data from macroeconomic consequences of the demographic transition ronald lee uc berkeley july 9,...

Data from

Macroeconomic Consequences of the Demographic Transition

Ronald LeeUC BerkeleyJuly 9, 2008

Talk prepared for Rand Summer InstituteResearch supported by NIA R37 AG025247

Thanks to Andy mason and NTA country teams

Data from

Plan

• Demographic transition

• Dependency ratios and support ratios

• Savings rates and capital intensification

• Human capital

Data from

I. The Demographic Transition

• A classic illustration: The transition in India, 1890-2100.

• Mixture of historical estimates, UN projections, and simulation based on fitted variations with time.

Data from

Data from

Pre fertility decline; child dependency ratio rises During fertility decline, child

dependency ratio declines

Population aging: old age dep ratio rises

Data from

The total dep ratio rises, falls, then rises again, ending up where it started.The changes in the total dependency ratio are transitory.

Data from

But there is a big permanentchange: At start, many children and few elderly.At end, few children and Many elderly.

Data from

Comments on simulation

• Assumed TFR stabilized at 2.1; but often has declined below replacement.

• Assumed e0 stopped rising at 80, but many countries already above this.

• Some countries experienced important baby booms and busts which distort this classic shape.

• Many countries now have declining populations and declining working age pops.

Data from

II. The economic life cycle:

• Age profiles of consumption and labor income

• Use estimates from the National Transfer Accounts project, or NTA.

• Consumption patterns are quite similar for Third World countries in Asia and Latin America.

• Consumption in Industrial populations looks different.

Data from An-Chi Tung

A Typical Asian Economic Lifecycle: National Transfer Accounts estimates for Taiwan, 1998

0

100

200

300

400

500

600

0 20 40 60 80

Age

Per

Cap

ita

Co

nsu

mp

tio

n a

nd

L

abo

r In

com

e

Consumption

Labor Income

Includes both private expends and in-kind public transfers (health, education, long term care)

Includes self employment, wages,unpaid family labor, & fringe benefits.Averages 0’s and both male and female.



0

100

200

300

400

500

600

0 20 40 60 80

Age

Per

Cap

ita

Co

nsu

mp

tio

n a

nd

L

abo

r In

com

e

Consumption

Labor Income

Flat cons age profile in adult years reflects extended family sharing.

Quite different than most industrial nations.



0

100

200

300

400

500

600

0 20 40 60 80

Age

Per

Cap

ita

Co

nsu

mp

tio

n a

nd

L

abo

r In

com

e

Consumption

Labor Income

Large deficits at young and old ages.



0

100

200

300

400

500

600

0 20 40 60 80

Age

Per

Cap

ita C

on

su

mp

tio

n a

nd

Lab

or

Inco

me Consumption

Labor Income

Reallocations from surplus to deficitages required.



0

100

200

300

400

500

600

0 20 40 60 80

Age

Per

Cap

ita C

on

su

mp

tio

n a

nd

Lab

or

Inco

me Consumption

Labor Income

Other income comes from assets, foreign loans, and remittances from abroad—its not all labor income.



0

100

200

300

400

500

600

0 20 40 60 80

Age

Per

Cap

ita

Co

nsu

mp

tio

n a

nd

L

abo

r In

com

e

Consumption

Labor Income

Asset income is particImpt for old age

Data from

Components of US Consumption, 2003

Public Other

Private Other

Private DurablesPrivate Health

Private Edu

Public Edu

Public Health

0

20000

40000

0 10 20 30 40 50 60 70 80 90

Age

Do

llar

s (U

S,

2000

)

Later I will measure HK investmentAs sum of pub and priv spendingon hlth and educ as shown here.

Unlike Taiwan and other Third World, in US cons rises strongly with age. True in other industrial too.

Data from

• Levels of age profiles change fast with economic development.

• Shapes of age profiles change slowly,

• Are broadly similar across countries at very different levels of development.

Data from

III. Dependency and Support

• Concern about pop aging is mostly about old age dependency.

• Sharpest concerns for age-sensitive public sector programs– pensions– health care– Long term care

• But should place these in broader context– Full range of public programs– Private consumption

• Use shape of estimated profile I just showed.

