Professional Documents
Culture Documents
Outline: 1. Summary of Kataoka (2018) 2. Theoretical Framework in Kataoka (2018)
Outline: 1. Summary of Kataoka (2018) 2. Theoretical Framework in Kataoka (2018)
Bengkulu
Sumatra
Lampung
South
Kalimantan
South
Southeast
Sulawesi Maluku
Papua
• The widening interprovincial pure labor productivity gaps
Sulawesi
Jakarta
Central
Java
Bali
may resist the decrease in T(y) and can be a further issue.
Banten East
East Nusa Tenggara
Java
Yogyakarta
West Nusa
Java provinces Tenggara
Process
•FA derives the production possibility frontier /cost DMU a
frontier. Input Output
production function and error terms, insensitive to Relative Actual output Minimum input
outliers
=
Efficiency Maximum output = Actual input
The frontier derives the maximum output or minimum input values.
2.2.3 Data Envelopment Analysis (DEA) 2.2.4 Efficiency scores, oe = pe ×se
• DEA: A liner programming method to measure the efficiency of
economic units, given actual inputs and outputs.
• DEA derives two frontiers under VRS/CRS assumptions. Overall tech. efficiency (oe, Y/Ye )
VRS: variable returns to scale / CRS: constant returns to scale
VRS where all units operate at maximum level, but not at optimal scale (ABCD1). = Pure tech. efficiency (pe, Y/Ys)
CRS where all units operate at optimal scale (0D2).
• Given 5 actual units A-E, the output-oriented DEA model estimates
× Scale efficiency (se, Ys/Ye)
D1 and D2 under VRS and CRS frontiers, respectively. 0< Score <1. (efficient: score = 1; inefficient: score < 1).
Case in D
Y: Actual observed output values pe: Efficiency in resource utilization (managerial performance)
Ys: Estimated output value without pure
tech. inefficiency se: Efficiency in resource allocation (scale optimality = operation
Ye: Estimated output value without total size)
inefficiency
2.3.1 Measurement of regional income 2.3.2 Income inequality measure: Theil L index
inequality Kataoka (2018) measures regional income inequality by Theil L index,
2 types inequality measures: Absolute / Relative measures. µy (3)
a. Absolute measures conceptualizes differences in money where andµ y is an per capita income in region i and provincial
terms (or the same unit as the variable). mean income. Theil L index is a relative measure that compares
b. Relative measure refers to the average disproportionality region i’s income with the mean income.
of the variable. Greater values in T(y) indicates the greater regional inequality.
Example: Standard deviation, SD; Coefficient of variation, CV Theil L index satisfies the following five properties.
a) Absolute measure: SD(y) = ⁄ ∑ − 1. The anonymity principle
b) Relative measure : CV(y) = ⁄ = ⁄ ⁄ ∑ − 2. Scale (Income) independence principle
∑ !
where = ⁄ and = are an per capita income in 3. Population size independence principle
region i and regional mean income, respectively. A nation consists of 4. Transfer principles
n regions in total. 5. Decomposability
SD(y) is expressed as the monetary term while CV(y) is
expressed as the relative ratio.
2.3.3 Scale independence 2.3.4 Desirable properties of inequality measures
Scale independence: The measurement values do not depend I(y) denotes an inequality measure of variable y.
on the scale of economy. We explain by using the follwing 1. The anonymity principle
– It does not matter who earns the income.
a) SD(y) = ⁄ ∑ − yi={0.5, 4.0, 1.0, 1.5, 3.0}, yi’={4.0, 0.5, 1.0,3.0, 1.5}, i = {a, b, c, d, e}
b) CV(y) = ⁄ = ⁄ ⁄ ∑ − yi’ has the same income distribution pattern as yi, but the recipients are different.
When every region’s income is multiplied by any positive constant λ, 2. Population size independence principle
The measurement value in (a) changes and in (b) remains unchanged. – It should not depend on the number of income recipients.
– The measurement value remains unchanged when the number of income
If all regions' income increases by λ (λ>1), recipients changes as long as it has the same income distribution pattern.
yi={0.5, 4.0, 1.0, 1.5, 3.0}, yj={4.0, 0.5, 1.0, 3.0, 1.5}, The economy in i and j has
SD(λy)= λ −λ = λ − =λSD. Thus, SD(y)≠SD(λy). the same income pattern I(yi)=I(yj). Thus, I(yi) = I(yi+yj) = I(yj) = I(2yi)
3. Transfer principles
– The measurement value decreased, when some incomes are transferred from the
CV(λy)= ⁄ ∑ λ −λ = λ − rich (e) to the poor (a).
