Vietnam: The Next Asian Tiger?Findings I Vietnam has grown impressively since 1986, but is still a...

Vietnam: The Next Asian Tiger?

Tom Barker1, Murat Ungor2

1Macro Financial DepartmentReserve Bank of New Zealand

2Department of EconomicsUniversity of Otago

3 July 2018

The views expressed in this paper are those of the authors and do notnecessarily reflect the views of the Reserve Bank of New Zealand.

1 / 66

Real GDP per capita indices

0 6 12 18 24 30 360

350

700

1050

1400

Years since base date

Japan (1950=100)

2 / 66

Real GDP per capita indices

0 6 12 18 24 30 360

350

700

1050

1400

Years since base date

Japan (1950=100)

Four Tigers (1960=100)

3 / 66

Real GDP per capita indices

0 6 12 18 24 30 360

350

700

1050

1400

Years since base date

China (1978=100)

Japan (1950=100)

Four Tigers (1960=100)

4 / 66

Real GDP per capita indices

0 6 12 18 24 30 360

350

700

1050

1400

Years since base date

China (1978=100)

Japan (1950=100)

Four Tigers (1960=100)

Latin America (1980=100)

5 / 66

Real GDP per capita indices

0 6 12 18 24 30 360

350

700

1050

1400

Years since base date

China (1978=100)

Japan (1950=100)

Vietnam (1986=100)

Four Tigers (1960=100)

Latin America (1980=100)

6 / 66

This paper

I This is the first detailed study conducted on Vietnam’sconvergence experience using newly available data from thePenn World Table (PWT) version 9.0.

I This is the first study to employ a two-sector generalequilibrium model of structural transformation looking at thesectoral allocation of Vietnam’s production factors and thesubsequent implications for aggregate productivity withalternative counterfactual experiments.

7 / 66

Findings

I Vietnam has grown impressively since 1986, but is still arelatively poor country in absolute terms.

I The improvement in GDP per capita since Doi Moi can largelybe attributed to progress made in labour productivity.

I Vietnam’s expansion has occurred with a notable absence oftechnological advancement since 1997. Given this fact, pastrates of economic growth are unlikely to be sustained if theyfail to improve their productive capacity and technologicalproficiency.

I A negative contribution from the TFP component acted as aheadwind to growth in output per worker during 1997-2014.There was a positive TFP growth before Doi Moi and anegative TFP growth after the first decade of Doi Moi.

8 / 66

Findings

I Growth in agricultural labour productivity plays a moreimportant role than growth in non-agricultural labourproductivity in Vietnam’s deagriculturalization experience.

I A significantly higher growth in aggregate labour productivitywould have been accomplished by feeding both sectors’productivity growth rates observed in China.

I The findings of our counterfactual experiments regarding theaggregate labour productivity growth in Vietnam can beconsidered in line with Johnston and Mellor (1961, p. 590):“[I]t is our contention that ‘balanced growth’ is needed in thesense of simultaneous efforts to promote agricultural andindustrial development”.

9 / 66

An overview of the Vietnameseeconomy

10 / 66

Real GDP per capita relative to the U.S. (%)

1950 1961 1972 1983 1994 2005 20160

2

4

6

8

10

%12

American military

involvement

Socialism

trialed

Doi Moi

enacted

Asian

Financial

Crisis

Ascension

to WTO

11 / 66

Peer Countries

China

I as the paragon emergingeconomy

ASEAN “Majors”

I ASEAN is a political andeconomic organisation of tenSoutheast Asian countriesestablished in 1967.

I The “Majors” refers to thesix largest economies in thearea: Indonesia, Thailand,Malaysia, the Philippines,Singapore, and Vietnam.

12 / 66

Peer Countries

China

I as the paragon emergingeconomy

ASEAN “Majors”

I ASEAN is a political andeconomic organisation of tenSoutheast Asian countriesestablished in 1967.

I The “Majors” refers to thesix largest economies in thearea: Indonesia, Thailand,Malaysia, the Philippines,Singapore, and Vietnam.

13 / 66

Peer Countries

China

I as the paragon emergingeconomy

ASEAN “Majors”

I ASEAN is a political andeconomic organisation of tenSoutheast Asian countriesestablished in 1967.

I The “Majors” refers to thesix largest economies in thearea: Indonesia, Thailand,Malaysia, the Philippines,Singapore, and Vietnam.

