Table 1: Types of local federal expenditure

Notes:

All variables are used per capita, based on county population .

source, Bureau of Economic Analysis 1990.

source, Bureau of the Census 1984-90.

source, Bureau of the Census 1986-91 and 1991; county totals are estimated as in Mebane 1993.

units, $1000 per person.

units, jobs per person.

Table 2: 95% confidence intervals for support values that maximize local federal expenditures

maximum is an elite-oriented targeting value maximum is not in the elite-oriented range

civilian procurements (.59, .61) military procurements (.00, 1.00)

military employment (.70, .71) civilian employment (.00, 1.00)

civilian salaries (.68, .68) military salaries

Federal welfare transfers (.61, .62) transfer payments no max

Federal education transfers (.60, .63) direct payments (.35, .44)

State highways transfers (.69, .75) Federal highways transfers (.00, 1.00)

State education transfers (.55, .59) Federal other transfers (.27, .37)

State welfare transfers (.35, .39)

State other transfers no max post-midterm

maximum is an elite-oriented targeting value maximum is not in the elite-oriented range

civilian procurements (.64, .71) military procurements (.79, .81)

civilian employment (.49, 1.00)

military employment (.00, 1.00)

civilian salaries (.02, 1.00)

military salaries no max

transfer payments (.00, 1.00)

direct payments no max

Federal highways transfers no max

Federal welfare transfers (.98, .98)

Federal education transfers no max

Federal other transfers (.70, .91)

State highways transfers no max

State welfare transfers (.92, .94)

State education transfers no max

State other transfers (.74, .98)

Source: Confidence intervals are computed using normal approximations and asymptotic standard errors obtained by the delta method from the standard errors of the coefficient estimates of the targeting polynomials.

The polynomial does not have any maximum values.

Table 3: 95% confidence intervals for support values that maximize changes in local federal expenditures

maximum is a voter-oriented targeting value maximum is not in the voter-oriented range

military employment (.38, .39) civilian procurements

civilian salaries (.37, .39) military procurements (.61, .79)

military salaries (.25, .55) civilian employment no max

transfer payments (.15, .75) direct payments no max institutionally complex LFEs

maximum is a voter-oriented targeting value maximum is not in the voter-oriented range

Federal welfare transfers (.29, .35) Federal highways transfers no max

Federal education transfers no max

Federal other transfers no max

State highways transfers no max

State welfare transfers no max

State education transfers no max

State other transfers (.00, 1.00)

Source: Confidence intervals are computed using normal approximations and asymptotic standard errors obtained by the delta method from the standard errors of the coefficient estimates of the targeting polynomials.

The polynomial does not have any maximum values.

Table 4: Presidential vote choice probit regression equation estimates

variable I II III

Constant -.252 -.313 -.318

(-.509, .001) (-.580, -.055) (-.581, -.063)

.521 .534 .533

(.467, .581) (.478, .595) (.478, 594)

.440 .524 .381

(.106, .778) (.053, .718) (.046, 715)

.540 .524 .526

(.220, .874) (.208, .866) (.210, .870)

.330 .330 .327

(.060, .607) (.057, .611) (.059, .608)

-- 1.072

(-.250, 2.401) (.155, 1.570)

-- .733

(-.435, 1.885) (.155, 1.570)

2.762 3.571 3.716

(.351, 5.328) (.960, 6.708) (1.124, 6.848)

731.94 723.86 724.06

Source: Equations for the probability of voting for the Republican rather than the Democratic presidential candidate, using 1988 ANES and contextual data. N = 975 respondents from 91 counties. In equation I, 85.6% correctly classified, . In equation II, 85.0% correctly classified, . In equation III, 85.1% correctly classified, . Coefficients are probit MLEs. 95% confidence intervals (percentile-t bootstrap) are in parentheses. Intercept-only: log-likelihood = 1348.76.

survey attitude variable.
county-level change (purged): total for 1987-88 minus total for 1985-86.
parameters constrained to equal one another.
(Goodman-Kruskal gamma) measures the rank correlation between observed responses and predicted probabilities.

