Manuscript ID: EER-D Public spending and growth: the role of government accountability. Online Appendix

Transcription

1 Manuscript ID: EER-D Public spending and growth: the role government accountability Online Appendix Contents: 1. Construction the fiscal dataset Data sources Stylized facts on the relationship between public spending and growth Estimation results with democracy proxies Exploiting time variations in institutional proxies Estimation results with freedom the press" as an institution proxy Growth effects different components current spending Estimations using proxies passive waste Estimations which exclude the most fiscally decentralized countries Estimations excluding ex-soviet and ex-socialist countries Estimations which exclude outliers References

2 1. Construction the fiscal dataset To explain how we construct unified disaggregated spending series following economic classifications, we start with clarifying the main differences in the exact definitions current and capital concepts under GFSM2001 and GFSM1986 (see Wickens (2002) for details). First, the capital expenditure concept under GFSM2001, denoted as net acquisition non-financial assets adopts a net concept, deducting government revenue from the sales fixed capital assets, while capital expenditure under GFSM1986, following a gross concept, does not deduct the revenue from capital sales, which is recorded as part total revenue. Second, while capital transfers were a part capital expenditure under GFSM1986, they are part the current expenditure concept, denoted as expense, under GFSM2001. Facing these differences, we first retrieved all historical spending data available for all countries that have reported data to the IMF's GFS yearbook from 1970 to 2010 and then converted spending items under GFSM1986 into the concepts defined by GFSM2001, so that the capital spending in our spending series deducts sales revenues and excludes capital transfers, with the latter included in the current spending. However, there is another key remaining issue to be dealt with, related to the fact that under GFSM 1986, statistics are reported on a cash basis (i.e., flows are recorded at the time cash is received or paid), while under GFSM2001, they are on an accrual basis (i.e., flows are recorded when economic value is created or extinguished). Specifically, the accrual concept consumption fixed capital, i.e., a decline in the value governments' fixed assets due to physical deterioration, obsolescence, or accidental damages, exists only under GFSM2001. This implies that even after the adjustments mentioned above, the capital spending concept under GFSM1986 and GFSM2001 are still not consistent, with the former not deducting this depreciation capital. To tackle this, for the data originally retrieved from GFSM2001, we move (i.e., add) the consumption fixed capital, initially categorized as current spending, to the capital spending component, so that the modified capital spending component becomes comparable to the ones from GFSM1986, i.e., without the depreciation deducted. More generally, however, the innate difference between cash- and accrual-based statistics still remains. Therefore, although the use time dummies in our analyses should help mitigate the effects possible systematic differences between them (remember that GFSM2001 was introduced between mid 1990s and early 2000s), it is important to acknowledge that our unification measure is not exact, but approximate. We also report that the level government covered in the unified dataset is at the central government (CG) level. This is because, under GFSM1986, countries report data at most at the CG 2

3 level, although under GFSM2001 they also provide data for the general government (GG) level. We primarily use consolidated, rather than budgetary, CG level data, yet when no budget deficits data are available at the consolidated level for a country over our sample period ( ), we use budgetary data for that country, to maximize the number countries and observations available. Note that this way using budgetary CG data ensures that when considering fiscal series for a given country, consolidated and budgetary CG data are never mixed over time, thus no potential jump in the series is created due to the usage data from different CG levels. Nonetheless, our robustness checks consider the case where no budgetary CG level data are included. Lastly, to construct consistent total revenue series spanning two methodologies, for the total revenue data retrieved from GFSM1986, we exclude the revenue from sales capital assets, to make it in line with the total revenue concept under GFSM2001. Having made the current and capital spending and total revenue comparable between the methodologies, we subsequently obtain the budget deficit as a difference between total expenditure, a sum current and capital spending, and total revenue. 3

