they promote human capital and related indicators.
   
2. She presents an error theory to explain why leftists, liberals, progressives and so on are not reacting much to these studies despite the fact that they are, based on her implicit evaluative framework (human capital promotion), devastating to the pro-cash-welfare argument. The error theory is essentially that they are dishonest actors who refuse to acknowledge inconvenient facts.

My response was specifically designed to clash with Piper’s on every argumentative level:

1. I make what debaters would call a kritik in which I challenge Piper’s implied framework for evaluating the success or failure of welfare programs. Rather than looking to see whether a welfare program promotes human capital (Piper’s preferred framework), I argue that we should look to the more traditional goals of the welfare state: eradicating class difference and social alienation, reducing inequality and leveling living standards, compressing and smoothing income and consumption, providing workers and individuals independence and refuge from coercion by reducing economic dependence on the labor market and the family, among other things. To make it clear that this is not some ad-hoc framework I am constructing to reach a preferred conclusion, I cite other authorities on the welfare state that also use it, including perhaps the most celebrated theoretical work on western welfare states,The Three Worlds of Welfare Capitalism.
   
2. I present an error theory to explain why Piper failed to account for these other goals in her argument. The error theory is that (a) she is unfamiliar with them, (b) she has unreflectively adopted recent ideas about the (mostly child-focused) welfare state that present it as an “investment” in human capital, (c) she is drawing upon an effective altruist idea to use direct cash transfers as an economic development strategy in the third world and mistakenly applying that logic to the developed-world welfare state in America.
   
3. I directly respond to Piper’s empirical claims by pointing out that there are contrary studies that do find positive results on the outcomes she focuses on. I anticipate her argument that those are “lower quality” because they aren’t RCTs by pointing to an RCT that was even higher quality than the ones she relied on.
   
4. I directly respond to Piper’s error theory by attacking the premise that nobody is acknowledging these studies by pointing out that the New York Times covered it. The kritik also functions as a response to the error theory: maybe it’s not that people dishonestly refuse to acknowledge the studies, but rather that the studies are just irrelevant to the main justifications for the welfare state.

In theory, if we care about evaluating arguments themselves, we could get some agreement that these were the moves and then from there get into the more contentious questions of how they were executed, who won the various clashes involved, etc. In practice, with a few exceptions, people don’t really digest arguments in this formalistic way.

One of the things I’ve found funny in the aftermath of the clash is watching people object to my argumentative style, which they describe as being mean or dismissive or personal or whatever. What these people are actually reacting to is my error theory about how Piper reached (what I regard as) an incorrect conclusion. My error theory is essentially that she is ignorant of the rationale for developed-world welfare states and has been led astray by recent liberal ideas about human capital and effective altruist ideas about third-world development policy.

The reason I find the reaction to that interesting is that Piper also spends a good deal of her piece on an error theory. This error theory is sprinkled throughout the piece but it essentially accuses others of dishonesty. Indeed, she concludes the piece by calling them liars:

We’re in a dangerous epistemic environment. One where widespread agreement on basic facts is scarce and noble lies have permeated the halls of truth-seeking organizations like the media. Those of us who care about ending poverty have to choose the integrity of our work over trying to play 5D chess and hoping no one else knows the rules.

As I said above, I designed my arguments to directly clash with Piper’s. She presented an error theory that her opponents are liars. I presented an error theory that she is ignorant. Why is my error theory offensive and hers isn’t?

There is no actual difference between the offensiveness of our error theories, but there is a psychological difference in how people process them. Piper called a general group of people liars who selectively disregard inconvenient facts while I called a specific person ignorant. Abstractly, it would seem Piper has actually done the more offensive thing because she’s said something sort of insulting about a lot of people while I have only said something sort of insulting about one person. But people don’t process it this way because they are more affected when they can personalize the insulted party than when they can’t.

In the last few years, sub-replacement fertility has made its way to the US and the fertility question has unsurprisingly made its way into the US discourse.

What I find most interesting about the fertility issue are the philosophical tensions that it brings to the surface. The ideal scenario for fertility is that individuals have whatever number of children they wish to have and this results in an aggregate fertility that is not too high and not too low. When this state of affairs obtains, fertility is not a matter of public concern, not something that could ever plausibly be seen as a legitimate subject of politics or policymaking. People will have their own personal opinions about ideal family sizes for themselves and others, just as they have other personal opinions like the value (or not) of stay-at-home parenting. But these remain personal opinions, something perhaps suited for lifestyle blogging, but not much else beyond that.