Data from

Support Ratios

• Effective labor is weighted sum of pop using labor income age profile.

• Effective consumers is similar.• Ratio of effective labor to effective consumers is

the “Support Ratio”.• Other things equal, consumption per effective

consumer is proportional to the support ratio.

0

0

Effective WorkersSupport Ratio

Effective ConsumerslP x y x

P x c x

Data from

Support Ratio for China, 1950-2100, Based on UN population projections and average LDC age profiles from NTA

0.5

0.6

0.7

0.8

0.9

1

1950 1970 1990 2010 2030 2050 2070 2090

Year

Eff

ecti

ve

Pro

du

cers

Per

Co

nsu

mer

2007

Population aging

First Dividend

Data from

Support Ratios for Five Less Developed Countries, 1950-2100, Based on UN population projections and average LDC age profiles from NTA

India

Brazil Niger

ChinaS. Korea

0.5

0.6

0.7

0.8

0.9

1

1950 1970 1990 2010 2030 2050 2070 2090

Year

Eff

ecti

ve

Pro

du

cers

Per

Co

nsu

mer

20072008

Data from

Support Ratios for Five Less Developed Countries, 1950-2100, Based on UN population projections and average LDC age profiles from NTA

India

Brazil Niger

ChinaS. Korea

0.5

0.6

0.7

0.8

0.9

1

1950 1970 1990 2010 2030 2050 2070 2090

Year

Eff

ecti

ve

Pro

du

cers

Per

Co

nsu

mer

2007

Niger S. Korea China India Brazil2050/08 1.20 0.78 0.86 1.09 0.96

Rate %/yr 0.43 -0.59 -0.35 0.22 -0.09

Data from

Spain

Italy

US

Japan

Germany

Spain, Low Fert.

Italy, Low Fert.

0.5

0.6

0.7

0.8

1950 1970 1990 2010 2030 2050 2070 2090

Year

Eff

ecti

ve P

rod

uce

rs P

er C

on

sum

erSupport Ratios for Five More Developed Countries, 1950-2100, based on UN long term population projections and the NTA age profile for the US.

Data from

Spain

Italy

US

Japan

Germany

Spain, Low Fert.

Italy, Low Fert.

0.5

0.6

0.7

0.8

1950 1970 1990 2010 2030 2050 2070 2090

Year

Eff

ecti

ve P

rod

uce

rs P

er C

on

sum

erSupport Ratios for Five More Developed Countries, 1950-2100, based on UN long term population projections and the NTA age profile for the US.

US Spain Italy Japan Germany2050/08 0.91 0.72 0.75 0.75 0.81

Rate %/yr -0.2 -0.8 -0.7 -0.7 -0.5

Data from

Proportionate Changes in the Support Ratio from 2007 to 2050 for Selected MDC and LDC

Proportionate Changes in the Support Ratio from 2007 to 2050 for Selected MDC and LDC

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

US Spain Italy Japan Germany Niger S. Korea China India Brazil

Country

Pro

po

rtio

nat

e C

han

ge

in S

up

po

rt R

atio

fro

m 2

007

to 2

050

Data from

Not written in stone. Many policy possibilities:

• Change the age profile of labor income– Later retirement– Earlier entry into labor force– Higher female labor force participation– Reform seniority system

• Change the age profile of consumption– In many industrial nations, the elderly consume much

more than younger adults.– Makes population aging more costly– Role of public transfer policy: pensions, health care,

long term care• Change the demographic trends: immig, fert

Data from

IV. Further on Interage Flows of Income

• Comparison of Japan and Indonesia

Data from Maliki (Indonesia) and H. Ogawa (Japan)

Per capita consumption and labor income by age for Indonesia and Japan

-

200,000

400,000

600,000

800,000

1,000,000

0 20 40 60 80 100

Age

Per c

apita

co

nsum

ptio

n or

labo

r in

com

e

-

100,000

200,000

300,000

400,000

500,000

0 20 40 60 80 100

Age

Per

cap

ita c

onsu

mpt

ion

or la

bor

inco

me

in Y

en

Indonesia, 2002

Japan, 2004

• Differences in consumption– Education in Japan– Rising consumption

in old age in Japan


Aggregate Life Cycle Deficit for Indonesia (2005) in Rupiah

(30,000)

