Behavioural Characteristics and Financial Distress

2. Secondly, the results show that behavioural characteristics matter. The behavioural
effects are economically large and add quite a bit to the fit of the model.

3. Finally, behaviour seems to be more important than financial literacy; being finan-
cially literate can reduce the probability of financial distress by up to 9 per cent,
while being patient can reduce it by 10 per cent. The effect of not being impulsive
is even larger. This is an important finding because it suggests that the recent drive
to improve financial literacy levels in the population may not be sufficient to prevent
financial difficulties. These efforts should be combined with tools to improve individ-
uals’ organisational skills and devices to, as much as is possible, minimise the impact
of behavioural and psychological traits on financial outcomes.

4.2 Do the Effects Differ by Degree of Struggle?

As discussed earlier, the dependent variable “Struggle to Keep Up” is constructed from
several responses to the question on how well people are keeping up with their bills and
credit commitments. These responses are mutually exclusive, suggesting that in addition
to examining the factors that cause people to get into financial difficulties, it is also possible
to examine if the effect of these factors differs by the degree of financial difficulty reported.
I therefore create a dependent variable “Degree of Struggle to Keep Up” (Y) which has four
outcomes, as follows:

Yi = 1, if “Falling behind with some/many.”

Yi = 2, if “Constant struggle.”

Yi = 3, if “Struggle from time-to-time.”

Yi = 4, if “No difficulties keeping up.”

I use a generalized ordered logit model to examine if the various effects of the demo-
graphic, economic and behavioural factors differ across these outcomes. This model, which
nests a number of more restrictive models such as the ordered logit model, is described in
detail in Williams (2006). ^9,10 In the current context, the generalized ordered logit model
⁹The ordered logit model is more restrictive because it imposes the parallel lines assumption, whereby
slope coefficients are deemed constant across the various outcomes of the ordered categorical dependent
variable. The generalized ordered logit model is able to nest this assumption for all or a subset of variables.
¹⁰The model is implemented in Stata using the gologit2 command. The results reported here are based
on the final specification chosen by the ‘autofit’ option.

21

<<   22   23   24   25   26   27   28   >>

More intriguing information