What is so troubling about a scenario where there is too-low overall fertility is that it confronts us with the reality that something we regard as a deeply personal and private matter is also, in the aggregate, very socially important. This observation is potentially destabilizing to liberal principles — personal autonomy, individual rights, and neutrality towards and acceptance of pluralistic conceptions of the good — that are dominant in American society. If we acknowledge that aggregate fertility is a matter of social concern, then some might conclude that individual fertility is something society and the public should have a say in, which is, among other things, repulsively illiberal and anti-feminist.

There are four ways to respond to this observed tension:

First, you can stake out philosophical positions that deny the tension altogether. This is Nathan Robinson’s move when he says that, even if sub-replacement fertility led to full-blown extinction, he does “not think that is a matter of moral concern.” For Robinson, the fact that the aggregate of individual fertility choices can have negative spillovers on the rest of the society does not in any way challenge the conclusion that the society has no business concerning itself with aggregate fertility. It remains entirely a matter of private, individual concern.

Second, you can stake out a factual or framing position that attempts to render the tension moot, at least for now. Perhaps aggregate fertility is not presently too high or too low. Perhaps one’s notion of aggregate fertility is too narrowly focused on particular geographies, like Japan or the United States, and instead should be focused on the globe, or should take into consideration the possibilities of migration into low-fertility geographies. Perhaps fertility has a cyclical characteristic to it that will match our current bust with a corresponding boom without having to publicly attend to it. Whatever the specific case may be, the way this move works is not by denying that there may be a philosophical tension underneath all of this, but rather by bracketing the philosophical question, and insisting that the blissful ideal scenario of things working out without having to make it a matter of social concern still obtains.

Third, you can navigate the philosophical tension in an illiberal way by adopting essentially authoritarian social and policy stances that attempt to force or shame people into having more children. You can think of this as the The Handmaid’s Tale option, even if the particular illiberal mechanisms one might pursue are different. This is obviously where much of the far right goes with the tension as they are already inclined towards rejecting pluralistic conceptions of the good and punishing ways of living that they personally dislike.

Fourth, you can navigate the philosophical tension in a liberal way by looking for policy mechanisms that increase aggregate fertility without sacrificing the ultimately private and individual nature of the millions of personal choices that make up the aggregate. This is where left-of-center (and really even most right-of-center) politicians in other countries, especially in Europe, wind up on the topic.

This last option might seem like a bit of a dodge in itself, or, at minimum, some kind of deception. How can you both respect individual choices while also enacting policy that changes them? Isn’t that a straightforward contradiction?

One way to respond to this charge is to point out that people always make decisions within the economic and social system they find themselves in. Changing that system will generate different decisions, but those decisions are no less “theirs” than the decisions they made in the prior system. There is no systemless baseline that can reveal what each person’s “real” uncorrupted decisions are. A system that relies more on out-of-pocket spending to cover children’s expenses will result in fewer births than a system that relies more on socialized spending to cover children’s expenses, but this observation does not tell us which (if either) of the two systems is liberty-respecting or liberty-destroying.

Another way to respond to this charge is by analogy. The fertility discourse tends to treat the question as if it is sui generis, but there are actually many other issues that have the same basic form as the fertility issue. The one that seems most analogous to me is the rural doctor shortage issue.

The rural doctor shortage issue is that there are rural areas in the country where there are too few doctors to provide adequate medical services. You could respond to this issue in all the four ways discussed above:

1. Philosophical Rejection. You could say it is not actually a matter of social concern because it pertains to something that should be left to individual choice. Doctors can locate wherever they’d like and it’s none of my business where they live.
2. Factual Rejection. You could say there is not really much of a medical services shortage at all in rural areas and that we have other mechanisms, like telemedicine, that allows for individuals living elsewhere to fill the gap.
3. Illiberal Solution. You could attempt to use illiberal mechanisms to force doctors to live in rural areas, such as by requiring medical school graduates to enter a draft where some are selected to work in rural areas and punished with prison if they do not do so. This sort of thing would finally make some sense of the claims made by libertarians like Rand Paul that a right to healthcare implies a corresponding ability to enslave.
4. Liberal Solution. You could attempt to use liberal mechanisms that result in more doctors deciding to work in rural areas but without requiring any of them to do so. This is in fact what we have done with initiatives like the National Health Service Corps that provides financial incentives for medical professionals to work in underserved areas. This type of thing is not understood by anyone as being repulsive or liberty-destroying.

Of course, the rural doctor shortage is not totally analogous to the fertility question as fertility decisions are perceived as being much more personal, private and individual than where in the country one decides to ply their craft. This difference explains why the fertility issue generates much more intensity of feeling, but it does not actually change any of the basic calculations involved.