(20,000)

(10,000)

-

10,000

20,000

30,000

40,000

0 20 40 60 80 100

Age

Ag

gre

gat

ed C

on

sum

pti

on

-

Lab

or

Inco

me

Aggregate Life Cycle Deficit for Japan (2004) in Yen

(5,000)

(4,000)

(3,000)

(2,000)

(1,000)

-

1,000

2,000

3,000

4,000

5,000

0 20 40 60 80 100

Age

Ag

gre

ga

ted

Co

ns

um

pti

on

- L

ab

or

Inc

om

e


Aggregate Life Cycle Deficit for Indonesia (2005) in Rupiah

(30,000)

(20,000)

(10,000)

-

10,000

20,000

30,000

40,000

0 20 40 60 80 100

Age

Ag

gre

gat

ed C

on

sum

pti

on

-

Lab

or

Inco

me

Aggregate Life Cycle Deficit for Japan (2004) in Yen

(5,000)

(4,000)

(3,000)

(2,000)

(1,000)

-

1,000

2,000

3,000

4,000

5,000

0 20 40 60 80 100

Age

Ag

gre

ga

ted

Co

ns

um

pti

on

- L

ab

or

Inc

om

e

• Green arrows show transfers from surplus of prime working years.

• Red arrows show asset income consumed by elderly out of earlier savings.


-

200,000

400,000

600,000

800,000

1,000,000

0 20 40 60 80 100

Age

Per c

apita

co

nsum

ptio

n or

labo

r in

com

e

-

100,000

200,000

300,000

400,000

500,000

0 20 40 60 80 100

Age

Per c

apita

con

sum

ptio

n or

labo

r inc

ome

in Y

en

Indonesia, 2002

Japan, 2004

• In Indonesia, average unit of income is earned at 39 and consumed at 30

• Travels 9 years down the age scale.

• In Japan, it is earned and consumed at nearly the same age.

Ac=30 Ayl=39

Ac=45 Ayl=45

100

0100

0

c

xPop x c xA

Pop x c x

Population weighted average age

Data from NTA Country Teams

Average Consumption-Earning Gap by Average Age of Population

Austria, 2000

Chile, 1997

Costa Rica, 2004

France, 2001

India, 1999Indonesia, 1999

Japan, 2004

Philippines, 1999

US, 2003

Uruguay, 1994Sweden, 2003

Thailand, 2004

Slovenia, 2004

Taiwan, 2003S. Korea, 2000

-12

-10

-8

-6

-4

-2

0

2

4

20 25 30 35 40 45Ave Age of Population

Av

Ag

e G

ap

13 yrs


How much of the difference in age gaps is due to the shapes of the

age profiles?


Average Age of Labor Income and Consumption with Population Held Constant (stationary, e0=75)

S. Korea, 2000Taiwan, 2003

Slovenia, 2004

Thailand, 2004

Sweden, 2003

Uruguay, 1994

US, 2003

Philippines, 1999Japan, 2004

Indonesia, 1999

India, 1999

France, 2001Costa Rica, 2004

Chile, 1997

Austria, 2000

38

39

40

41

42

43

44

45

46

38 39 40 41 42 43 44 45

Average Age of Consumption

Ave

rage

Age

of L

abor

Inco

me

Ac = Ayl


Average Age of Labor Income and Consumption with Population Age Distr. Constant (stationary,

e0=75)


Slovenia, 2004

Thailand, 2004

Sweden, 2003

Uruguay, 1994

US, 2003


Indonesia, 1999

India, 1999


Chile, 1997

Austria, 2000

38

39

40

41

42

43

44

45

46

38 39 40 41 42 43 44 45


Ave

rage

Age

of L

abor

Inco

me

Ac = Ayl

Low age of cons due to heavy spending on education.

High cons (health care) and work when old.

Generous old age support.

Very young Ayl

due to early start, early retirement.









Slovenia, 2004

Thailand, 2004

Sweden, 2003

Uruguay, 1994

US, 2003


Indonesia, 1999

India, 1999


Chile, 1997

Austria, 2000

38

39

40

41

42

43

44

45

46

38 39 40 41 42 43 44 45


Ave

rage

Age

of L

abor

Inco

me

Ac = Ayl

Difference in average ages is the distance above (-) or below (+) the diagonal.




