Post Labor Economics

Production = f(Tasks performed by Labor, Tasks performed by Capital)

Key dynamics:
1. DISPLACEMENT EFFECT
   - Machines replace humans in existing tasks
   - Reduces labor demand, depresses wages
   - Concentrated in routine cognitive and manual tasks

2. PRODUCTIVITY EFFECT
   - Automation lowers costs, increases output
   - Some gains flow to workers via cheaper goods
   - But distribution depends on market structure

3. REINSTATEMENT EFFECT
   - New tasks created that require human comparative advantage
   - Historically: ATMs → bank branch expansion → more tellers (temporarily)
   - Question: Is this time different? Will new tasks emerge fast enough?

4. NET EFFECT
   - Historical pattern: displacement < reinstatement (net job growth)
   - Current concern: AI attacks both routine AND non-routine tasks
   - Speed of displacement may exceed speed of reinstatement

Job polarization (Autor, 2015):

         High-skill (growing)
           /               \
          /     Hollowing    \
         /       out of       \
        /      middle-skill    \
       /                        \
Low-skill (growing)     Middle-skill (shrinking)

Examples by category:
- High-skill (growing): AI researchers, surgeons, lawyers (judgment tasks)
- Middle-skill (shrinking): Bookkeeping, data entry, assembly, driving
- Low-skill (growing): Care work, cleaning, food service (non-routine manual)

import pandas as pd
import numpy as np

def compute_automation_exposure(
    occupation_tasks: pd.DataFrame,
    ai_capability_scores: dict,
) -> pd.DataFrame:
    """
    Compute occupation-level AI exposure scores.

    Based on the methodology of Felten et al. (2021) and Eloundou et al. (2023).

    Parameters:
        occupation_tasks: DataFrame with columns [occupation, task, task_weight]
        ai_capability_scores: dict mapping task -> AI performance score (0-1)

    Returns:
        DataFrame with occupation-level exposure scores
    """
    # Map AI scores to tasks
    occupation_tasks["ai_score"] = occupation_tasks["task"].map(ai_capability_scores)

    # Weighted average exposure per occupation
    exposure = occupation_tasks.groupby("occupation").apply(
        lambda g: np.average(g["ai_score"].fillna(0), weights=g["task_weight"])
    ).reset_index(name="ai_exposure")

    # Classify risk levels
    exposure["risk_level"] = pd.cut(
        exposure["ai_exposure"],
        bins=[0, 0.3, 0.6, 1.0],
        labels=["low", "medium", "high"],
    )

    return exposure.sort_values("ai_exposure", ascending=False)

UBI design parameters:

AMOUNT:
- Subsistence: $12,000-15,000/year (US, ~poverty line)
- Moderate: $18,000-24,000/year (covers basic needs + participation)
- Generous: $30,000+/year (enables full non-employment)

FUNDING MECHANISMS:
1. Carbon tax + dividend (Alaska Permanent Fund model)
2. Value-added tax on automation (Andrew Yang proposal)
3. Sovereign wealth fund (Norway model, applied to AI rents)
4. Robot tax (Bill Gates proposal)
5. Land value tax (Georgist approach)
6. Consolidated existing transfers (replacing welfare bureaucracy)

EVIDENCE FROM PILOTS:
| Pilot | Location | Duration | Key Finding |
|-------|----------|----------|-------------|
| Finland (2017-2018) | National | 2 years | No employment effect, improved well-being |
| Stockton SEED (2019-2021) | City | 2 years | Employment increased, stress decreased |
| GiveDirectly (2016-) | Kenya | 12 years | Consumption up, no labor supply reduction |
| Mincome (1974-1979) | Manitoba | 5 years | Only new mothers and students worked less |
| Y Combinator (2024-) | US cities | 3 years | Results pending |