Transcript Determinants of Household Saving in China

Determinants of
Household Saving in
China
Marcos Chamon
Eswar Prasad
Disclaimer: The views expressed are those of the authors and do not necessarily
represent those of the IMF or IMF policy.
Motivation

Chinese households save a lot!
 About
25% of disposable income
Historically, households main contributor
to national savings
 Recently, enterprises have become largest
savers
 But household savings are still large:
about 16% of GDP

Savings as a percentage of GDP
50%
Gross domestic savings
40%
30%
Household savings
20%
Household savings
(based on Modigliani and Cao 2004)
10%
Enterprise savings
Government savings
0%
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
year
High household saving rate
somewhat puzzling
High enterprise savings can be justified by
attractive returns on retained earnings
 But households typically face small real
returns on their savings (sometimes
negative!)
 Moreover, rapid income growth suggests
households should be anticipating future
consumption/delaying their life-cycle
savings

Figure 9. Nominal and Real Interest Rates - 1 Year Deposit
12.5
Nominal
10
7.5
5
%
2.5
0
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
-2.5
-5
Real
-7.5
-10
-12.5
Year
Source: CEIC
Notes: Nominal one-year deposit rates (annual average) deflated by annual change in CPI (one year ahead)
2003
2004
Overview of presentation
Paper uses household-level data from a
subset of the Urban Household Survey
 Focuses on three determinants of savings:
1) Life-cycle effects
2) Transition effects from reform process
3) Credit constraints (durable good
purchases)

Life-cycle effects

In a fast growing economy people should:
 Borrow
against future income
 If credit constrained, at least delay
“retirement” savings

Paper presents a very simple OLG model
showing that interplay of credit constraints
and high income growth can actually
increase savings
Model set-up
 Agents
live for 3 periods, earn wages in first
two periods
 All wages in the economy grow at a geometric
rate g>1 every period:
Cohort born at t=0: w0=1, w1=g, w2=0
 Cohort born at t=1:
w1=g, w2=g2, w3=0
 Cohort born at t=2:
w2=g2, w3=g3, w4=0

can only borrow up to share b of their
second period income in the first period
 Agents
Simple example with no borrowing
(e.g. b=0)

Household born at t=0 has:
 wt=1,

wt+1=g, wt+2=0
If g≤2, household can perfectly smooth its
consumption by consuming:
 ct=(1+g)/3, ct+1=(1+g)/3, ct+2=(1+g)/3

If g>2, household would like to borrow against
future income in first period. Since it cannot, the
best it can do is not to save at t=0. Resulting
consumption path is:
 ct=1, ct+1=g/2, ct+2=g/2
With borrowing constraints, income
growth increases savings
1/ 3 if g  2

Share of human wealth saved for retirement =  g
 2(1 + g ) if g > 2

Aggregating across overlapping cohorts
yields:
 2 g 3 + 2g + 1
if g  2
3 
2
6g

Aggregate saving rate = 
 1  1 if g > 2
 4 4g
0
.05
.1
.15
.2
Aggregate savings rate in an OLG economy
as a function of growth rate of wages
1
2
3
Gamma
4
5
Relaxing borrowing constraints
(b>0 but still small) yields (for g>2)
 (1  b )g 1 
Share of wealth saved for retirement = max 
, 
 2(1 + g ) 3 
Aggregating across overlapping cohorts
yields:
 2 g 3 + 2g + 1
2

if
2

g


6g 2
1  3b
3
Aggregate saving rate = 
(1  b )  1  1   b (g  1) if g > 2



4
4
g
1  3b

 2
0
.05
.1
.15
.2
Aggregate savings rate in an OLG economy
as a function of growth rate of wages and
borrowing constraints
1
2
3
Gamma
Beta=0
Beta=5%
4
Beta=2.5%
Beta=7.5%
5
Empirical Evidence on life-cycle
effects
Use data from urban household survey.
Entire sample for 1986-1992, subset of 10
provinces/municipalities for 1993-2001.
 Limit analysis to households whose head
between 25 and 70 years old

Summary of Urban Household
Survey
Year
Observations
Income Per
Capita
(in 2005 RMB)
Consumption
per Capita
(in 2005 RMB)
Average Saving
Rate
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
11877
12700
13364
12806
13380
13508
16561
5992
6151
6159
6157
6144
6130
6135
5849
6047
3348
3450
3433
3447
3805
4008
4337
5086
5535
5686
5884
6091
6558
7199
7620
8186
2984
3038
3203
3036
3228
3442
3615
4306
4655
4828
4906
5078
5439
5904
6282
6504
10.9%
11.9%
6.7%
11.9%
15.1%
14.1%
16.6%
15.3%
15.9%
15.1%
16.6%
16.6%
17.1%
18.0%
17.5%
20.6%
Age and cohort effects
Following Deaton and Paxson (1994), we
compute average log(income) and
log(consumption) for each age*year
combination and regress on age, cohort
(age in 1986) and year dummies
 There is a linear relationship between age,
cohort and year. Year effects are
constrained to:

 Add
to zero
 Be orthogonal to a time trend
9.5
10
10.5
11
11.5
12
Age effects on income and consumption
20
30
40
50
60
Age
Log Y
Log C
Effects shown for household that was 10 years old in 1986
70
7
8
9
10
Cohort effects on income and consumption
0
20
40
Age in 1986
Log Y
60
Log C
Effects shown for 25 year old household
80
-.2
0
.2
.4
.6
Age and cohort effects on savings
0
20
40
Age and Age in 1986
Age effects
Cohort effects
Effects shown for 25 year old household in 2001
60
80
Time trend on income overwhelms
all other effects

Alternative approach:
 Give
up trying to identify cohort effects, and
regress log (income) and log(consumption) on
age dummies and unrestricted time trend
9.4
9.6
9.8
10
10.2
Age effects on income and
consumption
20
30
40
50
Age
Log Y
Effects shown for 2001
Log C
60
70
.22
.24
.26
.28
.3
Age profile of savings
20
30
40
50
Age
Effects shown for 2001
60
70
Qualitative results match our priors
Young households save substantially
(possibly to self-finance purchases of
durables)
 Savings increase sharply around mid 40s
(suggesting “retirement savings” begin
around that age)

Implications for future aggregate
saving patterns: Demographics
 In
the long run, population aging should lead
to a contraction in aggregate savings
 Share
of population in “prime saving” age
group will increase vis-à-vis “prime dissaving”
group in the short- and medium-term
Share of Chinese population by age group
45%
40%
1985
1995
2005
2015
2025
2035
2045
35%
30%
25%
20%
15%
10%
5%
0%
0-19
20-34
35-49
Age Range
50-64
65+
Precautionary Saving motives
Many observers emphasize role of
precautionary motives and uncertainty
related to reforms
 Several benefits traditionally provided by
State Owned Enterprises to their
employees:

 Health;

Education; Pensions; Housing;...
Provision of these benefits either lost or
became uncertain
Precautionary saving motives


Households may be saving a lot not only
because of higher uncertainty, but also to makeup for past savings that were not made
Different groups affected differently by this
uncertainty:
 SOE
workers have potentially a lot to lose vs
collective enterprise workers that didn’t have many
benefits to begin with
 Private sector workers face uncertainty but may also
face better income growth prospects
Percentage of households by type
of employer of head of household
Type of unit
SOEs
Collective Units
Other types of units
(including private)
Entrepreneurs
Employees of
individuals
Re-employed
retirees
Other employed
Retirees and others
1993
67.7%
13.1%
1994
66.4%
11.3%
1995
67.4%
10.3%
1996
67.2%
10.6%
1997
66.8%
10.3%
1998
65.2%
9.6%
1999
63.2%
10.0%
2000
59.4%
8.0%
2001
58.3%
7.3%
0.5%
0.4%
1.4%
0.5%
1.3%
0.5%
1.2%
0.7%
1.8%
1.0%
2.3%
1.3%
3.0%
1.6%
3.8%
2.6%
4.3%
3.0%
0.3%
0.2%
0.3%
0.2%
0.2%
0.3%
0.4%
1.1%
1.8%
3.3%
0.1%
14.6%
3.6%
0.1%
16.5%
3.4%
0.2%
16.6%
3.3%
0.1%
16.6%
2.9%
0.1%
16.8%
3.0%
0.2%
18.2%
3.4%
0.3%
18.0%
2.6%
0.5%
22.0%
2.7%
0.5%
21.9%
Estimated effect of employer type
on saving rates
Type of unit
SOEs
Collective Units
Other types of units
(including private)
Entrepreneurs
Employees of individuals
Re-employed retirees
Other employed
Dummy for spouse
working at SOE
1993-1997
-0.018
(0.006)**
1998-2001
-0.043
(0.008)**
1993-1997
-0.018
(0.006)**
-0.015 (0.012)
0.075
(0.022)**
-0.065 (0.034)
-0.020 (0.013)
-0.03 (0.010)**
-0.015 (0.012)
0.075
(0.023)**
-0.066 (0.034)
-0.020 (0.013)
-0.025 (0.010)*
-0.100 (0.051)
-0.106 (0.042)*
.000 (0.004)
0.012 (0.005)*
-0.100 (0.051)
-0.005 (0.016)
0.004 (0.024)
-0.036 (0.015)*
-0.111
(0.042)**
1998-2001
-0.04 (0.009)**
-0.001 (0.016)
0.009 (0.024)
-0.034 (0.015)*
Implications for future aggregate
saving patterns: Transition effects
 Shift
to a market economy and SOE reforms
likely contributed to the increase in household
savings
 The effect may weaken over time:
As households continue to accumulate savings, at
some point they will have enough assets to protect
them from most adverse shocks
 Eventual development of social safety net and
pension system should also lower savings

