The main reason to use "they" for preverbal children is that you don't know their internal experience of gender (if they even have one) until they're able to start expressing it. Using "they" allows the child to continue without an assigned gender until they start feeling like they have one and request otherwise, or they may choose to continue with "they" indefinitely. The experience of being assigned a gender that is contrary to your internal experience is one that trans folks have written a lot about. It's clearly a strong negative, and something I don't want to subject my children to.

A secondary reason to use "they" with preverbal children is to help build a norm of using "they" for people when you don't know how they identify. Even if our particular children are cis, some other children won't be. Using "they" for our children would help make "they" a usual way to refer to children, and make things easier for the children who need it.

On the other hand, in our current world, it's hard to go by "they". This is something non-binary people have written a lot about, and the general idea is that as long as lots of people think everyone is either male or female and try to categorize you by appearance and behavior, you get misgendered a lot. There's resentment of the work of needing to track the pronouns of people around them. There's resentment of weirdness. There's overt teasing with children, blending into (usually but definitely not always) more subtle harassment as adults.

None of this should happen, and I want to build a world where it doesn't. But I don't think it's fair to make my children fight this fight if it's not their fight. I want them to grow up to be people who do work to make things better, even when it's not something that harms them directly, but I don't think I should make them do it.

Overall, I see it as unfair to use a gendered pronoun which may conflict with the child's internal experience, and I also see it as unfair to subject a child to the unfair treatment nonbinary people often receive today. Yet in today's world we have to choose.

Part of how I think about it is, how likely is my child to end up with a gender identity that doesn't match the one we would guess? If it was almost certain then "they" would be an obvious choice, while it if was almost certainly not then similarly not. The best (according to Ozy) survey on this is How many adults identify as transgender in the United-States (2016, n=151,456) which asked simply "do you consider yourself to be transgender?". They estimate 0.58% of Americans identify as "transgender", with moderately higher numbers (0.66%) among younger (18-24) adults than older ones. [1]

This is lower than my experience, but it's hard to tell how much this comes from experience in environments that trans people are more likely to join (urban areas, software engineering) vs people being more likely to identify as trans in more accepting environments. I'm going to guess 2%, but this is very rough.

One way to think about this then would be: is being raised with an inaccurate assigned gender >50x worse [2] than undergoing the unfair treatment that non-binary children currently get today? I think it probably isn't, so I wouldn't go with "they" for a baby.

It's also possible that a child raised with "they" wouldn't experience it as "the people around me are avoiding gendering me" but instead as "the people around me are gendering me as nonbinary." Since even fewer people are nonbinary, this is even more likely to be felt as misgendering. It's really hard to figure out how much of an issue this is because I don't believe any children raised with "they" are old enough to talk about their experience with it yet.

I completely support parents who weigh these factors differently and do decide to go with "they", and I hope this goes well for them and their children. Whichever approach people take I hope they are attentive, notice if their child does start indicating discomfort, and are fully supportive.

(Expanded from a comment on this post.)

[1] If you take the study claims literally, MA is in right the middle of the pack. Except they actually only have data on nineteen states and used a complex regression I don't really trust to estimate data for the other states. I suspect MA is higher than their estimate.

[2] The idea is that we're comparing a 2% chance of raising the child with an inaccurate binary gender to a 98% chance of raising them with "they" unnecessarily.