What 41,544 Filipino Households Told Me About Money

I grew up hearing my parents talk about money in whispers. How much was coming in, how much was going out, whether there'd be enough left for tuition. It wasn't until I stumbled on the Philippine Statistics Authority's FIES dataset that I realized those private conversations had a public mirror — 41,544 households, all telling some version of the same story.

The FIES is one of the most detailed household surveys in the country. It tracks income, expenditure, and family characteristics across every region. I wanted to dig into it not just to see averages, but to understand the shape of inequality — where it's worst, why it persists, and what spending patterns reveal about daily survival.

What I Built

A Python-based analysis of the full FIES microdata. The project calculates the Gini coefficient from scratch, groups households into income deciles, and compares spending breakdowns across regions. I built visualizations for regional income distribution, expenditure share by category, and head-to-head comparisons between the richest and poorest areas.

Why This Project

Income inequality in the Philippines isn't a secret, but the specifics often get lost in broad statements like "the rich are getting richer." I wanted actual numbers. How big is the gap between NCR and ARMM? What percentage of income goes to food in the bottom decile versus the top? These are questions you can't answer with gut feeling.

I also wanted to practice computing inequality metrics myself instead of relying on pre-packaged statistics. Calculating the Gini coefficient by hand (well, by code) forces you to understand what it actually measures.

How I Put It Together

The raw FIES data came as a large CSV from the PSA. Cleaning it took a surprising amount of time — column names weren't intuitive, some income fields had to be derived from sub-components, and a few records had missing region codes that I had to map manually.

For the Gini coefficient, I sorted all households by total income, computed the cumulative income share, and compared it against a perfectly equal distribution. The calculation itself is just a few lines of Python with numpy, but setting up the data correctly was the real challenge.

Income decile grouping used pandas.qcut() to split households into 10 equal-sized bins. From there, I could calculate average income, average expenditure, and spending shares per decile. Regional breakdowns followed the same pattern but grouped by PSA region codes instead.

What the Data Showed

The numbers confirmed what most Filipinos already feel but rarely see quantified.

Households in the bottom income decile spend over 40% of everything they earn just on food. By contrast, the top decile spends closer to 25% on food — and they're eating much better. That gap matters because it means the poorest families have almost nothing left for education, health, or savings after keeping everyone fed.

The regional story was just as stark. NCR households earn roughly 3 to 4 times more than those in ARMM. That's not a small gap — it's an entirely different economic reality. Families in ARMM aren't just earning less; they're operating in a different economy with different prices, opportunities, and constraints.

Education spending showed an interesting pattern too. Middle-income families actually dedicate the highest share of their budget to education, not the wealthiest ones. The rich spend more in absolute terms, but as a percentage of income, it's the families stretching to send kids to better schools who feel the pressure most.

• NCR average household income is 3-4x higher than ARMM
• Food spending dominates for the bottom 3 income deciles
• Middle-income households spend the largest share of income on education
• The Gini coefficient confirms high inequality, consistent with World Bank estimates

Looking Back

This project changed how I read economic headlines about the Philippines. When someone cites average household income, I now think about which decile and which region they're talking about — because the averages hide enormous variation. A single national number smooths out the lived experience of a family in Tawi-Tawi compared to one in Makati.

If I were to extend this, I'd love to combine FIES data with price indices by region. Income differences tell half the story; the other half is what that income can actually buy where you live.