Durable goods and borrowing
constraints
Consumer finance very limited in China
 Development of consumer credit should
lower savings
 But magnitude of the effect may be small:

 If
household saves 20% of income and wants
to buy a new TV, it can do so just by saving
less.
 No need to rely on credit or even deplete past
savings!
Durable good consumption




Survey has detailed data on income and consumption
expenditures. We focus on 1993-2001 subsample
Exclude households with home purchasing/construction
expenditures (about 8% of households)
Durable good purchases correspond on average to 6.5%
of income (but distribution is very skewed due to their
“lumpiness”)
Durable good purchases exceed income minus other
expenditures for 33% of households (thus cannot finance
purchase just by saving less)
Financing sources for durable good
purchases

We break down the source of funds for
durable good purchases between:
(i) Income – nondurable consumption –
nonconsumption expenditures
(ii) Net financial dissavings (e.g. net saving
withdrawals)
(iii) Credit
Financing sources for durable good
purchases in 2001
Income
Durable consumption
Income - other expenditures
Net financial dissavings
Credit
Per capita income (in 2005
RMB)
Number of households
Age of head
Durable
Consumption/
Consumption
Below
Above
10%
10%
Below
Median
Above
Median
Below
35
35 or
older
4.5%
9.2%
-6.6%
1.8%
7.5%
23.2%
-16.5%
0.7%
7.2%
19.3%
-13.3%
1.3%
6.1%
17.1%
-12.2%
1.2%
1.4%
16.2%
-15.9%
1.1%
22.3%
21.1%
-0.4%
1.6%
5134
3901
13341
2106
7497
686
8078
5321
7533
4926
10191
1081
Note: Variables expressed as share of income unless otherwise noted. Negative
net financial dissavings indicates households net financial savers
Are net financial dissavings related
to large durable purchases?

We run probit regressions of a dummy
equal to one if household is net financial
dissaver (about 30% of households) on:
 Log
(Durable good purchases/Y)
 Log Y
 Dummy for household head below 35
Probit regression results
Probability of Being Net Financial Dissaver
All years
Durable Consumption/
Income
I
II
III
IV
3.569
4.201
3.514
4.225
(0.064)**
(0.069)**
(0.194)**
(0.208)**
Log (Income)
Age Below 35
Constant
2001
Year
Dummies
-0.423
-0.479
(0.012)**
(0.032)**
-0.103
-0.243
(0.017)**
(0.059)**
Year Dummies
-0.752
3.439
(0.020)**
(0.282)**
2
R
N
.064
.086
.059
.092
51746
51746
6007
6007
Marginal effects on probability of being
a net financial dissaver in 2001
All variables at their means
Pr(Net Financial
Durables / Income
Age Below 35
Log (Income)
Dissaver)
.278
1.415
-.077
-.160
All variables at their means, Durable/Income at 10%
.350
1.567
-.086
-1.773
All variables at their means, Durable/Income at 20%
.515
1.684
-.096
-1.908
Large durable purchases increase
likelihood of net financial dissavings

Magnitude of the effect non-negligible, but
relatively small
 Households
likely to remain net financial
savers even when making very large durable
purchases
Ownership of most durable goods
common except cars
Durable Good
2000
2005
Washing Machine
Refrigerator
Color TV
DVD Player
Mobile Phone
Automobile
90.8
80.5
116.7
37.1
18.3
0.6
95.5
90.7
134.8
68.1
137
3.4
Source: CEIC based on NBS data covering whole sample
Home ownership rates also very high
Year
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
10 Province/Municipalities
Sub-Sample
0.206
0.283
0.309
0.355
0.477
0.554
0.647
0.727
0.767
National Average
0.557
0.620
0.689
0.771
0.810
0.820
0.830
0.843
Implications for future aggregate
saving patterns
Development of consumer financing migth
only have limited impact on saving
behavior (with possible exception of auto
financing)
 Developments in housing market should
also have limited impact given very high
rates of home ownership

Conclusion




Precautionary saving motives seem to play an
important role
Demographic changes have contributed to
aggregate savings (high income growth
leverages that effect).
Demographics should continue to contribute to
aggregate savings over the next two decades
Developments in consumer credit may not have
a substantial effect on savings