14 / 66

A comparison of GDP per capita growth in Asia

0 3 6 %90

3

6

%9

Bangladesh

Cambodia

China

Hong Kong

India

Indonesia

Japan

MalaysiaMyanmar

Pakistan

Philippines

Singapore

South Korea

Sri Lanka

Taiwan

Thailand

Vietnam

(a): Asian Countries

1950-1985

19

86

-20

16

0 3 6 %90

3

6

%9

Vietnam

Indonesia

Thailand

Singapore

Malaysia

Philippines

(b): ASEAN Six Majors

1950-1985

19

86

-20

16

15 / 66

Delving into GDP per capita

16 / 66

Decomposing GDP per capita

(Y /P)t︸ ︷︷ ︸GDP per capita

= (Y /L)t︸ ︷︷ ︸Labour productivity

× (L/WP)t︸ ︷︷ ︸Employment rate

× (WP/P)t︸ ︷︷ ︸Demographics

Taking logarithms of both sides allows us to decompose theaverage annual growth rate of output per capita over z years (fromtime t to time t + z), where ln is the natural logarithm operator:

ln(YP

)t+z

− ln(YP

)t

z=

ln(YL

)t+z

− ln(YL

)t

z+

ln(

LWP

)t+z

− ln(

LWP

)t

z

+ln(WPP

)t+z

− ln(WPP

)t

z.

17 / 66

GDP per capita decomposition in Vietnam

Average annual changes, %

Period Y /P Y /L L/WP WP/P1970-1985 2.5 1.5 0.4 0.61986-2014 5.0 4.1 0.01 0.81970-2014 4.0 3.1 0.2 0.7

Source: PWT 9.0 and WDI (World Bank, 2017).

I During 1970-1985, per capita income grew at 2.5% per yearand output per worker increased 1.5% per year.

I The expansion in output per worker made up 61.2% of theincrease in per capita income between 1970 and 1985.

I GDP per capita grew 5.0% per year on average over the1986-2014 period.

I Labour productivity improvements accounted for 83.0% of thisgrowth.

18 / 66

Demographic dynamics in Vietnam

I Vietnam is currently enjoying a golden population structure,in which the working age population is nearly double thedependent population.

I The percentage of the population aged between 15 and 64rose from around 56% in 1986 to slightly over 70% in 2014.

I This is reflected in our results, with an increase in theworking-age population to total population ratio contributingroughly 15% of GDP per capita growth, on average, between1986 and 2014.

I This positive contribution is known as the “demographicdividend,” or demographics that promote economic growth(see Bloom et al., 2010).

19 / 66

Demographic dynamics in Vietnam, 1960-2015

1960 1988 201640

50

60

70

80

90

%100

(a): Two ratios

Dependency

Support

1960 1988 20160

10

20

30

40

50

60

70

80

90

%100

(b): Dependency

Old-age

Youth age

1960 1988 2016

2.6

3.6

4.6

5.6

6.6

(c): TFR

20 / 66

A comparison: GDP per capita decomposition (in PPP)

1970-1985 1986-2014

Average annual changes, % Average annual changes, %

Country Y/P Y/L L/WP WP/P Country Y/P Y/L L/WP WP/P

Vietnam 1.9 1.0 0.4 0.6 Vietnam 5.6 4.7 0.01 0.8ASEAN 4.2 3.1 0.5 0.6 ASEAN 4.4 4.0 -0.1 0.5China 3.8 2.3 0.6 0.9 China 6.0 5.6 -0.2 0.5

Source: PWT 9.0 and WDI (World Bank, 2017).

I Vietnamese GDP per capita expanded at an average annualrate of just 1.9% between 1970 and 1985. This was a poorperformance when compared to their East Asian counterparts.

I After the implementation of Doi Moi in 1986, Vietnamembarked on a high growth period.

I During 1986-2014, annual GDP per capita expanded at anaverage annual rate of 5.6%. This made Vietnam the bestperformer out of the ASEAN Six Majors.

21 / 66

Convergence Exercise

(Y /P)Vietnamt

(Y /P)USt=

(Y /L)Vietnamt

(Y /L)USt× (L/WP)Vietnamt

(L/WP)USt× (WP/P)Vietnamt

(WP/P)USt.

I In 1977, China and Vietnam had similarly poor levels of labourproductivity; 4.0% and 4.6% of the U.S. level, respectively.

I The remarkable convergence of China is visible, with labourproductivity reaching 19.0% of the U.S. level by 2014. At8.6% of the U.S. productivity level in 2014, Vietnam’s labourforce is the least productive compared to China and theMajors.

I The working-age population to total population ratio inVietnam consistently rises since the late 1970s in relation tothe U.S. This is evidence of the aforementioned “demographicdividend.”

22 / 66

GDP per capita components relative to the U.S.

