Education spending, economic development, and the size of government

Millin, Mark; Fielding, David; Owen, P. Dorian

doi:10.3326/pse.47.3.1.

1165
Views

539
Downloads

Education spending, economic development, and the size of government

Mark Millin*

David Fielding*

P. Dorian Owen*

Download citation

https://doi.org/10.3326/pse.47.3.1

FULL ARTICLE

FIGURES & DATA

REFERENCES

CROSSMARK POLICY

METRICS

LICENCING

PDF

Table 1
Data definitions and sources

Variable name	Description of the variable	Source
Dependent variables
pse/gdp	Public spending on education, total (% of GDP)	World Bank EdStats
pse/gov	Public spending on education, total (% of total government spending)	World Bank EdStats
Explanatory variables
ypc2015	GNI per capita country grouping in 2015, 21 OECD countries	World Bank (Atlas Method)
region	Richer (versus poorer) country regions	Authors’ compilation
poldemoc	Political democracy classification: yes; no	Freedom House
Control variables
pop024	Population aged 0-24 (% of total population)	World Bank EdStats
urban	Urban population (% of total population)	World Bank WDI
trade	Exports plus imports of goods and services (% of GDP)	World Bank WDI
hci	Human capital index	Penn World Table 9.0
pop65	Population aged 65 and above (% of total population)	World Bank WDI
military	Military expenditure (% of GDP)	World Bank WDI
fiscbal	Fiscal balance (% of GDP)	World Bank DPG
debt	General government gross debt (IMF, % of GDP)	World Bank TCdata360
Other variables
gdppc	GDP per capita, PPP (constant 2011 international $)	World Bank WDI
gini	Gini index (World Bank Estimate)	World Bank WDI

Notes: EdStats refers to the World Bank’s Education Statistics database (World Bank, 2017a).

TCdata360 refers to the World Bank’s TCdata360 database (World Bank, 2017b). WDI refers to the World Bank’s World Development Indicators database (World Bank, 2017c). DPG refers to the World Bank’s Development Prospects Group: A Cross-Country Database of Fiscal Space (World Bank, 2017d). The pop024 variable is the sum of pop014 and pop1524 variables from the World Bank EdStats database. Freedom House refers to the Freedom in the World survey data (Freedom House, 2016). See Feenstra, Inklaar and Timmer (2015) for the Penn World Table 9.0.

Table 2
Mean differences in the national effort and budget share by income group and regime type