Slovenia, 2004

Thailand, 2004

Sweden, 2003

Uruguay, 1994

US, 2003


Indonesia, 1999

India, 1999


Chile, 1997

Austria, 2000

38

39

40

41

42

43

44

45

46

38 39 40 41 42 43 44 45


Ave

rage

Age

of L

abor

Inco

me

Ac = Ayl

Indonesia -3.3 yrs

Austria +2.2 yrs

Data from

• total range in age gap was 13 years

• range due to differences in profiles is 5.5 years.

• So both population age distribution and shapes of age profiles help determine gap.

Data from

V. Wealth and the age gap: the golden rule case

• Demographic and economic steady state

• Saving and capital such as to maximize per capita consumption.

• r=n+g

Data from

Now suppose babies had to go into debt to feed themselves….

• At the start of life, c(x)>yl(x); dependency. • Suppose we keep a notional account of debt and

credit over the life cycle, discounted to age 0. • Credit gained (or lost) at age x is:

e-rx l(x) [yl(x)-c(x)]where r is interest rate, l(x) is survival from 0 to age x.

• Cumulated up to age x, we get W(x):

0

x rxlW x e l x y x c x dx

Data from

The average wealth per capita in the population may be pos or neg

• Now find the average level of per capita in the whole population, call it W

• W = pop(x)*W(x)/totpop

0

0

rx

rx

e l x W x dxW

e l x dx

Data from

The Willis result

W = c(Ac – Ay) , where c = per capita cons– If Ac>Ay then indivs need to hold onto some

output for later consumption, so wealth, W, is on average positive in the population.

– If Ac<Ay then indivs consume before they produce, and must go into debt on average, so W is negative.

• Alternatively: W/c = Ac – Ayl

• So wealth relative to consumption is roughly proportional to Ac - Ayl

Data from

• Given comparative analysis of Ac-Ayl, suggests that demand for wealth rises over the demographic transition.

• Why?– Older people hold more wealth; in old

population, there are more of them.– Longer life means workers need to

accumulate more wealth for longer old age.– Lower fertility means adults consume more

and need to save more to maintain in old age.

Data from

VI. The role of intergenerational transfers

• We just considered the wealth needed to achieve consumption targets.

• Wealth can be held in two forms: – Transfer wealth (expected future transfers

received minus expected future transfers made)

– Assets or Capital

Data from

NTA data on shares of old age support from different sources

• Asset income (land, equities, interest, etc.)

• Family transfers (not including bequests at death)

• Public transfers (Pay As You Go pensions, health care, and long term care)

• Triangle graph shows shares, not levels, so must add to 100%.

• Bequests not included; just old age cons.

Diagram from Andy Mason

Old-age Reallocation System, Selected Countries.

0

50

75

100

75

75

50

50

25

25

0

0

25

100

100

Asset-based (%)

Public transfers (%)

Family Transfers (%)

US

Thailand

Costa Rica

JapanTaiw an

Korea

Familial transfers equally important in Thailand, Korea, and Taiwan (36-

40%). Net familial

transfers near zero in US, CR, and J. Large

public transfers in CR and J

Net public transfers to elderly are zero in

Thailand; about 25% in Taiwan and Korea.



0

50

75

100

75

75

50

50

25

25

0

0

25

100

100

Asset-based (%)



US

Thailand

Costa Rica

JapanTaiw an

Korea

Public transfers: Thailand none,

Japan and Costa Rica around 70%

US, Korea, Taiwan, middling



0

50

75

100

75

75

50

50

25

25

0

0

25

100

100

Asset-based (%)



US

Thailand

Costa Rica

JapanTaiw an

Korea

Reliance on assets : Japan, Taiwan, C.R.

are low; Thailand high; US middling

Data from

VII. Demographic Transition and Capital Accumulation

• Changing dependency gets most attention for ec dev and pop aging.

• Changes in capital accumulation may be more important.

Data from

Calculating the demand for wealth and capital over the demographic transition

• Based on different theoretical models, approaches.

