DID vs. Synthetic Control: When to use which?

[bcdelodx]

I'm evaluating a state-level education policy change and trying to decide between difference-in-differences and synthetic control methods. Both seem potentially applicable, but I'm not sure which is more appropriate.

My setup:

• Treatment: One state adopted new teacher compensation policy in 2018
• Outcome: Student test scores (panel data 2010-2023)
• Data: All 50 states, annual observations
• Concern: Treated state had different pre-treatment trend from national average

My understanding:

• DID requires parallel trends assumption - but visual inspection shows my treated state trending differently
• Synthetic control can handle different pre-trends by constructing weighted counterfactual
• But synthetic control gives me just one treated unit, so wider confidence intervals?

Questions:

1. Does the different pre-trend automatically disqualify DID?
2. Is synthetic control always preferable when you have just one treated unit?
3. Can I run both and compare results, or is that p-hacking?

Would appreciate guidance on how to think through this choice systematically.