Transcript Social spending and health outcomes

Health, social expenditure
and inequality
Findings from a cross-national
empirical analysis
Alex Sutherland, Jennifer Rubin, Jirka Taylor,
Joachim Krapels, Charlene Rohr, Melissa Felician,
Jodi Liu and Lois Davis
25 May 2016
Increased spending on healthcare in the US
has not delivered the health outcomes seen
in European countries
Is social expenditure a key part of the health outcome story?
1. Population health and
spending on health and
social care:
• Bradley et al (2011)
• Bradley and Taylor (2013)
2. Social determinants
of health:
• UK: Marmot et al (2008)
• US: Woolf and Laudan (2013)
Relevant previous research
3. Inequality and health:
• Wilkinson & Pickett (2006)
• Pickett & Wilkinson (2015)
• US: Woolf and Braveman
(2011)
4. All of these…???
How robust is the
relationship between
social expenditure and
health?
Does the ‘type’ of social
spending, i.e. on
particular age groups or
social challenges,
influence the findings?
The focus of our research
Do wider contextual
factors – like inequality
or social capital – affect
the findings?
Do these relationships
hold within a country, in
this case the US?
How robust is the
relationship between
social expenditure and
health?
Does the ‘type’ of social
spending, i.e. on
particular age groups or
social challenges,
influence the findings?
The focus of our research
Do wider contextual
factors – like inequality
or social capital – affect
the findings?
Do these relationships
hold within a country, in
this case the US?
How did we look at this?
• Collated thirty years of expenditure data as % of GDP from 34 OECD
countries. About 800-1,000 observations depending on the measures.
• Combined this with data measuring:
Health outcomes from OECD and WHO, for example:
– Life expectancy at birth
– Low birth weight
– All-cause mortality
Inequality: Gini coefficient & Palma ratio
Social capital from the European Social Survey
and World Values Survey.
•
Additional data for sub-national analysis of the US came from the Bureau of
Economic Analysis, National Vital Statistics System, Behavioural Risk Factor
Surveillance System, Institute for Health Metrics and Evaluation, RAND State
Statistics.
We observed a significant association
between social expenditure and health Ratio of social
to health as
outcomes (as with Bradley and co.)
well overlaid –
OR JUST SAY IT
Model 1
Model 2
Model 3
life expectancy
log infant mortality
% low birthweight
b
0.05
-0.02
0.02
se
0.01
0.00
0.01
p
0.00
0.00
0.06
b
7.87
-1.12
1.35
se
0.24
0.06
0.21
p
0.00
0.00
0.00
b
-3.33
13.35
-7.77
se
2.48
0.62
2.14
p
0.18
0.00
0.00
319
308
312
0.94
0.95
0.86
Variable
Social spending % of GDP
Log GDP
Intercept
N country-years
rho variation between
countries over time
Results held when we
included more countries,
over longer time periods
and tested multiple health
outcomes
Incorporating a time lag
increased the strength
of relationships
Different ‘types’ of
spending had different
impacts – but not in the
way we expected
Public social expenditure
by governments seems
to have a particularly
strong association with
health outcomes
Public social expenditure by governments
seems to have a particularly strong
relationship with health outcomes
This table shows how many times larger the association between
public-only expenditure was to each outcome, compared to publicprivate expenditure…
Life
expectancy
Infant
mortality
% low
birthweight
All-cause
mortality
3.6
2.3
4.0
3.2
How robust is the
relationship between
social expenditure and
health?
Answer: relationships appear
robust to different approaches.
Does the ‘type’ of social
spending, i.e. on
particular age groups or
social challenges,
influence the findings?
Answer: type of social
spending matters
The focus of our research
Do wider contextual
factors – like inequality
or social capital – affect
the findings?
Do these relationships
hold within a country, in
this case the US?
How robust is the
relationship between
social expenditure and
health?
Answer: relationships appear
robust to different approaches.
Does the ‘type’ of social
spending, i.e. on
particular age groups or
social challenges,
influence the findings?
Answer: type of social
spending matters
The focus of our research
Do wider contextual
factors – like inequality
or social capital – affect
the findings?
Do these relationships
hold within a country, in
this case the US?
Wider contextual factors may also matter
As social spending increases, inequality decreases (and vice-versa)
Higher inequality is associated with an even
stronger association between social
spending and health outcomes
Lower = better
More unequal
So, social protection may be more important for health outcomes in
more unequal societies
How robust is the
relationship between
social expenditure and
health?
Answer: relationships appear
robust to different approaches.
Does the ‘type’ of social
spending, i.e. on
particular age groups or
social challenges,
influence the findings?
Answer: type of social
spending matters
The focus of our research
Do wider contextual
factors – like inequality
or social capital – affect
the findings?
Answer: other wider contextual
factors may matter, particularly in
how they interact with social
expenditure
Do these relationships
hold within a country, in
this case the US?
How robust is the
relationship between
social expenditure and
health?
Answer: relationships appear
robust to different approaches.
Does the ‘type’ of social
spending, i.e. on
particular age groups or
social challenges,
influence the findings?
Answer: type of social
spending matters
The focus of our research
Do wider contextual
factors – like inequality
or social capital – affect
the findings?
Answer: other wider contextual
factors may matter, but they do
not have such a large impact as
social expenditure
Do these relationships
hold within a country, in
this case the US?
Do these relationships hold within a single
country?
States that spent more
on social programmes
had better health
outcomes.
(Again, measured over
time.)
So,
where to
now?
A key limitation of the study is that we observe
correlations, not causal relationships…and we
focus on aggregate, not individual, outcomes
Need to continue building our understanding of how individual
circumstances and social contexts influence health outcomes
Broaden analysis to
incorporate other factors, eg
education, immigration,
ethnicity, race, poverty
Deepen analysis to examine
the role of the design and
implementation of social
programmes
Need to evaluate real-life experiments
of the impacts of social programmes,
on individuals, and think more about
‘what works, for whom, and why?’
These findings appear to highlight associations
strong enough and persistent enough to merit
further investigation:
What is the transmission mechanism?
Should governments explore shifting expenditure from health
systems to prevention and social programmes?
How politically feasible is this, given the tendency to ring-fence
health budgets but not welfare and wider social programmes?
If so, what would an optimal ratio be?
What about addressing wider contextual factors?
What could be done to increase levels of trust alongside any shifts in
expenditure?
And reducing inequality?
Others factors that merit attention?
Palma ratio results
Model structure
We used two modelling approaches:
𝑌𝑖𝑡 = 𝛼0 + 𝑏1 𝑥1𝑖𝑡 + 𝑏2 𝑧2 𝑖 + 𝜇0𝑖 + 𝜀𝑖𝑡 (1)
∆𝑌𝑖𝑡 = 𝛼0 + ∆𝑏1 𝑥1𝑖𝑡 + ∆𝜀𝑖𝑡 .
(1) Is a mixture of variation between and within countries.
(2) Is a within-country only approach.
(2)
Different ‘types’ of spending had different
impacts – but not in the way we expected
What next questions continued…
OECD SOCX dataset
Wider contextual factors may also matter
As social spending increases, trust increases – or is it the other way
around?