Dependent variable	pse/gdp						pse/gov
Estimation method	LSDV		Quantile		Robust		LSDV		Quantile		Robust
	(A)	(B)	(A)	(B)	(A)	(B)	(A)	(B)	(A)	(B)	(A)	(B)
1#0. Low income, not democratic	-1.551***	-1.635***	-1.787***	-1.885***	-1.979***	-2.013***	3.633***	-0.012	3.948***	-0.162	3.387***	-0.210
1#0. Low income, not democratic	(0.254)	(0.359)	(0.133)	(0.260)	(0.130)	(0.207)	(0.436)	(0.726)	(0.608)	(0.696)	(0.392)	(0.614)
1#1. Low income, democratic	-1.684***	-1.756***	-1.757***	-1.877***	-1.724***	-1.759***	4.796***	1.339*	5.135***	1.290**	4.821***	1.467**
1#1. Low income, democratic	(0.123)	(0.257)	(0.127)	(0.257)	(0.153)	(0.216)	(0.381)	(0.686)	(0.366)	(0.581)	(0.467)	(0.643)
2#0. Lower middle income, not democratic	-1.152***	-1.412***	-1.170***	-1.578***	-1.362***	-1.611***	4.304***	1.081*	4.231***	0.614	4.196***	0.850*
2#0. Lower middle income, not democratic	(0.130)	(0.220)	(0.209)	(0.259)	(0.116)	(0.171)	(0.368)	(0.600)	(0.595)	(0.531)	(0.353)	(0.511)
2#1. Lower middle income, democratic	-0.454***	-0.663***	-0.739**	-1.126***	-0.979***	-1.123***	4.852***	2.421***	4.137***	1.661***	4.224***	1.830***
2#1. Lower middle income, democratic	(0.158)	(0.199)	(0.318)	(0.287)	(0.117)	(0.159)	(0.369)	(0.446)	(0.470)	(0.419)	(0.348)	(0.468)
3#0. Upper middle income, not democratic	-1.090***	-1.389***	-1.357***	-1.669***	-1.432***	-1.764***	2.520***	-0.022	2.058***	-0.131	2.317***	-0.128
3#0. Upper middle income, not democratic	(0.155)	(0.230)	(0.166)	(0.177)	(0.129)	(0.146)	(0.455)	(0.503)	(0.748)	(0.371)	(0.405)	(0.450)
3#1. Upper middle income, democratic	-0.819***	-1.097***	-0.797***	-1.131***	-0.935***	-1.133***	3.643***	1.708***	3.556***	1.205***	3.411***	1.347***
3#1. Upper middle income, democratic	(0.097)	(0.132)	(0.106)	(0.156)	(0.106)	(0.127)	(0.279)	(0.354)	(0.331)	(0.179)	(0.318)	(0.374)
4#0. High income (non-OECD), not democratic	-1.123***	-1.987***	-1.478***	-1.933***	-1.329***	-2.047***	0.987**	-2.057***	0.034	-3.107***	0.683	-2.852***
4#0. High income (non-OECD), not democratic	(0.183)	(0.217)	(0.152)	(0.195)	(0.171)	(0.189)	(0.499)	(0.566)	(0.751)	(0.356)	(0.502)	(0.543)
4#1. High income (non-OECD), democratic	-0.814***	-0.768***	-0.644***	-0.654***	-0.777***	-0.721***	0.568**	-0.038	0.923***	0.308	0.471	-0.099
4#1. High income (non-OECD), democratic	(0.091)	(0.101)	(0.123)	(0.130)	(0.106)	(0.111)	(0.249)	(0.254)	(0.308)	(0.342)	(0.333)	(0.338)
5#0. High income (OECD), not democratic	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a
5#1. High income (OECD), democratic BASE	5.352***	3.668***	5.226***	3.107***	5.325***	3.449***	12.245***	4.332***	11.973***	0.885	12.217***	2.576***
5#1. High income (OECD), democratic BASE	(0.055)	(0.312)	(0.056)	(0.334)	(0.073)	(0.260)	(0.122)	(0.831)	(0.166)	(0.643)	(0.214)	(0.768)

pop024		0.016***		0.022***		0.018***		0.146***		0.191***		0.164***
pop024		(0.006)		(0.006)		(0.004)		(0.015)		(0.010)		(0.013)
urban		0.009***		0.011***		0.011***		0.029***		0.046***		0.040***
urban		(0.002)		(0.003)		(0.002)		(0.007)		(0.006)		(0.006)
trade		0.008***		0.008***		0.007***		0.017***		0.022***		0.021***
trade		(0.001)		(0.001)		(0.001)		(0.003)		(0.003)		(0.002)

R-squared	0.060	0.100	0.074	0.107	0.123	0.190	0.141	0.198	0.092	0.146	0.136	0.226
F-value	35.86***	34.89***	n/a	n/a	43.22***	48.51***	64.55***	56.40***	n/a	n/a	42.96***	54.58***
Countries	183	169	183	169	183	169	175	165	175	165	175	165
Years	27	26	27	26	27	26	27	26	27	26	27	26
Observations	2468	2288	2468	2288	2468	2288	2194	2069	2194	2069	2194	2069

Notes: BASE group is high-income (OECD) and democratic countries. A pseudo R-squared is reported for the quantile regression. Not applicable (n/a) means there are no applicable countries for these groups. Standard errors are given in parentheses. Huber/White heteroskedasticity-robust standard errors are reported for the LSDV and quantile estimators.

