Background & Motivation

QUESTION: Is metropolitan-level economic segregation a correlate of the pace of decline in heart disease mortality?

ASSUMPTIONS:

  1. Economic segregation is best measured at metropolitan scale.
  2. Economic segregation in 1970 is the closest ‘exposure’ period preceding county trends.
  3. Confounders or effect modifiers include individual (race, age), county (poverty, educational attainment), and metropolitan (population size) variables. Poverty and educational attainment are measured race-specific
  4. Data is 3-level: individual age/race groups clustered within counties clustered with CBSA’s/MSA’s

Economic segregation measured with the H index of income diversity

  1. ALL70_H: This is a single summary measure (H) of residential mixing across income categories in 1970
  2. ALL70_H10: This is a summary of the spatial separation of the lowest 10th percentile of income compared to the upper 90th percentile of income families. This is poverty concentration.
  3. ALL70_H90: This is a summary of the spatial separation of the highest 10th percentil of income from those in the lower 90th percentile. This is concentration of affluence.

Data overview

Distribution of county variables by quintiles of CBSA-level economic segregation, 1970

Q1 Q2 Q3 Q4 Q5 p
N Counties 81 85 126 188 226
HD % Decline (mean (sd)) 63.1 (12.9) 57.4 (12.1) 58.8 (13.4) 54.7 (14.7) 50.7 (19.3) <0.001
Total population, 1970 (mean (sd)) 137465.1 (135924.9) 166512.5 (192085.0) 166914.0 (261959.8) 193052.9 (501552.8) 170939.0 (663657.6) 0.936
% population change 1970 - 2010 (mean (sd)) 64.5 (76.9) 83.6 (135.6) 93.7 (139.0) 93.3 (104.3) 174.6 (219.8) <0.001
Total % in poverty, 1970 10.3 (4.7) 10.7 (6.1) 11.7 (8.0) 13.3 (7.5) 18.3 (9.9) <0.001
Black % in poverty, 1970 24.7 (13.1) 27.4 (15.6) 25.2 (14.1) 30.4 (17.1) 38.0 (17.4) <0.001
White % in poverty, 1970 9.9 (4.8) 9.6 (5.5) 10.2 (7.0) 10.6 (6.4) 12.3 (6.1) <0.001
Total % with college degree, 1970 8.6 (3.5) 9.8 (3.9) 8.8 (4.1) 8.4 (5.1) 8.5 (5.5) 0.270
Black % with college degree, 1970 9.2 (17.6) 5.4 (7.3) 6.0 (9.3) 4.6 (6.1) 4.1 (6.0) 0.001
White % with college degree, 1970 8.7 (3.5) 10.0 (3.9) 9.1 (4.2) 9.1 (5.7) 9.6 (6.0) 0.397

Model Results

Unadjusted models

Models are fit for Overall H, as well as H10 (poverty concentration) and H90 (affluence concetration).
Unadjusted multilevel model follow this form:

m<-lmer(decline ~ ALL70_H + (1 | FIPS) + (1 | CBSA), data = decline)
Here are results for unadjusted models with each of 3 segregation measures:
Predictors Model version
  Overall H   H10   H90
    B CI   B CI   B CI
Fixed Parts
(Intercept)   69.5 *** 65.5 – 73.4   65.4 *** 61.3 – 69.5   68.4 *** 63.9 – 72.9
ALL70_H   -100.2 *** -140.6 – -59.9            
ALL70_H10         -70.0 ** -121.2 – -18.8      
ALL70_H90               -63.4 *** -96.9 – -30.0
Random Parts
σ2   156.295   156.275   156.275
τ00, FIPS   36.594   36.730   36.890
τ00, CBSA   43.983   50.457   47.077
NFIPS   708   708   708
NCBSA   171   171   171
ICCFIPS   0.154   0.151   0.154
ICCCBSA   0.186   0.207   0.196
Observations   2731   2731   2731
R2 / Ω02   .484 / .455   .486 / .455   .485 / .455
AIC   22180.659   22195.485   22190.147
Notes * p<.05   ** p<.01   *** p<.001



Adjusted models

The adjusted, no interaction models are of this general form:

m<-lmer(decline ~ ALL70_H + race + age + POP_CHANGE + race_specific_poverty + race_specific_college + log(TOTPOP) + (1 | FIPS) + (1 | CBSA), data = decline)



And here are results of adjusted (no-interaction) models for all 3 indices:

