That's silly. Most people with a PhD don't get academic jobs. It would be foolish not to plan for that possibility.
I don't know much about EE, but I'll talk about math. It's unlikely you would enjoy research on nonlinear wave equations but hate stochastic PDEs. If one has far better job prospects, it's silly not to focus on that one unless you are dead certain you'll win the academic tournament.
PhDs make marginal improvements on one's employment prospects vs a Masters degree.
If you're doing it for the cash don't even undertake a PhD. Go get an MSc, go get a job, save four years of your life.
If you're gambling on getting into academia, you won't get the publication record you need if aren't motivated to do the work, and you won't be motivated to do the work if you don't enjoy what you're doing. If you don't get into academia but do have a quantitative PhD you will get a job quite easily, so why worry about your particular field.
My conclusion is that there is simply no point doing a PhD unless you really enjoy the work you're doing.
...you won't get the publication record you need if aren't motivated to do the work, and you won't be motivated to do the work if you don't enjoy what you're doing.
The point I'm making is that if you are unmotivated in elliptic PDEs, you'll probably be unmotivated in control theory as well. The fact is that research is research.
You might hate experimental physics and love theory, or hate computation but love experimentation, but these distinctions between subfields (condensed matter vs photonics) that seem huge in academia are not that big in reality.
If these small distinctions can give you a big advantage 50% of the time, it's a tradeoff worth making.
whatever makes you good in software and statistics. I did a PhD in signal processing (statistical reconstruction for medical imaging) worked as a quant for 2 years, and then sofware contracting for another 2. I'm now doing some other startup stuff.
the top fields that come to mind that would be good on the EE side are machine learning, signal processing, control theory, communications/information theory. machine learning is probably the best.
and also - don't just be a matlab jocky, matlab is used heavily in industry, but if I had to do it again, i'd have learned more C++, python, and R
R and matlab are both problematic when you need to integrate with production systems, but at least R is free. Python, in my opinion is best