Significance levels: * p < 0.10, ** p < 0.05, *** p < 0.01.

Table 3
Mean differences in the national effort and budget share by country region and regime type

Dependent variable	pse/gdp						pse/gov
Estimation method	LSDV		Quantile		Robust		LSDV		Quantile		Robust
	(A)	(B)	(A)	(B)	(A)	(B)	(A)	(B)	(A)	(B)	(A)	(B)
0#0. Poorer country regions, not democratic	-1.298***	-2.031***	-1.633***	-1.756***	-1.681***	-1.760***	3.880***	1.492**	4.046***	0.726	3.757***	1.108*
0#0. Poorer country regions, not democratic	(0.147)	(0.216)	(0.117)	(0.234)	(0.104)	(0.199)	(0.305)	(0.612)	(0.499)	(0.764)	(0.303)	(0.597)
0#1. Poorer country regions, democratic	-0.804***	-1.412***	-1.163***	-1.235***	-1.110***	-1.178***	5.077***	3.321***	4.833***	2.284***	4.923***	2.945***
0#1. Poorer country regions, democratic	(0.104)	(0.166)	(0.121)	(0.180)	(0.098)	(0.166)	(0.233)	(0.458)	(0.287)	(0.496)	(0.283)	(0.491)
1#0. Richer country regions, not democratic	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a
1#1. Richer country regions, democratic BASE	5.169***	2.095***	5.114***	2.047***	5.213***	2.520***	11.944***	5.333***	11.849***	2.093**	11.943***	3.705***
1#1. Richer country regions, democratic BASE	(0.069)	(0.375)	(0.062)	(0.404)	(0.074)	(0.348)	(0.127)	(1.193)	(0.160)	(0.960)	(0.212)	(1.036)

pop024		0.034***		0.025***		0.019***		0.104***		0.163***		0.127***
pop024		(0.007)		(0.007)		(0.006)		(0.021)		(0.019)		(0.019)
urban		0.016***		0.022***		0.020***		0.029***		0.044***		0.037***
urban		(0.003)		(0.003)		(0.002)		(0.008)		(0.008)		(0.007)
trade		0.013***		0.011***		0.011***		0.018***		0.017***		0.021***
trade		(0.002)		(0.002)		(0.001)		(0.004)		(0.004)		(0.003)

R-squared	0.057	0.167	0.090	0.170	0.154	0.313	0.195	0.222	0.140	0.171	0.189	0.247
F-value	52.61***	63.60***	n/a	n/a	135.10***	125.25***	270.00***	113.21***	n/a	n/a	158.19***	84.82***
Countries	102	97	102	97	102	97	99	96	99	96	99	96
Years	27	26	27	26	27	26	27	26	27	26	27	26
Observations	1486	1382	1486	1382	1486	1382	1360	1299	1360		1360	1299

Notes: BASE group is richer country regions that are democratic. See the notes for table 2. Standard errors are given in parentheses.

Significance levels: * p < 0.10, ** p < 0.05, *** p < 0.01.