1970 1992 20140

5

10

15

20

%25

(a): Y/P

1970 1992 20140

5

10

15

20

%25

(b): Y/L

1970 1992 201480

90

100

110

120

%130

(c): L/WP

1970 1992 201480

90

100

110

120

%130

(d): WP/P

Vietnam

China

ASEAN

23 / 66

An investigation of theVietnamese productivity

24 / 66

Growth accounting

Consider the following aggregate production function:

Yt = AtKαt (hE )

1−αt

In per-worker terms the production function can be rewritten as:

yt = Atkαt ht

1−α

Taking logarithms of the terms and decomposing the averageannual growth rate of output per worker over z years (from time tto time t + z) yields:

lnyt+z − lnytz

=lnAt+z − lnAt

z+ α

lnkt+z − lnktz

+ (1 − α)lnht+z − lnht

z.

25 / 66

Decomposing Vietnam’s growth: 1970-2014

Growth (%) Contributions (%)

Period y A k h y A k h

1970-1985 1.5 1.2 -0.3 0.6 100 81.0 -17.4 36.31986-1996 4.0 1.1 2.6 0.3 100 27.2 66.1 6.71997-2014 4.0 -0.7 3.7 1.0 100 -17.4 92.9 24.51970-2014 3.1 0.4 2.1 0.6 100 12.5 66.8 20.7

Source: PWT 9.0. Note: α = 0.5.

I For the pre-1986 period, there was no capital deepening.

I TFP growth accounted for more than four-fifth of the growthof output per worker between 1980 and 1985.

I Post-1986, the economic fortunes of Vietnam changedconsiderably.

I In the post-1986 period, growth has been mainly coming fromincreases in both physical and human capital rather thanincreases in TFP.

26 / 66

Labour productivity components relative to the U.S.

Expressing country i ’s performance relative to that of the U.S.leads to the following re-write labour productivity:

y ityUSt

=Ait

AUSt

×(

k itkUSt

)α

×(

hithUSt

)1−α

,

I In 1970, TFP in Vietnam was 47.9% of the U.S. level. Itincreased to 53.4% of the U.S. level in 1980 and decreased to48.7% of the U.S. level in 1986.

I In 2014, TFP in Vietnam was 37.3% of the U.S. level.

I TFP in China was 47.1% of the U.S. level in 2014 and thecorresponding figure for the ASEAN group was 48.1% in thatyear.

27 / 66

Labour productivity components relative to the U.S.(%)

1970 1992 201430

44

58

%72

(c): TFP

1970 1992 20140

10

20

%30

(a): K/L

1970 1992 201442

52

62

%72

(b): h

Vietnam

China

ASEAN

28 / 66

Conventional measure of human capital

I Human capital is proxied by years of schooling (and returns toeducation).

I Data are from PWT 9.0. PWT 9.0 follows the procedureimplemented by Hall and Jones (1999) and Caselli (2005).

I Specifically, data on average years of schooling are convertedinto human capital, using the formula h = ϕ(s), where s isaverage years of schooling, and the function ϕ(s) is piecewiselinear and is defined as:

ϕ(s) =

0.134 × s, if s ≤ 4,

0.134 × 4 + 0.101 × (s − 4), if 4 < s ≤ 8,

0.134 × 4 + 0.101 × (s − 4) + 0.068(s − 8), if s > 8.

29 / 66

Average scores of 15-year-old students in the PISA 2012

Mathematics Reading Science Average Relative to the U.S.

Shanghai - China 613 570 580 588 1.19

Indonesia 375 396 382 384 0.78

Malaysia 421 398 420 413 0.84

Singapore 573 542 551 555 1.13

Thailand 427 441 444 437 0.89

Vietnam 511 508 528 516 1.05

U.S. 481 498 497 492 1.00

OECD average 494 496 501 497 1.01

Source: OECD, PISA 2012 (http://www.oecd.org/pisa/).

30 / 66

Adjusted human capital

We extend the usual Mincer model that considers only years ofschooling to one that additionally considers PISA scoreperformance and suggest the following measure of human capitalfor 2012:

h∗ih∗US︸︷︷︸

Considering years of schoolingand PISA scores

=hihUS︸︷︷︸

Considering onlyyears of schooling

× PISAi

PISAUS︸ ︷︷ ︸Pisa scores

31 / 66

Human capital relative to the U.S. in 2012 (%)

0

20

40

60

80

100%

Vietnam China Malaysia Indonesia Singapore Thailand

Conventional measure

PISA-adjusted measure

32 / 66

Can we learn more about therole of productivity by

examining the sectoral data?