Table 5: Probabilities of a pro-incumbent (pro-Republican) vote at minimum and maximum values of local-area employment changes and subjective economic evaluations

national family

nongovernmental civilian military economy finances

employment employment employment evaluations evaluations

at mean of PID .35 .71 .39 .63 .36 .61 .43 .64 .45 .58

strong Democrat .03 .18 .04 .13 .03 .12 .05 .14 .06 .10

Democrat .09 .36 .11 .28 .10 .26 .14 .28 .15 .23

independent-Democrat .22 .57 .25 .47 .23 .46 .29 .49 .30 .42

independent .40 .76 .44 .68 .42 .67 .49 .69 .50 .63

independent-Republican .61 .89 .65 .84 .63 .83 .69 .85 .71 .81

Republican .79 .96 .82 .94 .80 .93 .85 .94 .86 .92

Strong Republican .91 .99 .93 .98 .92 .98 .94 .98 .95 .97

.43 .65 .46 .60 .44 .59 .49 .60 .50 .56

Source: Computed using the vote choice probit regression equation point estimates of Table 4. The probability for each level of party ID and each minimum or maximum value is computed with the other variables in the equation at their mean values in the 1988 ANES sample. Employment-changes values are for the purged data.

Minimum value. For evaluations, 0 means ``worse.''

Maximum value. For evaluations, 1 means ``better.''

Average of the probabilities over the seven party ID categories, weighting by the sample proportions for the PID categories among voters in the ANES data. In order from strong Democrat through strong Republican the proportions are .199, .165, .112, .063, .128, .148, .186.

Table 6: Effects of LFEs on individuals' subjective economic evaluations

national economy family finances

variable evaluations evaluations

Intercept 1 -.956 -.628

(.069) (.090)

Intercept 2 .468 .328

(.065) (.089)

.166 .111

(.017) (.019)

-- .010 (.0019)

.126 -- (.064)

-- -.143 (.066)

-- 6.492 (3.264)

-1.843 -- (.440)

-2.442 -- (1.169)

-- 35.15 (17.95)

-91.46 -- (43.05)

3.722 2.900

(1.286) (1.393)

Source: Ordered probit regression equations for the probability of giving a more favorable response, using 1988 ANES and contextual data. National economy N = 975 ( ) and family finances N = 901 ( ) respondents. Coefficients are MLEs. Standard errors are in parentheses.

survey attitude variable.
county-level change: total for 1987-88 minus total for 1985-86.
(Goodman-Kruskal gamma) measures the rank correlation between observed responses and predicted probabilities.

Table 7: Probabilities of evaluating the national economy as ``better'' at minimum and maximum values of the LFE changes that have heterogeneous effects

social welfare civilian military

transfers employment employment SPEND value

increase all index items .067 .298 .029 .389 .011 .349

.848 .026 .190 .190 .192 .192

decrease all index items .9998 .0004 .552 .070 .709 .087

Source: Computed using the ordered probit regression equation point estimates of Table 6. The probability for each level of SPEND and each minimum or maximum value is computed with the other variables in the equation at their mean values in the 1988 ANES sample. LFE-changes values are for unpurged data.

Minimum value.

Maximum value.

Table 8: Effects of LFEs on individuals' partisanship and spending tastes

variable party identification spending scale

Intercept -1.854 -.278

(.300) (.019)

.371 -- (.085)

1.121 .170

(.291) (.037)

-15.22 -- (6.33)

-- -.981 (.520)

-2.065 -- (.747)

5.832 .482

(1.638) (.193)

-- .891 (.357)

.161 -- (.055)

Source: Ordinary least squares regression equations, using 1988 ANES and contextual data. N = 901. Party identification equation residual mean square error = 4.71 ( ). Spending scale equation residual mean square error = .084 ( ). Standard errors are in parentheses.

survey attitude variable.
county-level change: total for 1987-88 minus total for 1985-86.

Figure 1: Effects of support on federal local expenditures, pre-midterm

Notes: Quasi-likelihood estimates. The number in parentheses shows the degree of the targeting polynomial.

Figure 2: Effects of support on local federal expenditures, post-midterm

Notes: Quasi-likelihood estimates. The number in parentheses shows the degree of the targeting polynomial.

Figure 3: Effects of support on changes in local federal expenditures, by support

Notes: Computed using the targeting polynomial estimates shown in Figures 1 and 2.

• About this document ...