Table 4
Wald tests for parameter equality of the factor-variable interactions

Wald tests for parameter equality in table 2
	Robust estimator
	(A)				(B)
Dependent variable: pse/gdp	Wald Statistic	Uncorrected Significance (α = 0.05)	Corrected Significance (α = 0.0125)	Wald Statistic	Uncorrected Significance (α = 0.05)	Corrected Significance (α = 0.0125)
Test parameter 1#0 = 1#1	F(1, 2459) = 2.19	No	No	F (1, 2276) = 2.33	No	No
	p = 0.1388			p = 0.1273
Test parameter 2#0 = 2#1	F(1, 2459) = 8.90	Yes	Yes	F(1, 2276) = 14.79	Yes	Yes
	p = 0.0029			p = 0.0001
Test parameter 3#0 = 3#1	F(1, 2459) = 14.38	Yes	Yes	F(1, 2276) = 24.17	Yes	Yes
	p = 0.0002			p = 0.0000
Test parameter 4#0 = 4#1	F(1, 2459) = 10.20	Yes	Yes	F(1, 2276) = 51.44	Yes	Yes
	p = 0.0014			p = 0.0000
	Robust estimator
	(A)			(B)
Dependent variable: pse/gov	Wald Statistic	Uncorrected Significance (α = 0.05)	Corrected Significance (α = 0.0125)	Wald Statistic	Uncorrected Significance (α = 0.05)	Corrected Significance (α = 0.0125)
Test parameter 1#0 = 1#1	F(1, 2185) = 7.33	Yes	Yes	F(1, 2057) = 11.37	Yes	Yes
	p = 0.0068			p = 0.0008
Test parameter 2#0 = 2#1	F(1, 2185) = 0.00	No	No	F(1, 2057) = 6.75	Yes	Yes
	p = 0.9438			p = 0.0094
Test parameter 3#0 = 3#1	F(1, 2185) = 6.88	Yes	Yes	F(1, 2057) = 13.88	Yes	Yes
	p = 0.0088			p = 0.0002
Test parameter 4#0 = 4#1	F(1, 2185) = 0.16	No	No	F(1, 2057) = 25.58	Yes	Yes
	p = 0.6852			p = 0.0000
Wald tests for parameter equality in table 3
	Robust estimator
	(A)			(B)
Dependent variable: pse/gdp	Wald Statistic	Uncorrected Significance (α = 0.05)	Corrected Significance (α = 0.05)	Wald Statistic	Uncorrected Significance (α = 0.05)	Corrected Significance (α = 0.05)
Test parameter 0#0 = 0#1	F(1, 1483) = 33.66	Yes	Yes	F(1, 1376) = 35.80	Yes	Yes
	p = 0.0000			p = 0.0000
	Robust estimator
	(A)			(B)
Dependent variable: pse/gov	Wald Statistic	Uncorrected Significance (α = 0.05)	Corrected Significance (α = 0.05)	Wald Statistic	Uncorrected Significance (α = 0.05)	Corrected Significance (α = 0.05)
Test parameter 0#0 = 0#1	F(1, 1357) = 16.59	Yes	Yes	F(1, 1293) = 37.70	Yes	Yes
	p = 0.0000			p = 0.0000

Notes: The “#” naming convention accords with that in the respective table of results. Using interaction models with applicable controls, “Yes” means the relevant interaction parameters are statistically significantly different (“No” means not significantly different) from one another for the respective pairwise comparison at the conventional (uncorrected) 5% level of significance or Bonferroni (corrected) level of significance. (A) refers to the model with no controls (unconditional mean differences) and (B) refers to the model with controls (conditional mean differences). Because there is only one pairwise test of parameter equality performed on the estimates from table 3, α (= 0.05) is the same for both the uncorrected and corrected critical level of significance.

Table A1
List of countries and territories by GNI per capita group in 2015 (ypc2015)

Low income
(31)

Lower middle income
(52)

Upper middle income
(56)

High income
(mostly non-OECD) (57)

High income
(OECD) (21)

Afghanistan

Benin

Burkina Faso

Burundi

Central African Republic

Chad

Comoros

Congo, Dem. Rep.

Eritrea

Ethiopia

Gambia, The

Guinea

Guinea-Bissau

Haiti

Korea, Dem. People’s Rep.

Liberia

Madagascar

Malawi

Mali

Mozambique

Nepal

Niger

Rwanda

Senegal

Sierra Leone

Somalia

South Sudan

Tanzania

Togo

Uganda

Zimbabwe

Armenia

Bangladesh

Bhutan

Bolivia

Cabo Verde

Cambodia

Cameroon

Congo, Rep.

Cote d'Ivoire

Djibouti

Egypt, Arab Rep.

