Table 3 Non-parametric RD estimates of hypertension-diagnosis effects on daily nutrient intake (individuals ≥ 18 years old).

From: More information, better diet?―regression-discontinuity evidence from hypertension diagnoses in China

Estimator

Mean

[SD]

RD estimates based on SBP readings (cutoff: SBP = 140)

RD estimates based on DBP

readings (cutoff: DBP = 90)

RD estimates based on the distance to the joint cutoff (SBP = 140, DBP = 90)

 

(1)

(2)

(3)

(4)

(5)

(6)

Sample

Non-hypertensive

(SBP < 140,

DBP < 90)

Diastolic

normal

(DBP < 90)

Full

Systolic

normal

(SBP < 140)

Full

Full

A. Energy (kcal)

2209.24

−57.335

−46.668

−97.036

23.963

45.152

 

[722.24]

(78.649)

(65.697)

(144.083)

(77.995)

(79.879)

Optimal bandwidth

 

16.004

13.158

5.991

12.054

23.033

B. Carbohydrates (g)

316.18

23.005

14.902

0.926

4.937

4.763

 

[113.86]

(15.776)

(11.930)

(17.195)

(12.526)

(18.656)

Optimal bandwidth

 

13.414

13.524

7.995

11.772

16.332

C. Fat (g)

72.73

−12.855***

−9.441***

−7.419

−0.753

−2.389

 

[46.84]

(4.394)

(3.312)

(11.842)

(6.463)

(4.664)

Optimal bandwidth

 

12.455

15.938

6.665

11.155

22.270

D. Protein (g)

66.81

−4.326

−3.838

2.313

1.320

1.358

 

[23.92]

(3.136)

(2.426)

(3.506)

(1.885)

(2.940)

Optimal bandwidth

 

14.681

13.721

5.600

15.153

20.721

N

14,850

15,680

17,817

16,054

17,817

17,817

  1. Data were drawn from the China Health and Nutrition Survey (1997, 2000, 2004, 2006, 2009, and 2011). Analytical samples are restricted to observations within ±40 mmHg intervals of the respective cutoffs. The sample size (N) is the same for all panels in each column of the table; the analytical sample size for each estimate is determined by the optimal bandwidth chosen by minimizing the mean squared error (Calonico et al. 2014). Non-parametric RD estimates using the triangular kernel and the associated optimal bandwidths are reported. Standard deviations are reported in brackets. Robust standard errors are reported in parentheses, clustered at the community level.
  2. ***p < 0.01.