4 2. Data sources The GDP growth rate is obtained as the log difference over 8 years (for our reference regressions) real GDP per capita taken from the Penn World Tables (PWT 8.0, Feenstra et al. (2015)). Initial real GDP per capita is from the same source. All the fiscal variables are originally from the IMF's GFS yearbook. To calculate fiscal data as a ratio to GDP, GDP figures are taken from the World Economic Outlook (WEO), while exchange rate data, required for unit conversion, are from both WEO and the International Financial Statistics (IFS) databases the IMF. Turning to the other explanatory variables, years schooling (for the population aged between 25 and 64) is from Barro and Lee (2010), which is interpolated to proxy initial human capital in each period. The private investment ratio is calculated as a difference between the total investment ratio (the ratio gross fixed capital formation to GDP, from WEO) and the share capital spending in GDP that we assembled. The population growth rate is from WEO. Percentages the population below 15 and above 65 years old, used in the robustness checks, are from the World Bank's World Development Indicators (WDI). The inflation rate is calculated as the relevant percentage change in CPI, from WEO. The degree openness is obtained as the ratio the sum values imports and exports to GDP, all which are from WEO. Private credit, defined as the ratio domestic credit to private sector to GDP, is from WDI. Black market exchange rates, which we interpolated to address the scarcity observations, are from the Economic Freedom the World Annual Report (EFW, Gwartney et al. (2013)). The party ideology variable is from the 2012 version the Database Political Institutions (Beck et al. (2001)). Fiscal decentralization on total spending is calculated using the World Bank s fiscal decentralization indicators. Price level investment (capital formation) is from PWT 8.0, while the real interest rate is from WDI. Last, real effective exchange rates data, used to create the uncertainty measure, are from WDI and WEO. The government accountability proxies executive constraints, democracy/autocracy, and voice and accountability are from Polity IV (Marshall et al. (2013)), Polity IV, and the Worldwide Governance Indicators (WGI, Kaufmann et al. (2010)), respectively, while freedom the press is from Freedom House (Dunham et al. (2015)). The law enforceability proxy is from EFW. Corruption proxies, corruption and control corruption, are from the International Country Risk Guide and WGI, respectively. Last, PPP-adjusted real GDP per capita data, used to classify countries by income level, is from WEO. 4

5 3. Stylized facts on the relationship between public spending and growth This section the online appendix complements section 2.3 Stylized facts - the paper, by presenting further simple evidence indicative the key role government accountability in the nexus between public spending and growth. Figure 1 plots public capital and current spending as a share GDP against growth rates, both based on the 8-year non-overlapping averages data used in the regression analysis the paper (covering 80 countries). Constraints is used to classify countries by accountability levels. Being in line with Figure 1 the paper, the upper-left subfigure shows that particularly under high accountability, the relation between capital spending and growth is significantly positive. Although excluding 3 observations for, located in the north east the subfigure, weakens the relation (see dashed fitted line), the statistical significance remains. Still, it is important to acknowledge that this is an indication that the relation is sensitive to the inclusion/exclusion certain observations. Figure 1: Scatterplot: public spending/gdp and growth (with non-overlapping