• Model with Social Planner maximizing discounted social welfare function with full foresight.

• Model with individuals saving and consuming over their life cycles to maximize their life time utility.

Data from

Here take a different approach – no optimization--emphasizes institutional

setting• Assume

– share of old age consumption supported by asset income stays constant over time.

– altruistic sharing maintains the shape of the cross sectional consumption age profile.

– Demography is known in advance.

• Can solve recursively for unique growth path and asset holdings.

Data from

Two scenarios: high level of transfers to elderly (65%); or low level (35%)

• Other assumptions– Productivity growth raises income age profile by 2%

per year.– Open economy; rate of return on assets is 3%.

• Aggregate saving is calculated to maintain asset share of old age consumption support.

• Results will be shown relative to a 2% growth trajectory from prod gr.

From Mason, Lee and Lee (2008)

0

0.05

0.1

0.15

0.2

0.25

1940 1960 1980 2000 2020 2040 2060

Net

Savin

g R

ate

High IG Transfers

Low IG Transfers

Simulated Saving Rate, ASEAN (S.E. Asian countries), 1950-2050


Simulated Assets/Labor Income, ASEAN

0

2

4

6

8

1940 1960 1980 2000 2020 2040 2060

Ass

ets/

Labo

r In

com

e .

Low IG Transfers

High IG Transfers

Ratio of assets to labor income rises greatly in any case, but 3 or 4 times as much with low IG transfers.


Simulated Consumption, ASEAN

60

80

100

120

140

160

1940 1960 1980 2000 2020 2040 2060

Co

nsu

mp

tio

n In

dex

(19

50=

100)

.

Low IG Transfers

High IG Transfers

With low IG transfers, saving is higher from 1990 to 2020, reducing consumption.

Thereafter, it is higher.

Data from

These sorts of results are qualitatively like those from optimization approaches

• Timing of swings differs• Level of savings rates differs• Capital/labor income ratios differ

Big picture is the same:1. The demographic transition leads to a major increase

in capital per worker.2. The greater the role of transfers to the elderly, the

smaller is the increase in capital intensity.3. Eventually consumption rises with lower transfers, but

initially it is lower.4. Population aging leads to a decline in savings rates but

an increase in capital intensity.

Data from

VIII. Human capital and the demographic transition

• Measure public and private expenditures on health and education at each age.– Sum these for health ages 0-18– Sum for education ages 0-26– Gives synthetic cohort HK investment per

child

• Construct ratio of HK to average yl(x), ages 30-49.

• Plot log of HK/w against log of TFR.

Data from NTA country teams

Figure 1. Per Child HK Spending (Public and Private) vs. Fertility

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1.80

2.00

0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40

ln(TFR)

ln(H

K p

er C

hild

/Av

Lab

Inc

30-

49)



AustBrz

Chl

CR

Fin

FrHng

India

Indonesia

Jpn

Mex

Phil

Slv

Kor

SwdTwn

Thai

Urg

US

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1.80

2.00

0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40

ln(TFR)

ln(H

K p

er C

hild

/Av

Lab

Inc

30-

49)



0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1.80

2.00

0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40

ln(TFR)

ln(H

K p

er C

hild

/Av

Lab

Inc

30-

49)

y = -1.05*x + 1.92R2 = 0.62

Data from

Now calculate total HK spending on all children

• Multiply TFR times HK per child, and plot its log against log(TFR).


Figure 5. Total Expenditures Per Woman for All Children's HK vs. Fertility

for 18 NTA countries (log scale)

0.00

0.50

1.00

1.50

2.00

2.50

0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40

ln(TFR)

ln(T

FR

X P

er C

hil

d H

K S

pen

din

g/

Av

Lab

In

c 3

0-49

)

Roughly a horizontal cloud, perhaps slightly negative.

6.8 years of labor income are invested in total HK on average.

1/12 of lifetime labor income for a couple.

Data from

Association is non-causal

• We don’t know whether fertility decline causes rising HK investments per child.

• Desire to make bigger HK investments causes fertility decline.

• Some other factor like rising income causes both fertility and HK changes.

• Here is one theory about a causal path from income growth to other changes. In some models the HK growth causes income growth.

