The least bad answer would be to use local-currency bond yields. But good luck trying. Local issuance in any real size is a relative novelty of the last decade or so. There's no guarantee there's always a nice clean 10 year note live for every country, like a Treasury, Bund, JGB, Gilt etc.; and no guarantee that if it did exist 20 years ago that it was a veyr liquid instrument with reliable signals that people assume from Tsys et al.
Of course, these rates will be (almost) risk-free in local currency terms; but the currency itself is hardly "risk-free" ;-) I say almost because I recall Russia defaulting on local paper in 1998. Probably a few others I've missed, but the defaults on local are definitely outliers. Hey, Germany and Japan defaulted too on the DM side of the ledger.
Whatever the theoretical niceties, using local is always better than trying to use EMBI/$ paper, for which there is a genuine credit risk. Meanwhile, trying to keep everything in dollars and matching vs Tsys/T-Bills is all well and good, but it requires you to dollarise future earnings on the equity side. Good luck forecasting with confidence where BRL, ZAR or TRY will be in 10 year's time ;-) Absent such forecasts, using currency forwards as the obvious naive default is simply equivalent to using local rates.
So the problem you'll get trying to use this data is the nominal public-vs-real private "mismatch". Many years ago, I used to be a stockmarket strategist and we built a suite of models to try to estimate the equity risk premium of different markets. For us, Italy was always the bugbear... because it was quite normal for Italian stocks to exhibit a negative risk premium for prolonged periods of time. It's hard to torture a PE of 10x against a bond yield of 14% and get any different answer. But it's very hard to explain to a client or banker, who just needs an intuitive model input they can sell to their own client.
The short answer is that the ERP was negative! The BTPs were riskless in Lira. Investors were guaranteed their coupon and principal. The yield compensated them for the inflationary risk of getting their principal back in Lira, with an uncertain purchasing power given Rome's fiscal predilictions. Italian stock did have bankruptcy risk. However, these were real assets, measured against a nominal bond yield. Tank the Lira, and their Lira value rises. The negative ERP was simply telling us that the ITL inflation/deval risks were greater than the risks to long-term real earnings. It was safer to own risky factories than riskless paper in a dangerous currency.
With ~5% real rates seen recently in the likes of Brazil, I'd be amazed if you didn't experience similar issues with applying this kind of model to EM.
If's there no way to sanitize your interest rate baseline to make it intuitive and comparable to what the textbooks teach about the US, then Ockham's Razor suggests using zero, and just looking at absolute returns. If these fluctuate as wildly as the inflationary histories of the countries in question, then your equity model is probably mis-specified in the first place.
hope this helps