Based on this above analysis, it is easy to see why the fertility issue is one that could favor the economic left. The most liberty-respecting policies that increase fertility are the left-wing ones that socialize the cost of children by enacting things like paid parental leave, a monthly cash benefit for kids, and free child care, pre-k, education, and health care. These benefits also happen to be favored by the left for totally other reasons rooted in opposition to inequality and poverty. Indeed, there is such a consensus in favor of the benefits that when people like Nathan Robinson go out of their way to explain that they philosophically reject the idea that fertility can ever be a matter of social concern, they still nevertheless feel the need to indicate that they support all these policies anyways, just for other reasons.

Of the the four types of responses identified above, it is ironically the pronatalist right where you will find the most hostility towards these proposals, not because they don’t realize that they will, at the margin, increase fertility, but because, for many of them, their commitment to market income distributions takes precedence over their pronatalism. Pushing family benefits would thus have the added benefit of revealing how full of shit they really are.

Despite all of this, I suspect that this issue will, if it has not already, fall prey to a dynamic I call overlapping dissensus.

In Political Liberalism, John Rawls explicates a concept he calls “overlapping consensus,” which is meant to address the question of how a diverse, pluralistic society with many divergent moral, religious, and philosophical worldviews can nonetheless operate as a liberal democratic society. Rawls’s answer is that people with divergent underlying worldviews and motivations can still develop an overlapping consensus in favor of certain principles (or, I’d add, policy goals). This overlapping consensus allows for governing majorities and relatively peaceful governance to be established because it manages to achieve agreement over governance-relevant things without requiring agreement on more fundamental questions.

Put simply, if you and I both support principle or policy goal A, it doesn’t matter that I support it because of worldview B and you support it because of worldview C, even if the worldviews B and C are totally at odds and repulsive to one another.

As elegant as Rawls is on this point, I have found that in a lot of public discourse at least, things often work out in exactly the opposite way. When people see that a person, reason, or worldview that they find repulsive is in favor of a particular principle or policy goal, they will oppose that goal purely because they do not wish to overlap with that person, reason, or worldview. If their own pre-existing worldview seems to favor the same policy goal but for different reasons, they will respond by altering their worldview (including by inserting a contradiction into it) rather than accepting that they are part of an overlapping consensus with a social grouping they despise.

Thus, overlapping does not necessarily result in a plurality-respecting consensus but often in deeper and more vicious dissensus that actually makes policymaking and governance harder, not easier.

Chronology

Yglesias’s first critique is about the chronology of it. NAFTA was enacted in the 1990s while this most recent political realignment did not really crystallize until 2016:

It is unquestionably true that NAFTA was very unpopular in the industrial Midwest and that it continues to have a very bad reputation there. But it’s not possible that it was more salient in 2024 than it was a quarter century ago. This doesn’t explain anything about why Kamala Harris did so much worse in the industrial Midwest than Gore did.

The counter to this is that there was an intervening event that made NAFTA, a term that is used to refer to trade liberalization and offshoring generally, especially salient in a way that promoted realignment. That intervening event was a Republican candidate for president, Donald Trump, who made it one of his two main issues. Prior to that point, Republicans were supporters of NAFTA and free trade. What opposition to it that did exist was on the Democratic side of things, though this was a minority position among elected Democrats.

I don’t think it’s at all inconceivable that dislike of trade and offshoring was a simmering aspect of working class politics that only became politically actionable in the form of party realignment when Trump came along.

Gore’s opponent was George W. Bush who expanded the number of countries the US had free trade agreements with from three to sixteen. Obama’s first opponent was John McCain who the UFCW characterized in 2008 at someone who “consistently voted for unfair trade agreements, including the North American Free Trade Agreement (NAFTA) and the Central American Free Trade Agreement (CAFTA), and supported China’s entry into the World Trade Organization despite its ongoing history of human rights and workers’ rights violations.” Obama’s second opponent was Mitt Romney who released the following campaign position on trade in 2011:

It shouldn’t be surprising that none of these Republican candidates were able to electorally capitalize on anti-trade sentiments in the way that Trump did.

Narrow Impact

Yglesias’s second critique is that the negative impact of NAFTA (and trade liberalization more generally) was narrowly contained to a specific region of the country while the geography of the political realignment has been much broader.