33 / 66

Employment share in agriculture vs GDP per worker

7 8 9 10 11 12 130

10

20

30

40

50

60

70

80

%90

Vietnam

Log aggregate GDP per worker (2005)

Em

plo

ym

en

t sh

are

in

ag

ricu

ltu

re (

20

05

)

34 / 66

Deagriculturalization: Historical comparisons

1800 1905 2010

0

10

20

30

40

50

60

70

80

%90

U.S.

35 / 66

Deagriculturalization: Historical comparisons

1800 1905 2010

0

10

20

30

40

50

60

70

80

%90

U.S.

Japan

36 / 66

Deagriculturalization: Historical comparisons

1800 1905 2010

0

10

20

30

40

50

60

70

80

%90

U.S.

Japan

China

37 / 66

Deagriculturalization: Historical comparisons

1800 1905 2010

0

10

20

30

40

50

60

70

80

%90

U.S.

Japan

China

ASEAN

38 / 66

Deagriculturalization: Historical comparisons

1800 1905 2010

0

10

20

30

40

50

60

70

80

%90

U.S.

Japan

China

ASEAN

Vietnam

39 / 66

Deagriculturalization: A comparison with China

0 8 16 24 3235

45

55

65

%75

Years since base date

China (base=1978)

40 / 66

Deagriculturalization: A comparison with China

0 8 16 24 3235

45

55

65

%75

Years since base date

China (base=1978)

Vietnam (base=1990)

41 / 66

Driving forces of deagriculturalization

I Labour Push:

Improvements in agricultural productivity releaseresources from agriculture.

I Labour Pull:

Improvements in non-agricultural technology attractlabour out of agriculture.

We investigate if it is rising productivity in agriculture ornon-agriculture (or both) that is responsible for the pace ofdeagriculturalization and overall productivity in Vietnam.

42 / 66

We examine

I If the pace deagriculturalization of the Vietnamese economy isdue to productivity growth in agriculture or non-agriculture orboth

I Counterfactual: What would have happened to aggregateproductivity if the Vietnamese economy had

I productivity in agriculture same as China

I productivity in both sectors same as China

43 / 66

A Two-Sector Model

44 / 66

I At each date there are two goods produced by two sectorsagriculture (A) and non-agriculture (call it industry, I). Theproduction function for sector j is given by

Yj ,t = θj ,tNj ,t (1)

I The instantaneous utility is defined over the compositeconsumption good (Ct), which is derived from agriculture andnon-agriculture

Ct =(

γ1/ηA (At − A)(η−1)/η + γ

1/ηI I

(η−1)/ηt

)η/(η−1)(2)

I A represents the subsistence level of food consumption andsatisfies

θA,t > A > 0 (3)

45 / 66

I At each date, and given prices, the household choosesconsumption of each good to maximize his lifetime utilitysubject to the budget constraint

pA,tAt + pI ,t It = 1 (4)

I Markets clear in equilibrium

At= YA,t , It= YI ,t (5)

46 / 66

I The equilibrium employment share in agriculture is given by:

NA,t =

(γAθ

η−1A,t

γAθη−1A,t + γI θ

η−1I ,t

)+

(γI θ

η−1I ,t

γAθη−1A,t + γI θ

η−1I ,t

)A

θA,t

I η determines the amount of substitution between agriculturaland industrial goods

47 / 66

Quantitative Analysis

48 / 66

Data

Our sample period is between 1990 and 2013, since there is nopublicly available data on employment and GDP by sector prior to1990 even at a high level of aggregation.

This point is also made by McCaig and Pavcnik (2013).

Data are from the Asian Productivity Database (30 September2015 version) (APO, 2015).

Although this is only a 24-year period, Vietnam’s employmentshare in agriculture fell from 71.4 to 46.8% over this time.

49 / 66

Calibration

We calibrate the model to the Vietnamese data.

All series are de-trended using the Hodrick-Prescott filter with asmoothing parameter of 6.25 before any ratios are computed.

We normalize the level of labour productivity in both sectors toone in 1990, that is, θA,1990 = θI ,1990 = 1.

We need the values of the following four parameters in our model:γA, γI , η, and A.

50 / 66

Calibration

For a given value of η, we jointly determine the value of γA, γI ,and A to match the following set of statistics of the Vietnamesedata:

1. the employment share in agriculture in 1990;

2. the employment share in non-agriculture in 1990;

3. the aggregate labour productivity growth between 1990 and2013.

We set η=0.45 following the structural transformation literature.

Then we determine the values for the remaining three parameters;γA = 0.4982, γI = 0.5018, and A = 0.4104.

