A survey of 12,000 households achieves a 55% response rate (6,600 respondents). To adjust for unit nonresponse bias, researchers use inverse probability weighting (IPW). A logistic regression propensity model estimates:

• High-income households: $P(\text{respond}) = 0.70$
• Low-income households: $P(\text{respond}) = 0.35$
• Middle-income households: $P(\text{respond}) = 0.55$

In the IPW approach, which group receives the largest survey weights?