El Salvador

Ghana

Guatemala

Honduras

India

Indonesia

Kenya

Kiribati

Kosovo

Kyrgyz Republic

Lao PDR

Lesotho

Mauritania

Micronesia, Fed. Sts.

Moldova

Mongolia

Morocco

Myanmar

Nicaragua

Nigeria

Pakistan

Papua New Guinea

Philippines

Samoa

Sao Tome and Principe

Solomon Islands

Sri Lanka

Sudan

Swaziland

Syrian Arab Republic

Tajikistan

Timor-Leste

Tonga

Tunisia

Ukraine

Uzbekistan

Vanuatu

Vietnam

West Bank and Gaza

Yemen, Rep.

Zambia

Albania

Algeria

American Samoa

Angola

Argentina

Azerbaijan

Belarus

Belize

Bosnia and Herzegovina

Botswana

Brazil

Bulgaria

China

Colombia

Costa Rica

Cuba

Dominica

Dominican Republic

Ecuador

Equatorial Guinea

Fiji

Gabon

Georgia

Grenada

Guyana

Iran, Islamic Rep.

Iraq

Jamaica

Jordan

Kazakhstan

Lebanon

Libya

Macedonia, FYR

Malaysia

Maldives

Marshall Islands

Mauritius

Mexico

Montenegro

Namibia

Palau

Panama

Paraguay

Peru

Romania

Russian Federation

Serbia

South Africa

St. Lucia

St. Vincent and the Grenadines

Suriname

Thailand

Turkey

Turkmenistan

Tuvalu

Venezuela, RB

Andorra

Antigua and Barbuda

Aruba

Bahamas, The

Bahrain

Barbados

Bermuda

British Virgin Islands

Brunei Darussalam

Cayman Islands

Channel Islands

Chile

Croatia

Curacao

Cyprus

Czech Republic

Estonia

Faroe Islands

French Polynesia

Gibraltar

Greenland

Guam

Hong Kong SAR, China

Hungary

Iceland

Isle of Man

Israel

Korea, Rep.

Kuwait

Latvia

Liechtenstein

Lithuania

Luxembourg

Macao SAR, China

Malta

Monaco

Nauru

New Caledonia

Northern Mariana Islands

Oman

Poland

Puerto Rico

Qatar

San Marino

Saudi Arabia

Seychelles

Singapore

Sint Maarten (Dutch part)

Slovak Republic

Slovenia

St. Kitts and Nevis

St. Martin (French part)

Trinidad and Tobago

Turks and Caicos Islands

United Arab Emirates

Uruguay

Virgin Islands (U.S.)

Australia

Austria

Belgium

Canada

Denmark

Finland

France

Germany

Greece

Ireland

Italy

Japan

Netherlands

New Zealand

Norway

Portugal

Spain

Sweden

Switzerland

United Kingdom

United States

Notes: Groups are adapted from the World Bank’s Country and Lending Groups for the 2015 calendar

year, based on GNI per capita calculated using the World Bank Atlas Method, except for the high-income (OECD) group, which includes the 21 countries comprising the “core” OECD countries that have been categorised as OECD for the entire study period, from 1989 to 2015 (i.e., excluding Chile, Czech Republic, Estonia, Hungary, Iceland, Israel, Korea, Latvia, Luxembourg, Mexico, Poland, Slovak Republic, Slovenia and Turkey, which are mostly included in the second high-income group). Numbers in parentheses show the total number of countries in each group.

The historical classification is available from: https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups.

Source: Adapted from the World Bank’s historical classification.