51 / 66

Benchmark results: Sectoral employment shares

1990 2002 201446

55

64

%73(a): Agriculture

1990 2002 201427

36

45

%54(b): Non-agriculture

Data Model

52 / 66

Sensitivity: Sectoral employment shares

1990 2002 201446

55

64

%73(a): Agriculture

1990 2002 201427

36

45

%54(b): Non-agriculture

Data

η=0.45

η=0.3

η=0.1

53 / 66

Sectoral productivity differences, China versus Vietnam

1990 2002 20141.0

2.5

4.0(a): Agriculture

Vietnam

China

1990 2002 20141.0

2.5

4.0(b): Non-agriculture

Vietnam

China

54 / 66

Counterfactuals

The success of the benchmark model motivates severalcounterfactual experiments to further investigate the structuraltransformation experience of Vietnam.

We use the model to assess the quantitative role of sectoral labourproductivity growth rates in the behaviour of (i) employment sharein agriculture; and (ii) aggregate labour productivity in Vietnam.

To do so, we perform a series of counterfactual experimentswhereby we replace observed sectoral labour productivity growthrates in a sector with the ones observed in China.

55 / 66

Counterfactuals: Employment share in agriculture

1990 1994 1998 2002 2006 2010 2014

31

45

59

%73

Data Benchmark A only

56 / 66

Counterfactuals: Employment share in agriculture

1990 1994 1998 2002 2006 2010 2014

31

45

59

%73

Data Benchmark A only I only

57 / 66

Counterfactuals: Employment share in agriculture

1990 1994 1998 2002 2006 2010 2014

31

45

59

%73

Data Benchmark A only I only A&I

58 / 66

Aggregate labour productivity in Vietnam (1990=1)

1990 1994 1998 2002 2006 2010 2014

0.8

1.7

2.6

%3.5

Data

59 / 66

Aggregate labour productivity in Vietnam (1990=1)

1990 1994 1998 2002 2006 2010 2014

0.8

1.7

2.6

%3.5

Data A only

60 / 66

Aggregate labour productivity in Vietnam (1990=1)

1990 1994 1998 2002 2006 2010 2014

0.8

1.7

2.6

%3.5

Data A only I only

61 / 66

Aggregate labour productivity in Vietnam (1990=1)

1990 1994 1998 2002 2006 2010 2014

0.8

1.7

2.6

%3.5

Data A only I only A&I

62 / 66

Estimates of distortions to agricultural incentivesI Focuses on government imposed distortions that create a gap

between the domestic price and what they would be underfree markets.

I NRA (Nominal Rate of Assistance): the percentage by whichgovernment policies have raised gross returns to farmersabove what they would be without the governmentsintervention (or lowered them, if NRA < 0).

I RRA (Relative Rate of Assistance):

RRA = 100 ∗[ 100 +NRAag

t

100 +NRAnonagt

− 1]

,

where NRAagt is the NRA to producers of tradables in the

agricultural sector, and NRAnonagt is the NRA to producers of

tradables in the non-agricultural sector.

If the RRA is below (above) zero, a country’s policy regime has ananti-(pro-) agricultural bias.

63 / 66

Evolution of agricultural distortions

1986 1994 2002 2010-40

-30

-20

-10

0

10

20

30

%40(a): Vietnam (1986-2005)

NRA, agricultural tradables

NRA, non-agricultural tradables

RRA

1986 1994 2002 2010-60

-50

-40

-30

-20

-10

0

10

20

30

%40(b): China (1986-2010)

64 / 66

Evolution of agricultural distortions

I The policy regime in Vietnam was characterized by asignificant bias against agriculture between 1986 and 1993.

I The RRA averaged 21% during 1986-1989 and 22% during1990-1993.

I These RRA values suggest that farmers received, on average,less than 80% of the prices they would have received hadmarkets for both farm and non-farm goods been free between1986 and 1993.

I An agricultural bias remained in the post-Asian crisis, but it issmall relative to the corresponding bias that prevailed between1986 and 1994. Vietnam’s agriculture has become much lessdistorted since then. Farmers in Vietnam received slightlypositive assistance, with an average RRA of 1.2% between2002 and 2005.

65 / 66

Conclusions and policy suggestions

I Vietnam must decrease its reliance on factor accumulation asits source of growth and increase its technological capabilities.

I Economic policies should equally target both agricultural andnonagricultural sectors to increase sectoral productivitygrowth rates in Vietnam.

I Our findings suggest that the problem of convergence inVietnam is broad-based and research and policy makingshould place the emphasis on an economy-wide explanationrather idiosyncratic factors.

I Although promising signs, Vietnam’s government must buildon these foundations in order to avoid slipping into theso-called middle-income trap.

66 / 66