Getting Historical Exchange Rates for Peso to Dollar
Pulling historical peso to dollar rates isn't as simple as you'd think. Most people grab whatever rate shows up on Google Finance and call it a day. That works for casual curiosity. It falls apart fast if you're doing anything that requires precision. Most people assume historical rate data is just a single number per day. It's not. The Mexican peso is traded in multiple formats — spot, forward, interbank — and the USD/MXN pair has both a mid-rate and a bid-ask spread that changes depending on which data source you use. The Mexican central bank, Banxico, publishes daily reference rates, but those are end-of-day snapshots. If you need intraday movement, you're looking at Bloomberg, Reuters, or OANDA's historical feed. The peso also had a dramatic shift in its exchange rate regime. Before 1994, Mexico essentially pegged the peso to the dollar within a band. The 1994 Tequila Crisis changed everything and the peso went floating. Any historical analysis that spans across that date needs to account for the structural break or your regression coefficients will be garbage.
How I Actually Get the Data
I used to rely on Yahoo Finance's download feature. It was free, it was convenient, and it was wrong more often than I cared to admit. Here's what happened last year. I pulled a CSV of daily MXN/USD rates from Yahoo going back twenty years. Everything looked fine until I cross-referenced a few dates with Banxico's official records. Three specific dates in the dataset had swapped open and close values. Not all of them. Just three out of 5,000+ rows. That kind of silent corruption is the worst possible outcome because the data looks clean. My workaround was straightforward. I started pulling the same dates from two independent sources — Banxico's API and the Federal Reserve Economic Data (FRED) series for MXNUSD. When the numbers disagreed, I defaulted to Banxico. For periods before Banxico's digital records existed, FRED's historical reconstruction was reliable enough. I automate this now with a small Python script that downloads from both feeds, compares the values, flags any discrepancies above 0.001 pesos, and writes the final dataset to a local CSV with a confidence score appended to each row. Takes about four minutes on a good connection. If you need a quick fix right now without writing code, the World Bank's database provides MXN/USD annual averages that are generally accurate for macro-level work. The IMF's IFS database has monthly data. Neither is great for high-frequency or intraday needs.
Common Mistakes People Make
The biggest issue I see is directionality confusion. The peso dollar rate can be quoted two ways — how many pesos per dollar, or how many dollars per peso. Most US-based platforms show MXN per 1 USD. European and Mexican sources sometimes flip it. If you don't check which convention a dataset uses before you start modeling, your results will be inverted and you won't catch it immediately. I've seen this happen in client work more than once. A financial model was built on inverted rates and nobody noticed until someone actually tried to convert a number and the output was off by a factor of roughly eighteen. Another subtle problem is survivorship bias in rate history. Many online platforms only show current tradable pairs. If you're looking at historical data for the Guatemala quetzal, the Nicaraguan córdoba, or smaller Central American currencies paired against the peso, the data quality drops off sharply. Some of these pairs didn't exist on certain platforms during earlier periods. You'll hit gaps that look like missing data but are actually just never-recorded data.
Get the Full Details
When This Approach Doesn't Work
The two-source comparison method I described breaks down during periods of extreme volatility or capital controls. During the March 2020 crash, the spread between interbank rates and retail rates for MXN/USD widened to levels that made any single reference rate misleading. If you were hedging a position and used only the Banxico reference rate, you could have been off by several percent from what a actual bank would have charged you on the same day. There's no clean workaround for that except pulling actual transaction-level data from a broker or bank statement, which most individual researchers don't have access to. For pre-1990s data, electronic sources are thin. Physical archives of the Banco de Mexico's daily bulletins are the most reliable source for that era, but they require visiting in person or requesting scans. The data exists. Accessing it is the bottleneck. The Banxico API itself has a limit of roughly 10,000 requests per hour. If you're building a large dataset that requires repeated queries, you'll get blocked. I learned this the hard way and now I cache everything locally with a daily refresh rather than making live calls for every run.
I generally point people toward the FRED API for US-side data and the Banxico XML or JSON endpoints for Mexico-side data. Both are free. Both require an API key. Both are reasonably well-maintained. Between those two, you can reconstruct nearly a complete history of the peso-dollar relationship with minimal friction. Just make sure you're reading the documentation carefully before you start coding. The parameter names are not intuitive and the date format expectations vary between the two systems.
