Giffen Good

The mechanism

The law of demand says: when the price of a good rises, people buy less of it. Giffen goods are the classic counterexample. The mechanism rests on the Slutsky decomposition. Any price change affects demand through two channels:

1. Substitution effect. The good is now relatively more expensive. Switch to alternatives. This effect always points away from the good.
2. Income effect. A price rise makes the consumer's real income smaller. For a normal good, less real income means less consumption (reinforces substitution). For an inferior good, less real income means more consumption — because the consumer can no longer afford the things they used to substitute upward to.

For an inferior good, the income effect points toward the good while the substitution effect points away. If the income effect is large enough — large enough means the good is strongly inferior and takes a big share of the budget — it can overwhelm the substitution effect. Net effect: price up, quantity up. That is the Giffen good.

The Hunan rice experiment

Jensen and Miller (2008) ran a field experiment among 1,300 households in two Chinese provinces. In Hunan, rice was the staple — a poor household might spend 30–40% of food expenditure on it. The researchers issued price-discount vouchers to randomly selected households over five months, effectively lowering the price of rice for treatment but not control families.

The standard prediction (rice is a normal good): a lower price should raise rice consumption. The Giffen prediction (rice is a Giffen good for poor households): a lower price should reduce rice consumption, because the discount frees up budget to buy more meat and vegetables, displacing rice.

Subgroup	Effect of lowering rice price on rice consumption	Interpretation
Very poor Hunan households	Decrease (statistically significant)	Giffen behavior — rice is the inferior staple
Moderately poor Hunan households	Roughly zero / mixed	Income and substitution effects offset
Less poor Hunan households	Increase	Standard demand — rice behaves normally
Gansu households (wheat staple)	Decrease for poorest tier	Same Giffen pattern, different staple
Households near subsistence	No effect	Too poor to substitute even with discount
Households well above subsistence	Standard increase	Income effect smaller; substitution dominates

The interpretation: Giffen behavior emerges in a specific income window. Below subsistence, the household is locked in. Above some threshold, the substitution effect wins. In between — poor enough that meat is a luxury, rich enough that they actually choose between rice and meat — rice behaves as a Giffen good.

A toy numerical example

Consider a household with a $20 weekly food budget. Two foods: rice ($1/kg) and meat ($5/kg). Subsistence requires 2,000 kcal/day, ≈ 14,000 kcal/week. Rice provides 3,500 kcal/kg; meat provides 2,500 kcal/kg.

At $1 rice and $5 meat: the household's optimal bundle is, say, 12 kg rice + 1.6 kg meat = $20 spent, ≈ 46,000 kcal — comfortably above subsistence and including some preferred meat.

Now rice rises to $1.50/kg. The household tries to keep meat consumption: 1.6 kg meat costs $8, leaving $12 for rice = 8 kg rice. Total: 8 × 3,500 + 1.6 × 2,500 = 32,000 kcal. Still above subsistence — for now.

But the household's real-income loss is severe. To maintain subsistence calories, they need to give up meat: drop to 0.5 kg meat ($2.50), use $17.50 on rice = 11.7 kg rice. Rice consumption rose from 12 kg to ~12 kg even though price rose — the household substituted rice for the meat they could no longer afford. That's the Giffen mechanism in cartoon form.

Giffen good vs related concepts

	Normal good	Inferior good	Giffen good	Veblen good
Demand slope	Downward	Downward	Upward	Upward
Income elasticity	Positive	Negative	Strongly negative	Positive (luxury)
Substitution effect	Negative	Negative	Negative (overwhelmed)	Negative (overwhelmed)
Income effect	Reinforces	Offsets, partially	Offsets, dominates	N/A — driven by signaling
Typical example	Restaurant meals	Bus tickets	Hunan rice for the very poor	Rolex, Birkin
Driver of upward slope	—	—	Poverty + budget-dominance	Status signaling

The historical case — and why it failed

Alfred Marshall's Principles (1895) credits Sir Robert Giffen with observing that during the Irish Potato Famine (1845–1852), potato prices rose and yet consumption among the very poor also rose. The explanation: potatoes were a budget-dominant inferior staple; price rises made families too poor to afford meat, and they fell back on more potatoes. The story spread through textbooks for nearly a century.

Sherwin Rosen and Gerald Dwyer's 1982 reexamination of the historical Irish data found the case rests on weak evidence. Famine reduced the potato supply; prices and quantities moved in opposite directions for that reason, not because of upward-sloping demand. George Stigler made similar criticisms earlier. The Irish potato example survived in textbooks long after economic historians stopped believing it.

That's why Jensen and Miller's 2008 paper was such a significant contribution — they finally produced clean, randomized, modern data showing genuine Giffen behavior in a controlled setting.

Counterarguments and skeptical takes

Even Jensen-Miller is contested. Some commentators argued the rice subsidy operated through informational or labeling channels rather than pure price effects. Subsequent work has refined the estimates but the basic pattern — Giffen behavior among the very poor — has held up.

Rare in practice. Even granting the existence of Giffen goods, they are exceedingly rare. The required conjunction (strongly inferior + budget-dominant + no close substitute) is unusual outside subsistence settings. For everyday goods in developed economies, Giffen behavior is essentially nonexistent.

Aggregate vs individual. Even if some individual consumers exhibit Giffen behavior, aggregate market demand may still slope downward because most consumers are not in the Giffen window.

Variants and related theory

• Strict vs weak Giffen. Strict: dq/dp > 0. Weak: dq/dp ≥ 0 (no response to a price rise). Real data tends to show the weak form more often.
• Marshallian vs Hicksian demand. The Slutsky decomposition is cleaner with Hicksian (compensated) demand. Marshallian demand is what's actually observed; Hicksian is what would be observed at constant utility.
• Giffen behavior for time, not money. Some labor-economics models predict Giffen-like patterns when wages are very low (more wage cuts mean more hours worked because consumption-floor binds).

Common pitfalls

• Confusing Giffen with Veblen. The single most common error. Giffen = inferior + poor; Veblen = luxury + status. Both have upward demand but for opposite reasons.
• Calling something Giffen because demand looks upward in the data. Supply shifts, anticipated future prices, and quality variation can all generate upward-sloping observed demand without any Giffen behavior. The Slutsky test requires controlled price variation.
• Forgetting the inferiority requirement. A normal good cannot be a Giffen good — the income effect points the wrong way.
• Forgetting the budget-dominance requirement. If the good is a small budget item, the income effect is too small to overpower the substitution effect, even when the good is strongly inferior.
• Citing the Irish potato uncritically. The textbook example doesn't actually hold up under historical reanalysis. Cite Jensen-Miller instead.