Working With Argentina's Religious Demographics
Most people trying to project Argentina's religion percentages hit the same wall on day one. The INDEC census questions changed mid-decade, and the data isn't consistent across 2001, 2010, and 2022. You think you have clean time series data, then you open the actual tables and realize the self-identification categories were reshuffled. Here's what I've learned doing this properly. The baseline comes from INDEC's Censo 2022 results, specifically the "Religión o creencias" module. That survey found roughly 62.9% Catholic, up from about 59% in 2010 and 77.6% in 2001 when you count nominal affiliation. The evangelical/Protestant sector sits at approximately 16.5%, the unaffiliated or no religion group climbed to roughly 22-24% depending on whether you count agnostics separately, and the remainder breaks into Jewish (around 0.4%), Muslim, other religions, and people who refused to answer. These numbers are directionally right, but they're self-reported and self-identification in Argentina carries different weight than in Europe. Someone raised Catholic who hasn't set foot in a church since confirmation still checks "Catholic" on the census. That gap between institutional practice and statistical identity is where most projections go wrong. I built a similar model for a provincial demographic study a few years back and ran into a specific problem with intergenerational transmission. The 2010 census showed evangelicals at about 9%, and by 2022 the official figure was roughly 16%. That's a massive jump for twelve years. But when I cross-referenced with the Pew Research Center's Argentina dataset and the own-resources survey data, the apparent growth slowed dramatically. The issue was methodological. INDEC changed their sampling frame and weighting between the two census rounds, and the 2010 evangelical figure was probably understated because the prior census had undercounted non-Catholic households in urban peripheries where evangelical churches are densest. My workaround was to anchor the projection to the 2010 census for the starting point but use the Pew historical trajectory as a smoothing layer, then apply the raw 2022 figure as an upper bound rather than a precise midpoint. It made the projection less clean but significantly more honest about uncertainty.
You should also know that the provincial breakdown tells a completely different story from the national aggregate. Buenos Aires Province, the City of Buenos Aires, Córdoba, and Santa Fe account for the bulk of both population and religious change. Meanwhile, provinces like Formosa, Chaco, and Corrientes still show Catholic dominance above 80%. A national projection masks regional volatility. If your use case requires subnational granularity, don't average the country. Model the five or six major provinces separately and then weight them together at the end.
How Projection Actually Works In Practice
A proper projection isn't a spreadsheet with three columns. It needs at least four variables: baseline cohort composition, religious switching rates, fertility differences by religious group, and migration patterns. The first three are the main drivers. Migration matters less for religion than for most demographic categories because Argentina's recent immigrant flows are overwhelmingly secular or Catholic from neighboring countries, so they tend to reinforce the existing base rather than shift it. Fertility is a quiet force here. Evangelical households in Argentina have historically had slightly higher fertility than the national average, but the gap has narrowed to maybe 0.1 to 0.3 children per woman over the last decade. That won't move the needle much in a ten-year window. The real lever is switching. Catholics converting to evangelical churches, people leaving religion entirely, and the small but steady growth in non-Christian groups through conversion and immigration are what actually change the percentages year to year. I used a Markov chain transition matrix for a client project a couple of years ago. You take the 2022 baseline and build a transition probability table showing what share of Catholics become unaffiliated each year, what share of Catholics convert to evangelical, what share of evangelicals stay, and so on. The 2022 data suggested an annual Catholic attrition rate somewhere between 0.4% and 0.7% of the Catholic population moving to unaffiliated or other religions, and evangelical retention above 85%. Running that model forward ten years gives you a range, not a point estimate. Catholicism lands somewhere between 56% and 60%, evangelicals between 18% and 21%, and the unaffiliated between 25% and 29%. The width of that range is the honest answer. Anything claiming a single number is either guessing or hiding its assumptions.
Get the Full Details
There are real bottlenecks with this approach. The biggest one is that INDEC doesn't publish detailed switching matrices. You're estimating transition rates from period data that doesn't directly measure conversion. Another problem is that the unaffiliated category isn't monolithic. It includes people who are atheist, agnostic, spiritual but not religious, and people who simply refuse to answer. These groups have different trajectories. The atheist segment in Argentina is small, maybe 2-3%, and grows slowly. The spiritual-but-not-religious segment is harder to pin down and tends to fluctuate with economic conditions. During the 2001 crisis, unaffiliated identification spiked. It didn't fully reverse afterward, but the pattern shows that economic stress correlates with religious disaffiliation more than anything else. If you need more precision than a Markov model can give you, the alternative is microsimulation. You build individual-level records with religion, age, sex, province, and socioeconomic status, then apply probabilistic transitions at the micro level. It's more accurate but requires a level of data that INDEC doesn't freely provide. The granular census microdata is available through restricted access, and getting it takes time and institutional credentials. For most practical purposes, the aggregate transition model is sufficient, but you should flag the limitation explicitly if anyone is going to cite your numbers.
A Note On How Much You Can Trust Any Projection
No model captures a sudden policy change, a major religious revival, or a demographic shock. Argentina's religious landscape could shift faster than any projection if, for example, a prominent political figure publicly reconverts to Catholicism and it triggers a media cycle that changes social norms around affiliation. Those events don't show up in the data until after they happen. The best you can do is state your assumptions, show a range, and treat the output as a plausible scenario rather than a forecast. That's how I present it to clients now. I used to try to nail down exact percentages. It just didn't hold up under scrutiny. The underlying trend is clear enough without perfect precision. Catholicism is declining as a share of the population but remains the dominant identifier by a wide margin. Evangelical Christianity is growing steadily, mostly through conversion rather than natural increase. The unaffiliated are the fastest-growing segment relative to their starting size. Non-Christian religions remain small but are slowly increasing through immigration and conversion. These directions are unlikely to reverse quickly. The exact numbers in five or ten years will depend on switching rates you can't fully observe today, which is why ranges matter more than point estimates.