The alternative explanation of why working class voters have swung against Democrats — they disagreed with 21st-century cultural liberalism — has much broader and deeper explanatory power. Firefighters on Staten Island didn’t have their wages undercut by imports from Mexico. Neither did retirees on the Gulf Coast in Florida. The localized economic impact of NAFTA on union manufacturing workers and factory towns is a very good explanation for why elected officials from those areas were against NAFTA, but the political trends are much too widespread to be explained by something so specific.

What’s interesting about this move is that Yglesias decides to compare the NAFTA explanation with an alternative cultural-liberalism explanation, concluding that the latter is broader than the former. Then, when explaining what he means by saying NAFTA is narrow, he offers the criteria of whether someone was directly impacted by it.

But what happens if we apply this directly-impacted criteria to cultural liberalism? How many firefighters on Staten Island or retirees on the Gulf Coast have actually been directly impacted by transgender athletes? Is voter opposition to immigration the strongest in areas where a lot of illegal immigrants live or areas where not very many live?

What Yglesias must recognize with these cultural issues is that various constituencies can develop strong opinions about them separate from whether they have any actual impact on their own lives. But something similar is also possible with NAFTA. Working class constituencies across the country can end up having strong opinions about trade liberalization and offshoring that motivate their electoral behavior even if they personally were spared the negative impacts.

In fact, if you listen to certain people talk about NAFTA, trade liberalization, and offshoring, you’ll actually hear a discourse that sounds a lot like the immigration discourse, which contains a mix of economic concerns (foreigners taking jobs and pushing down wage levels) alongside more cultural appeals to nationalism, nativism, and xenophobia. The foreign aid discourse looks that way too.

In this sense, trade liberalization is part of 21st century cultural liberalism, i.e. just another way in which liberal cultural sensibilities have them prioritizing foreigners over the native working class. Indeed, this is exactly the rhetorical synthesis that people like Trump and Vance have pursued about foreign aid, trade, and immigration, which in their telling are not just economic issues but also about the sick priorities of bleeding-heart liberals.

Strategic Considerations

With that all said, it’s not at all clear to me what the strategic way forward is. Just because you can tell a history where X caused Y, that does not mean that, going forward, not-X will reverse Y. And of course, as Yglesias points out, there are real tensions at play here between the goals of onshoring through trade deliberalization, counteracting monopoly and concentration through competition, and keeping prices low.

Per my usual bit, I’ll end here by just pointing out that if you install egalitarian economic institutions that result in a compressed distribution of income and wealth, the precise regional and sectoral composition of the economy becomes much less important. During trade liberalization, US output per capita did not decrease. In fact, it increased. It is the refusal to use economic institutions that ensure that these increases are broadly distributed that is ultimately the problem.

My System

I got a 7.56 kW solar panel system consisting of 21 LONGi panels and 1 Solar Edge inverter. The unsubsidized all-in price for the system, including labor, was $25,525. After applying a federal tax credit equal to 26 percent of that price, the price fell to $18,889. Initially, I planned to get a loan to finance the system and the payments for that loan would have been $129 per month.

As discussed in the prior piece, due to various timing and contractual snafus, I ended up negotiating down the price a further $5,525 and then purchased the system on my cash back credit card (which I then paid off), saving me a further $400. So my all-in price was $12,963. But this was an unusual situation, so in the calculations below, I am mostly going to compare the amount of money I saved to the $129/mo solar loan.

Solar Production and Utility Bill Savings

In their first three years of operation, the solar panels produced 32,266 kWh of electricity, which is an average of 10,755 kWh per year. As illustrated in the graph below, production is higher in summer months and lower in winter months.

Over this same period, the electricity rates for my area were as follows.

Because my state does net metering, I can calculate utility bill savings by multiplying solar production by the electricity rates. I do this in the graph below. The dashed grey line is the $129/mo loan payment, which is provided for comparison.

Here is the same graph but with cumulative figures rather than monthly figures.

Over this period, the solar panels saved me $8,935 on my electricity bill, which averages to $2,978 per year. Had I been making the $129/mo loan payments, the net savings would have been $4,308 over this three year period.

Had I not used the loan but instead paid the full price up front, minus the federal tax credit, then I would have paid $18,889. With $2,978 of annual savings, it would have taken a total of 6.3 years to recoup the upfront cost. With the actual amount I personally paid — $12,964 — it will take a total of 4.4 years to recoup the upfront cost. The solar panels have a 20-year lifespan.

So ultimately, this is proving to be a very wise financial decision for me. The solar panels save me money and have required no maintenance at all. In a little over a year, I will have fully recouped my costs and have over 15 years of production still ahead of me. Even if I had not lucked into my own personal arrangement and instead taken out a loan on the $18,889 price, I’d still be saving $1,436 per year on net.