averages) GDP p.c. growth (over 8 yrs, percent) High accountability Low accountability Korea, Republic Thailand Korea, Republic Lesotho Estonia Singapore Singapore Latvia Mauritius Korea, Republic Lithuania Chile Egypt Cyprus Slovak Republic Luxembourg Mauritius Cyprus Korea, Republic Croatia Austria Kingdom Austria Denmark France Luxembourg GermanyItalyLuxembourg Germany Netherlands Japan Norway Sweden Canada Finland Lithuania Singapore Russian Thailand Cameroon Tunisia Federation PolandPoland Slovenia Cyprus India Japan Finland Slovak Spain Uruguay Mauritius Italy Norway Spain Portugal Spain Finland Colombia Sri Republic Romania Lanka Sri Lanka Chile Thailand Indonesia Malawi Uganda South Australia Czech Cyprus Israel Kingdom Netherlands Germany Denmark Belgium Austria Africa Republic Estonia Sri Lanka Hungary Fiji Chile Egypt Panama Bahrain, Jordan Kingdom Iran, I.R. Pakistan Thailand Tunisia Egypt Mongolia Singapore Egypt Iran, I.R. Morocco Kingdom Austria Kingdom Germany India Australia Italy Israel Belgium Denmark France Sweden France Italy Portugal Norway Luxembourg Mauritius Germany South Australia Costa Sweden Canada Netherlands Finland Israel Africa Rica Sri Fiji Lanka Russian Federation Bahrain, India Kingdom Spain Sweden Luxembourg Cyprus Mexico Guatemala Mexico El Salvador Bolivia Panama Bahrain, Kingdom Panama Paraguay Panama Peru Kingdom Tunisia Hungary Bulgaria Jordan Pakistan Morocco Kenya Bolivia Indonesia Thailand Tunisia Morocco Denmark France Portugal Spain Finland Norway New South Zealand Uruguay Africa Fiji Philippines Chile Mexico Iran, Cameroon Tanzania Swaziland I.R. Italy South Africa Kenya Ukraine Kenya Namibia South Uruguay Africa Hungary Burundi Paraguay Kuwait Gabon Malawi Gabon PeruBurundi Jordan Togo Iran, I.R. Bahrain, Kingdom Capital spend/gdp (avg over 8 yrs, percent) GDP p.c. growth (over 8 yrs, percent) Japan India India High accountability Estonia Mauritius Latvia Cyprus Lithuania Mauritius Cyprus Luxembourg Slovak Republic Lithuania Croatia Cyprus Japan Austria France Germany Italy Luxembourg Norway Luxembourg Finland India Mauritius Japan Portugal Sri Lanka Slovenia Spain Uruguay Sri Sri Lanka Norway Denmark Kingdom Sweden Canada Finland Finland Portugal Spain Spain Slovak Republic Fiji Lanka ColombiaEstonia Portugal Australia Kingdom Germany Germany Netherlands Austria Austria Kingdom Austria Italy Luxembourg Portugal Denmark Denmark Norway Kingdom Sweden Belgium Germany Belgium Germany Denmark France France Italy France Italy Netherlands Australia Australia South Czech Africa Republic Cyprus Israel Sri Lanka Mauritius Costa Rica Fiji Australia Sweden Israel Canada Finland Spain South Africa Israel Luxembourg Netherlands Kingdom Netherlands Finland Norway Sweden Denmark France Cyprus South Africa Fiji Spain New Zealand Uruguay Portugal South Africa Uruguay Italy South Africa Korea, Thailand Republic Korea, Lesotho Republic Singapore Singapore Korea, Republic Chile Korea, Republic Cameroon Singapore Tunisia Russian Federation Thailand Poland Poland Indonesia Uganda Malawi Romania Bahrain, Chile Chile Kingdom Jordan Iran, I.R. Egypt Pakistan Thailand Thailand Tunisia Singapore Iran, I.R. Mongolia Morocco Egypt Panama Bahrain, Guatemala Mexico Mexico Bolivia Peru Bahrain, El Salvador Panama Kingdom Kingdom Jordan Pakistan Kenya Russian FederationTunisia Paraguay Indonesia Thailand Bolivia Panama MoroccoMorocco Cameroon PanamaTunisia Bulgaria Swaziland Philippines Iran, Mexico I.R. Tanzania Chile Kenya Kenya Ukraine Namibia Paraguay Gabon Malawi Burundi Gabon Peru Burundi Jordan Togo Iran, I.R. Bahrain, Kingdom Current spend/gdp (avg over 8 yrs, percent) Low accountability Egypt Hungary Egypt Hungary Kuwait Hungary Note: Dashed fitted line is for a sample excluding Source: Authors' calculations, using 8-year periods data 5

