---
title: Monthly Employed Persons by Occupation and DAIOEs
emoji: 🌍
colorFrom: yellow
colorTo: indigo
sdk: docker
pinned: false
license: mit
---

# Monthly Employed Persons by Occupation and DAIOEs

![AI-Econ Lab logo](logos/lab.svg)

**[Live app on Hugging Face Spaces](https://huggingface.co/spaces/joseph-data/app_months)**

## Overview

This repository builds and deploys the **AI Exposure & Employment Dashboard** — an
interactive Shiny app for exploring monthly Swedish employment by occupation alongside
**DAIOE** (Dynamic AI Occupational Exposure Index) scores.

**DAIOE** measures the potential applicability of AI to occupational content over time.
It tracks annual progress across nine AI subdomains (games, vision, language, speech)
and links capability advances to occupational requirements via O*NET abilities, weighted
by ability importance and social-skill intensity. It is a measure of AI *exposure*, not
adoption or automation probability. See the [AI-Econ Lab](https://www.ai-econlab.com/ai-exposure-daioe) for full methodology.

The dashboard filters observations by year, sex, occupation, DAIOE metric, and
employment-change horizon, then shows summary value boxes (average AI exposure, median
employment change, observation count), a Plotly scatter plot with an OLS trendline, and
a filterable data table.

## Runtime Files

The deployable app is intentionally small:

- `app.py` defines the Shiny Express app.
- `_brand.yml` defines the Shiny theme and points to the lab logo.
- `logos/lab.svg` is shown in the app sidebar and README.
- `data/scb_months_lvl1.parquet` is the app dataset.
- `Dockerfile`, `.dockerignore`, `pyproject.toml`, and `uv.lock` define the containerized runtime.

## Dependencies

Key Python packages (see `pyproject.toml` for pinned versions):

| Package | Purpose |
| --- | --- |
| `shiny[theme]` | Shiny for Python web framework |
| `shinywidgets` | Plotly widget integration |
| `plotly` | Interactive scatter plots |
| `polars` | Fast dataframe operations |
| `pandas` | DataFrame conversion for Plotly |
| `pyarrow` | Parquet file I/O |
| `statsmodels` | OLS trendline in scatter plot |

Requires **Python ≥ 3.14**.

## Local Development

Install dependencies with `uv`:

```bash
uv sync
```

Run the app locally:

```bash
uv run shiny run app.py --reload
```

Or run with the project virtual environment directly:

```bash
.venv/bin/python -m shiny run app.py --reload
```

## Docker

Build the image:

```bash
docker build -t ai-econ-daioe-months .
```

Run it locally:

```bash
docker run --rm -p 7860:7860 ai-econ-daioe-months
```

The container serves the app on `http://127.0.0.1:7860`.

## Branch Workflow

This repository uses separate branches for each stage of the data and deployment
pipeline.

| Branch | Purpose | Main output |
| --- | --- | --- |
| `scb_pull` | Pulls or prepares the monthly SCB employment data. The `SCB Pull -> DAIOE Pull` workflow runs `main.py` on this branch. | `data/scb_months.parquet` |
| `daioe_pull` | Receives the SCB output and enriches or merges it with DAIOE exposure data. The `DAIOE Pull -> Development` workflow runs `main.py` here. | `data/scb_months_lvl1.parquet` |
| `development` | Integration branch for the merged dataset and deployable app files before promotion. The `Development -> Main` workflow promotes the deploy bundle from here. | deploy-ready app files |
| `main` | Production/deployment branch. This branch contains the Dockerized Shiny app and syncs to Hugging Face Spaces. | running dashboard |

The pipeline is therefore:

```text
scb_pull -> daioe_pull -> development -> main -> Hugging Face Spaces
```

## GitHub Actions

The repository contains four workflows:

- `.github/workflows/01_scb_pull_to_daioe_pull.yml` builds the base SCB parquet and pushes it to `daioe_pull`.
- `.github/workflows/02_daioe_pull_to_development.yml` builds the DAIOE-enriched parquet and pushes it to `development`.
- `.github/workflows/03_development_to_main.yml` promotes deployable app files to `main`.
- `.github/workflows/sync_to_hub.yml` syncs `main` to the Hugging Face Space `joseph-data/app_months`.

The scheduled workflows run daily at `00:00 UTC`, and each can also be run manually
with `workflow_dispatch`.

## Data Shape

The app dataset (`data/scb_months_lvl1.parquet`) has monthly occupation-level rows with:

- `year`, `month`, `sex`, `occupation`, `emp_count` — dimensions and employment count
- `pct_chg_1m`, `pct_chg_3m`, `pct_chg_6m` — percentage employment change over 1, 3, and 6 months

DAIOE weighted-average exposure columns (`daioe_*_wavg`) used as x-axis metrics:

| Column | Label |
| --- | --- |
| `daioe_allapps_wavg` | All AI Applications |
| `daioe_stratgames_wavg` | Strategic Games |
| `daioe_videogames_wavg` | Video Games |
| `daioe_imgrec_wavg` | Image Recognition |
| `daioe_imgcompr_wavg` | Image Compression |
| `daioe_imggen_wavg` | Image Generation |
| `daioe_readcompr_wavg` | Reading Comprehension |
| `daioe_lngmod_wavg` | Language Models |
| `daioe_translat_wavg` | Translation |
| `daioe_speechrec_wavg` | Speech Recognition |
| `daioe_genai_wavg` | Generative AI |