Table A2
List of countries by two regional country groups (region)

Poorer country regions
Central Africa (8)	Central America (8)	East Africa (12)	South America (12)	South Asia (8)
Cameroon Central African Republic Chad Congo, Dem. Rep. Congo, Rep. Equatorial Guinea Gabon Sao Tome and Principe	Belize Costa Rica El Salvador Guatemala Honduras Mexico Nicaragua Panama	Burundi Comoros Djibouti Eritrea Ethiopia Kenya Rwanda Somalia South Sudan Sudan Tanzania Uganda	Argentina Bolivia Brazil Chile Colombia Ecuador Guyana Paraguay Peru Suriname Uruguay Venezuela, RB	Afghanistan Bangladesh Bhutan India Maldives Nepal Pakistan Sri Lanka
Southeast Asia (11)	Southern Africa (13)	West Africa (16)
Brunei Darussalam Cambodia Indonesia Lao PDR Malaysia Myanmar Philippines Singapore Thailand Timor-Leste Vietnam	Angola Botswana Lesotho Madagascar Malawi Mauritius Mozambique Namibia Seychelles South Africa Swaziland Zambia Zimbabwe	Benin Burkina Faso Cabo Verde Cote d’Ivoire Gambia, The Ghana Guinea Guinea-Bissau Liberia Mali Mauritania Niger Nigeria Senegal Sierra Leone Togo
Richer country regions
North America (3)	Nordic Countries (5)	Western Europe (22)
Bermuda Canada United States	Denmark Finland Iceland Norway Sweden	Andorra Austria Belgium Channel Islands Faroe Islands France Germany Gibraltar Greece Greenland Ireland Isle of Man Italy Liechtenstein Luxembourg Monaco Netherlands Portugal San Marino Spain Switzerland United Kingdom

Notes: The numbers in parentheses show the total number of countries in each sub-group of the respective country regions. For the poorer country regions, Equatorial Guinea, Chile, Uruguay, Brunei Darussalam, Singapore and Seychelles (the countries in bold) are excluded because they are classified as high-income (non-OECD) countries for most or all of the time period under investigation (from 1989 to 2015).

Source: Authors’ compilation.

Table A3
Descriptive statistics

Variable name	Data coverage	N	Countries	Years	Mean	Std dev.	Minimum	Maximum
pse/gdp	1989-2015	2551	193	13.2	4.505	2.007	0.781	44.334
pse/gov	1989-2015	2255	181	12.5	14.849	5.036	2.563	47.279
ypc2015	1989-2015	5859	217	27.0	n/a	n/a	n/a	n/a
region	1989-2015	3024	112	27.0	n/a	n/a	n/a	n/a
poldemoc	1989-2015	5105	193	26.5	n/a	n/a	n/a	n/a
pop024	1990-2015	4714	184	25.6	49.977	13.687	20.160	73.288
urban	1989-2015	5799	215	27.0	55.788	24.901	5.342	100.000
trade	1989-2015	4785	193	24.8	86.996	52.290	0.021	531.737
hci	1989-2014	3703	143	25.9	2.342	0.694	1.028	3.734
pop65	1989-2015	5234	195	26.8	7.073	4.814	0.697	26.342
military	1989-2015	3870	166	23.3	2.433	3.210	0	117.388
fiscbal	1990-2015	4184	191	21.9	-2.299	13.715	-505.442	122.188
debt	1989-2015	3796	186	20.4	57.015	49.714	0	789.833
gini	1989-2014	1188	155	7.7	39.875	9.871	16.23	65.76
gdppc	1990-2015	4803	195	24.6	15111	18507	247	137164

Notes: Years refers to the average number of years (time-series observations) for each country. Std. dev. refers to the overall standard deviation. Two changes were made to the original data for the pse/gdp variable. The zero observation for Turkey in 1998 was deleted (because there were no other 0% values in the dataset; nil or negligible appeared in the original UNESCO source data for this observation) and the observation for Tuvalu in 1997 (3730833.5%) was deleted as an obvious mistake; the extreme value for this observation also appeared in the original UNESCO source data. It was subsequently noted that this observation was deleted from the World Bank’s Education Statistics (EdStats) data as of the update dated 21 May 2018. Descriptive results are not reported for ypc2015 (21 OECD countries), region and poldemoc because these are sets of binary variables used to characterise broad political and economic categories.