Trying to get a sense of the following ...
In some emerging markets such as Argentina, there isn't any observable IRS swap curve, but only Xccy. I noticed in the place I work that FX NDF are used to construct the xccy curve.
I don't understand the rational of using NDF for the xccy curve... Does anyone have any info on xccy curve construction? or good bibliography
In Argentina (and a few other emerging markets), a cross-currency swap is somewhat liquid (much less so than in was before the most recent sovereign default).
You can find someone to trade 2 year fixed ARS for floating USD (LIBOR; will probably be SOFR soon).
2 years ago you could easily find someone for 5 year fixed ARS for floating USD (LIBOR).
The ARS fixed rate for such swaps is observabe (in interdealer brokers runs if not always Bloomberg terminal). But you can't decompose it into onshore swap rate plus cross-currency basis the way you can in most markets.
Argentinians (and many other emerging markets) just don't trade fixed-for-float ARS swaps. There are some floater bonds that reset from a rate called BADLAR. But no one trades BADLAR swaps. I tried to have a BADLAR curve, but there is not enough data in the prices of the floater bonds to get a curve.
Cross-currency swaps and FX forwards in most EM currencies, not just ARS, are usually offshore non-delivery (you observe the spot rate 2 days before each cash flow, calcuate net USD flow, and pay or receive that) and external-law rather than local-law. Only major currencies like EUR/JPY/GBP/CHF are usually physical delivery.
The rationales for offshore NDFs are: it's an operational pain to physically pay/receive non-CLS currencies. If there's a dispute, you want to deal with it in London or New York court, not local EM court. You don't want the local government in the future to restrict what you can do with the currency (cross-border risks sometimes called transferability and convertibility). Even if you're trying to hedge some cash flows where you expect to pay or receive physical EM currency, you don't quite flatten the cross-border risk with an offsetting physical settlement trade. Argentina is nothing special with respect to these rationales.
FX NDF forwards are settled in USD, rather than ARS, so they'll be more liquid than on-shore ARS forwards (if these are even traded, probably not).
Xccy swaps are traded and liquid only from a certain tenor onwards: usually 3 years or 5-years. For the shorter maturities, the NDF FX forwards would be used - that hopefully answers your question on "why to use FX forwards to build an interest rate curve".
As far as actually building the ARS rates curve goes, mathematically, FX forwards satisfy the following equation:
$$S_{USDARS}*(1+r_{ars})=(1+r_{usd})*F_{USDARS}$$
Above, $S$ is the FX spot rate, $F$ is the forward rate, whilst $r_{usd}$ & $r_{ars}$ are the respective interest rates. If you plug in the USD Libor rate for $r_{usd}$, the spot rate for $S_{USDARS}$ and the Forward rate for $F_{USDARS}$, you can solve for the ARS Libor rate term $r_{ars}$: but bear in mind that the forward $F_{USDARS}$ also contains the cross-currency basis, and that will be "hidden" in the $r_{ars}$ term that you'll be getting out as the output from the equation.
A decent book on bulding curves, including Xccy basis curves, is this one here from JM Darbyshire, but it's a bit expensive.