λ λ
yi= {0.5, 4.0, 1.0, 1.5, 3.0}, sd(yi)= 1.304; yi’= {0.5+1, 4.0, 1.0, 1.5, 3.0−1},
λ "# sd(yi’) = 1.048, I(yi)>I(yi’)
= = . Thus, SD(y)=SD(λy).
λ 4. Decomposability
CV(y) satisfies the property of scale independence. – Inequality value can be disaggregated into several interpretative factors.
I(y)= I(x) + I(z): Inequalities in x and z influence inequality in y.
2010
Maximum Minimum
Variables
Value Province Value Province
GDP, Y (IDR trillion) 410.8 W. Java 7.3 Maluku
Physical Capital , K ( IDR trillion) 1,311.9 Jakarta 10.7 Bengkulu
Labor, L (Million) 24.2 W. Java 0.9 Bengkulu
Average years of education, H 10.4 Jakarta 5.3 Papua
Population, P (Million) 53.7 W. Java 1.7 Bengkulu
GDP per capita, y 41.2 Jakarta 2.7 E. Nusa Tenggara
Labor productivity, x 75.0 Jakarta 5.9 E. Nusa Tenggara
Physical capital per capita, k 136.5 Jakarta 4.7 Maluku
Labor participation rate, l 0.577 Bali 0.423 Maluku
3.3.1 Findings: Kernel density distribution in 3.3.2 Findings: Change in average efficiency
efficiency score scores
Mean 1990 2010 # of provinces with mean
in 1990 < mean in 2020
oe 0.509 0.771 22
pe 0.647 0.843 22
se 0.794 0.918 18
3.3.3 Findings: Overall tech. efficiency (oe) 3.3.4.a Classifying provinces by RTS
Provinces with the lowest & highest OE Output-oriented VRS model
Annual Growth 1990/2010 D1: Projected inputs/outputs derived by VRS Frontier
scores D2: Projected inputs/outputs derived by CRS Frontier
Rank Province (%) 1990 2010
Lowest
1st Highest Bengkulu 8.4 Province Score Province Score Classifying a province by RTS (Returns to Scale)
2nd Highest Jambi 6.9 1st Prov. Bengkulu 0.184 C. Kalimantan 0.407 VRS frontier though ABCEF consists of Segments AB (IRS), BC(CRS) and CEF(DRS).
3rd Highest C. Sulawesi 6.9 2nd Prov. W. Nusa Tenggara 0.187 W. Nusa Tenggara 0.506
3rd Prov. C. Sulawesi 0.187 Yogyakarta 0.510
D locates below frontier:D can be improved to D1. Thus, D is classified as DRS.
24th Highest Riau -0.8 Provinces 6 provinces: Aceh, Riau, 5 provinces: Bengkulu, W. VRS Frontier
CRS Frontier se=oe/pe =(D0 D1 / D0D2) if D moves to
25th Highest Aceh -1.2 with score = W. Java, Jakarta, E. Java, Java, Jakarta, E.
1.0 E. Kalimantan Kalimantan, S. Kalimantan leftward, se increases.
26th Highest Papua -1.8 Output
1. Higher oe growth provinces (Bengkulu 8.4%, Jambi 6.9%, C. Sulawesi 6.9%) ・D2 E F IRS : Increasing returns to scale. Increase in
C ・ operation size improves se.
• are resource-poor off-Java provinces. D1 A0ABB0Blue
• Show the lower oe scores in 1990 (Bengkulu 0.184, Jambi 0.191, C. Sulawesi0.187).
⇒ Efficiency convergence: Negatively correlated between the 1990’s oe scores and B CRS: Constant returns to scale. No change in
the growth rates: -0.879. operation size improves se.
2. Negative oe growth provinces (Aceh -1.2%, Riau -0.8%, and Papua -1.8%) ・D B0BCC0 Orange
• Those are resource-rich provinces DRS: Decreasing returns to scale. Decrease
• Those show the higher oe scores in 1990 (Aceh 1.0, Riau 1.0, Papua 0.818). IRS CRS DRS in operation size improves se.
