WikiSummary, the Social Science Summary Database

Brown. 1995. Party cleavages and welfare effort in the American states. American Political Science Review 89: 23-34.

Scholars have found inconclusive results when studying whether Democratic control of the state government correlates with higher welfare spending. Brown points out that scholars have neglected an important variable: the nature of each state's cleavage system.

X: STATE PARTY CLEAVAGES

Using a novel technique, Brown shows which cleavages matter most in most states (see Table 2, pg 26). He then identifies three ideal types (see Table 3, pg 28):

• Southern partisan cleavage: driven primarily by race, but also low-income and rural support for Democrats.
• New Deal cleavage (class-based): driven primarily by a Catholic-Protestant split, but also income, unionization, education, and sex.
• Post-New Deal cleavage: essentially, a majority vs minority cleavage, with the Democratic party a strange coalition of many "out" groups (blacks, Catholics, urbanites, unions, low income, female, Jewish).
• A few states have no significant cleavages (Arkansas, Kentucky, West Virginia, and Wyoming)

Y: STATE WELFARE EFFORT

Hypothesis: States with a class-based (New Deal) cleavage will spend the most on welfare.

CONTROLS

State resources (income per capita vs size of welfare-receiving population), elite values (the liberalism of those in government), and racism (which is poorly measured; see below).

RESULTS

See Table 4, pg 30, for regression results. "New Deal" is the baseline category. Thus:

• New Deal states: Party control has a +0.00039 (significant) effect on welfare effort
• Southern states: Party control has +0.00007 (insignificant) effect (0.00039 - 0.00032 = 0.00007)
• Post-New Deal states: Party contrl has a +0.00004 (insignificant) effect (0.00039 - 0.00035 = 0.00004)

IN ENGLISH: The effect of party control is 9.5 times greater in New Deal states than in post-New Deal states, and 5.5 times stronger in New Deal states than in Southern states. All predictions are borne out.

FOOD FOR THOUGHT

• Uses a pooled cross sectional regression (combines data from different surveys from 1976-85), which presents risks of heteroskedasticity and autocorrelation (i.e. inflated results). Claims to control for this.
• In an effort to control for a state's "population's attitude toward [black] welfare recipients," Brown includes "a variable for the percentage of the state's population that is black" (29-30). What on earth?!!?