Consumer Spending and Customer Satisfaction: Untying the Knot

Table 1

Estimated MARS model.

Parameter	Estimate	T-ratio	Hansen stability test
Constant	8.23207	1525.89813	0.044	(0.90)
Basis function 1	0.83131	21.82990	0.053	(0.85)
Basis function 2	−0.59353	−2.99735	0.016	(1.00)
Basis function 3	0.39318	5.31771	0.042	(0.92)
Basis function 4	−0.30214	−5.29911	0.029	(0.98)
Basis function 6	2.33858	4.55403	0.012	(1.00)
Basis function 8	−6.02882	−3.58748	0.088	(0.63)
Basis function 10	−0.73692	−4.76604	0.017	(1.00)
Hansen variance test			0.163	(0.34)
Hansen joint test			1.015	(0.88)
Basis functions
BF1 = max(0, ln(RPDI) − 8.8473)
BF2 = max(0, 8.8473 − ln(RPDI))
BF3 = max(0, ln(DSR) − 2.59525)
BF4 = max(0, 2.59525 − ln(DSR))
BF6 = max(0, 1.46047 − ln(CCNR)) ∗ BF4
BF8 = max(0, 1.56752 − ln(CCNR)) ∗ BF1
BF10 = max(0, 1.05934 − ln(CCR))

RPDI, DSR, CCNR, and CCR denote real personal disposable income, the debt-service ratio, outstanding nonrevolving consumer credit, and outstanding revolving consumer credit, respectively. Estimation spans 1994 Q4 to 2008 Q3 inclusive allowing a maximum of ten basis functions and two variable interactions. The MARS sum of squared residuals based criterion function was adjusted by a penalty value of two with the dependent variable discretionary personal consumption expenditures. All series were transformed into their natural logarithms. Hansen stability tests refer to the Hansen [11] test (null: no structural change) and the associated probability values (in parentheses) for individual parameter stability, variance stability, and the hypothesis of joint stability. MARS is available at http://www.salford-systems.com/.