Predictors Model version
  Overall H   H10   H90
    B CI   B CI   B CI
Fixed Parts
(Intercept)   36.2 *** 18.4 – 54.1   33.3 ** 13.2 – 53.5   37.7 *** 18.6 – 56.8
ALL70_H   -146.4 *** -188.3 – -104.5            
race (White)   2.7 *** 1.6 – 3.8   2.6 *** 1.5 – 3.7   2.6 *** 1.5 – 3.7
age (65+)   -8.5 *** -9.3 – -7.7   -8.5 *** -9.3 – -7.7   -8.5 *** -9.3 – -7.7
POP_CHANGE   1.8 *** 1.4 – 2.3   1.8 *** 1.3 – 2.2   1.8 *** 1.4 – 2.2
race_specific_poverty   2.0  -2.1 – 6.0   1.8  -2.3 – 5.8   1.9  -2.1 – 5.9
race_specific_college   -1.3  -8.7 – 6.0   -1.3  -8.7 – 6.0   -1.3  -8.6 – 6.1
log(TOTPOP)   2.8 *** 1.4 – 4.2   2.8 *** 1.2 – 4.4   2.7 *** 1.2 – 4.2
ALL70_H10         -135.0 *** -191.5 – -78.5      
ALL70_H90               -101.6 *** -140.1 – -63.2
Random Parts
σ2   95.782   95.794   95.777
τ00, FIPS   30.945   30.814   30.972
τ00, CBSA   31.995   40.154   38.007
NFIPS   633   633   633
NCBSA   139   139   139
ICCFIPS   0.195   0.185   0.188
ICCCBSA   0.202   0.241   0.231
Observations   2331   2331   2331
R2 / Ω02   .611 / .597   .612 / .598   .612 / .598
AIC   17895.292   17914.022   17910.876
Notes * p<.05   ** p<.01   *** p<.001



Interaction Models

Finally, there is evidence for signifcant interaction of Economic Segregation with race (black v white) and age (35-64 v 65+). Interaction models are of this form:

m<-lmer(decline ~ ALL70_H*race*age + POP_CHANGE + race_specific_poverty + race_specific_college + log(TOTPOP) + (1 | FIPS) + (1 | CBSA), data = decline)



Predictors Model version
  Overall H   H10   H90
    B CI   B CI   B CI
Fixed Parts
(Intercept)   31.2 *** 13.1 – 49.2   29.8 ** 9.5 – 50.0   32.8 *** 13.4 – 52.2
ALL70_H   -102.9 *** -151.1 – -54.8            
raceWhite   12.1 *** 8.0 – 16.1   9.9 *** 6.0 – 13.7   12.2 *** 7.4 – 16.9
age65+   3.7  -0.5 – 7.9   1.0  -3.0 – 5.1   3.5  -1.5 – 8.5
POP_CHANGE   1.8 *** 1.4 – 2.3   1.8 *** 1.3 – 2.2   1.8 *** 1.4 – 2.2
race_specific_poverty   2.5  -1.5 – 6.4   2.3  -1.7 – 6.3   2.2  -1.8 – 6.2
race_specific_college   -0.9  -8.1 – 6.4   -0.9  -8.1 – 6.4   -0.7  -8.0 – 6.5
log(TOTPOP)   2.9 *** 1.5 – 4.2   2.8 *** 1.2 – 4.4   2.7 *** 1.2 – 4.3
ALL70_H:raceWhite   -86.0 *** -123.1 – -48.8            
ALL70_H:age65+   -111.7 *** -151.1 – -72.3            
raceWhite:age65+   -22.8 *** -28.4 – -17.1   -18.9 *** -24.3 – -13.5   -23.2 *** -29.9 – -16.5
ALL70_H:raceWhite:age65+   211.5 *** 158.3 – 264.7            
ALL70_H10         -99.9 ** -162.4 – -37.3      
ALL70_H10:raceWhite         -76.1 *** -118.0 – -34.2      
ALL70_H10:age65+         -101.3 *** -145.6 – -57.1      
ALL70_H10:raceWhite:age65+         204.3 *** 144.5 – 264.1      
ALL70_H90               -71.1 ** -114.9 – -27.3
ALL70_H90:raceWhite               -63.0 *** -95.4 – -30.6
ALL70_H90:age65+               -79.5 *** -113.9 – -45.1
ALL70_H90:raceWhite:age65+               155.9 *** 109.7 – 202.1
Random Parts
σ2   92.576   93.334   93.456
τ00, FIPS   31.611   31.369   31.444
τ00, CBSA   32.038   40.121   38.016
NFIPS   633   633   633
NCBSA   139   139   139
ICCFIPS   0.202   0.190   0.193
ICCCBSA   0.205   0.243   0.233
Observations   2331   2331   2331
R2 / Ω02   .626 / .613   .623 / .610   .623 / .609
AIC   17815.666   17848.013   17848.632
Notes * p<.05   ** p<.01   *** p<.001



Because it is hard to interpret the interaction models, here are plots of the segregation x race x age effects for ALL70_H, the overall segregation index (results are similar for H10 and H90, the indices of poverty and affluence segregation resepectively).

Interaction plot of segregation x race x age

Conclusion

  1. I’m not sure why these results are different from past results. Possibly explanations include: new updated data to 2015; adjustment for race & age rather than race & gender; prior (or perhaps current) error in data preparation/merge.
  2. I will continue to look into possibly spurious reasons for new results
  3. Assuming they hold, this is a substantially different story for this paper.