Data from

The standard Quantity-Quality model

• Assume that the share of total labor income spent on HK is fixed, consistent with scatter plot.

• Draw budget constraints for differing levels of income.

• Quantity and quality interact multiplicatively in the budget constraint, both with positive income elasticities for constant price.

Data from

The Standard Model: Rising Income Leads to Choice of Lower Fertility and Higher HK Investment per Child

0

1

2

3

4

5

6

7

1 2 3 4 5 6 7 8

Number of children

Hu

man

Cap

ital

In

vest

men

t p

er

chil

d

Yn=1

Yn=6

Yn=4

Data from

The Standard Model: Rising Income Leads to Choice of Lower Fertility and Higher HK Investment per Child

0

1

2

3

4

5

6

7

1 2 3 4 5 6 7 8

Number of children

Hu

man

Cap

ital

In

vest

men

t p

er

chil

d

Yn=1

Yn=6

Yn=4

B

A

C

With same data, plot ln(HK/w) instead of HK, against ln(TFR) instead of n.

The budget lines collapse onto a single straight line.

Data from

Ln(n)

Ln(pqq/w)

A

C

B

D

ln(d)

Figure: The transformed budget constraint showing different quantity-quality choices.

Slope (elasticity) = -1

d is HK expenditure expressed in years of work at rate w

Quite similar to empirical scatter

Intercept of scatter indicates years of work expended on HK is 6.8.

Share of lifetime labor income is 1/12.

Data from

• So our scatter plot shows a common transformed budget constraint with different fertility-HK choices.

• Differing incomes is one possible cause.

• Many others.

Data from

Production and Human capital

• Human capital (HK)– Portion of wage, W(t),

workers invest in their children is inversely related to their fertility, F(t)

– Human capital of workers one period later is

– HK(t+1) = h(F(t)) W(t)

• Wage (W)– Wage is increasing in

human capital– W(t) = g(HK(t))

112

W tHK t

F t

.33W t HK t

Baseline Specifications

Data from

Other sources of variation in fertility/HK choice

• Pref for HK: Rate of return to HK; survival rates; consumption value of HK.

• Price of HK due to medical technology, transportation improvements, etc.

• Price of number: family allowances, fines for second child, changing access to effective contraceptives

• Cultural influences on varying share of income allocated to total HK expenditures and on number.

Data from

Model—basic structure

• Take fertility variations as given, trace out consequences for HK, w, consumption.

• 3 generations: children, workers, retirees; usual accounting identities.

• No saving or physical capital.

• HK drives wage growth; wage growth drives HK growth. (Lee and Mason 2008)

From Lee and Mason (2008)

Figure 6. Macro Indicators: Baseline Results

60.0

70.0

80.0

90.0

100.0

110.0

120.0

130.0

0 1 2 3 4 5 6

Period

Va

lue

(p

erc

en

t o

f y

ea

r 0

)

Support ratio

C/ EA

Boom (demoraphic

dividend)

Fertility bust, but consumption remains high

Fertility recovers: modest effect on C/EA

Bottom line: Low fertility leads to higher consumption. Human capital investment has moderated

the impact of fertility swings on standards of living.

From Lee and Mason (2008)

Figure 6. Macro Indicators: Baseline Results

60.0

70.0

80.0

90.0

100.0

110.0

120.0

130.0

0 1 2 3 4 5 6

Period

Va

lue

(p

erc

en

t o

f y

ea

r 0

)

Support ratio

C/ EA

During first dividend phase, consumption does not rise as much as support ratio.

The difference is invested in HK.

That is why ih later periods, consumption is proportionately higher than the support ratio.

Data from

Conclusions for changes over the transition

• Support ratios change over demographic transition; ending where started, roughly. – Importance in long view may be exaggerated. – In shorter view, we are in painful payback phase.

• Bigger effect is on capital intensity– These raise productivity per worker– Raise wealth and asset income

• Increased human capital results from low fertility—so closely related to aging: same cause for both.– Raises productivity.

• However, increased demand for wealth can be met either by increased asset holdings or through increased transfer wealth.

• Major role for policy and institutions at every point; nothing inevitable.

data from macroeconomic consequences of the demographic transition ronald lee uc berkeley july 9,...

Documents