⇒ Resource curse thesis in efficiency: Rich natural resource provinces show poor A Input C0CEFF0 Purple
efficiency growth. 0 A0 B0 C0 D0 E0 F0
3.3.4.b Classifying provinces by RTS in 2010 3.3.5 Reference sets
# of provinces by RTS(CRS:6, IRS:15, DRS:5) Efficient peer groups, referred by inefficient provinces
• Inefficient F refers to efficient sets of B and C. RS shows the # of the reference stets (Efficient
– Provinces exhibiting IRS: 15; Provinces exhibiting DRS: 5 • Inefficient D refers to efficient sets of C and E.
– The fact that majority of provinces improves se by increasing operation size infers province + inefficient followers).
B, C, and E are the role model while A, F, and D are
that those face the resource constraints. followers. 1990 2010
2. Provinces with lower se exhibit IRS. ⇒Necessary for improving the access to the Provinces RS Provinces RS
resources in economic activities. Aceh 15 S. Kalimantan 15
2010 2010 Papua 10 E. Kalimantan 12
Provinces Provinces E. Kalimantan 7 W. Java 7
RTS SE RTS SE
1 Aceh DRS 0.992 14 Bali DRS 0.988 W. Java 6 Papua 7
2 N. Sumatra IRS 0.995 15 W. Kalimantan IRS 0.802 E. Java 6 Bengkulu 6
3 Riau* IRS 0.987 16 C. Kalimantan IRS 0.468 SE. Sulawesi 6 E. Nusa Tenggara 4
4 W. Sumatra DRS 0.994 17 E. Kalimantan* CRS 1.000 Riau 1 Jakarta 1
5 Jambi CRS 1.000 18 S. Kalimantan CRS 1.000 E. Java 1
6 Bengkulu CRS 1.000 19 N. Sulawesi* IRS 0.991
7 S. Sumatra* IRS 0.942 20 C. Sulawesi IRS 0.899 • 1990: Resource rich provinces, Aceh, Papua, E.Kalimantan have many followers.
8 Lampung IRS 0.960 21 S. Sulawesi* IRS 0.967
• 2010: Off-Java lower income provinces, Bengkulu and E. Nusa Tenggara have
9 W. Java* CRS 1.000 22 SE. Sulawesi IRS 0.969
several followers.
10 Jakarta CRS 1.000 23 W. Nusa Tenggara IRS 0.763
11 C. Java IRS 0.969 24 E. Nusa Tenggara IRS 0.652
• Jakarta and E. Java has no followers as those are regarded as extraordinary provinces
12 Yogyakarta DRS 0.995 25 Maluku* DRS 0.998
in terms of input-output efficiency.
13 E. Java IRS 0.968 26 Papua* IRS 0.557
Appendix 2.1 Cheng & Li’s (2006) inequality Appendix 2.2 Cheng & Li’s (2006) inequality
decomposition decomposition
The covariance of li and xi (cov(l, x)) can be expressed as follows:
Recall Slide 2.3.2 that Theil L is expressed as:
E = ⁄ ∑ ln µ y⁄ E ≥0 (1) cov +, - = 1⁄D ∑ + − J - − L = − J⋅ L (4)
Per capita income is expressed as Dividing both sides in Eq. (4) by (µ i・ µ x), we get
=+ ·- (2) =
QRS J,L
+1 (5)
where + and - are labor participation rate and labor productivity. Let I⋅ K I⋅ K
µ y, µ l, and µ x, are provincial mean values of yi (µ y [= (1/n)Σyi]) and its Substituting Eq. (5) into Eq. (3), we obtain.
corresponding two multiplicative terms. QRS J,L
E = E + + E - + ln + 1 = E + + E - + / +, - (6)
Then, substituting Eq. (1) into Eq. (2) and multiplying the quotient I⋅ K
inside the natural logarithm by (μl · μx / μl · μx) yields QRS J,L
where I(l, x)=ln + 1 is the interaction term that can be
E = ⁄ ∑ ln I ⋅ K ⋅
M I⋅ K
J L I⋅ K positive, negative, or zero if + and - are correlated positively,
I K
= ⁄ ∑ ln + ⁄ ∑ ln + ln negatively, or not correlated.
J L I⋅ K