6 Further, we report that, even when other variables are controlled for, OLS estimators still confirm the econometric analysis based on system-gmm estimators reported in the paper. Specifically, Figures 2 and 3, which are added variable plots 1 based on OLS estimations the regression equation Table 3 the paper, with standard errors clustered by country, show that capital spending (scaled by GDP) is still positively associated with growth, particularly under highgovernment accountability. Notice that this is the case, even when financing factors are specified, for both constraints (Figure 2) and voice (Figure 3) as accountability proxies. Table 1 presents the corresponding estimation results. 1 These added variable plots (also known as partial regression plots) are made using the Stata command avplot. Note that this avplot command works with OLS regressions, but not with other estimation methods, such as system-gmm. 6

7 Figure 2: Added variable plots: the capital spending-growth nexus across accountability levels ( constraints used as a proxy) Funded by Current Spending High Constraints Low Constraints e(gdppc growth X) Korea, Republic Kuwait Egypt Poland Lesotho Thailand Jordan Korea, Singapore Panama Republic Chile Korea, Republic Bahrain, Luxembourg Mauritius Iran, Kingdom I.R. Slovak Thailand Cameroon Bahrain, Pakistan Latvia Uganda Republic Morocco Lithuania Kingdom Malawi Indonesia Tunisia Thailand Russian Bahrain, Singapore Denmark South Israel Israel Italy Finland Kingdom Uruguay Croatia Mexico Poland Morocco Egypt Federation Estonia Kingdom Cyprus Guatemala Italy Egypt Sweden Kingdom Belgium Guatemala Africa India Netherlands Kenya Sweden Kingdom Germany Swaziland Canada Egypt Spain Thailand Luxembourg Mexico Luxembourg Colombia Lithuania Cyprus Mauritius Belgium Tunisia Romania Chile Japan HungaryCyprus Netherlands France Netherlands South Israel Austria France Italy Norway Portugal Bolivia Spain Austria Tunisia Finland Iran, Portugal Kingdom Korea, Kenya Denmark Netherlands Germany Panama Costa Australia Slovenia Canada Bolivia Austria Italy Spain Chile Mauritius Spain Hungary Luxembourg I.R. Mauritius Sri Lanka Republic Rica Africa India Pakistan Germany France Germany Tunisia Norway Panama Malawi Jordan Mexico Singapore Russian Finland Portugal Kingdom Chile Denmark Paraguay Cameroon Kenya Portugal Mongolia Australia Namibia Panama Norway Iran, Japan Sri I.R. Lanka Sri Lanka Bulgaria Cyprus Fiji Fiji Federation Morocco Peru Slovak Luxembourg Sri Lanka Republic Indonesia Uruguay Japan Thailand Estonia Italy El Salvador Sweden Czech Denmark Denmark Spain New Republic Finland Zealand Australia Philippines South Burundi South Sweden Singapore Iran, I.R. Africa Gabon India Norway Africa Fiji Tanzania Burundi Hungary Uruguay Finland South PeruAfrica Paraguay Jordan Gabon Togo Ukraine Bahrain, Kingdom e(gdppc growth X) Kuwait Korea, Republic Egypt Lesotho Poland Thailand Jordan Panama Singapore Korea, Korea, Republic Chile Republic Pakistan Bahrain, Mauritius Thailand Slovak Luxembourg Kingdom Latvia Cameroon Republic Bahrain, Kingdom Uganda Iran, Morocco I.R. Thailand Poland Bahrain, Singapore Mexico Kingdom Tunisia Lithuania Indonesia Malawi Morocco Italy Guatemala Egypt Mexico Hungary Guatemala South Africa India Thailand Colombia Denmark Finland Kingdom Tunisia Chile Kingdom Israel Uruguay Egypt Croatia Cyprus Russian Luxembourg Belgium Italy Egypt Hungary Portugal Panama Kenya France Mauritius Kingdom France Spain Kenya Tunisia Sweden Austria Germany Netherlands Spain Costa Italy Norway Finland Mauritius Japan Spain Belgium Romania Cyprus Sweden Lithuania Estonia Federation Germany Luxembourg Mauritius France Netherlands Cyprus Canada Austria Italy Portugal Austria Bolivia Kingdom Cyprus Swaziland Iran, Chile Korea, Rica Panama IndiaMexico South Bolivia Austria Netherlands Denmark Israel Sri Germany Republic Australia Canada I.R. Africa ChileSingapore Pakistan Paraguay Bulgaria Portugal Germany France Germany Tunisia Jordan Slovenia Lanka Panama Kenya Morocco Russian Portugal Namibia Finland Kingdom Norway Sri Lanka Japan Iran, I.R. Malawi Cyprus Norway Federation Uruguay Luxembourg Sri Mongolia Australia Denmark Australia Sri El Salvador Denmark Italy Japan Fiji Lanka Peru Fiji Lanka Slovak Cameroon Denmark Burundi IndiaSouth Finland Philippines Australia Spain Czech Sweden Indonesia Estonia Republic Republic Thailand Sweden Africa New Zealand Singapore Iran, I.R. South Tanzania Africa Gabon Norway Hungary Burundi Fiji Uruguay Finland South Africa Peru Paraguay Togo Jordan Ukraine Bahrain, Kingdom Gabon e(cap spend*highconst X) coef = , (robust) se = , t = e(cap spend*lowconst X) coef = , (robust) se = , t =.47 Funded by Revenue High Constraints Low Constraints e(gdppc growth X) Korea, Republic Egypt Kuwait Poland Lesotho Jordan Thailand Korea, Singapore Panama Republic Korea, Republic Bahrain, Chile Mauritius Luxembourg Iran, Kingdom I.R. Slovak Pakistan Thailand Cameroon Bahrain, Latvia Uganda Republic Morocco Lithuania Kingdom Malawi Indonesia Thailand Tunisia Russian Bahrain, Kingdom Mexico Singapore Denmark Italy Finland Israel Guatemala Israel Italy Egypt Mexico Uruguay Croatia Poland Morocco Egypt Federation Kingdom Estonia Cyprus Luxembourg Luxembourg Colombia Lithuania Cyprus Mauritius South Guatemala Thailand Sweden Japan Kingdom Tunisia Chile Romania Belgium Belgium Africa Netherlands India Sweden Swaziland Canada Kingdom Kenya Germany Netherlands Spain Iran, Egypt France Netherlands Kenya France Austria Italy Portugal Norway Germany South Israel Denmark Panama Australia Bolivia Canada Austria Finland Tunisia Spain Hungary Cyprus Cyprus Korea, Netherlands Costa Slovenia Spain Chile Portugal Kingdom Spain Mauritius I.R. Mauritius Sri Lanka Hungary Luxembourg Republic Africa Bolivia Austria ItalyRica Mexico Panama Tunisia Malawi India Germany Pakistan Germany France Paraguay Singapore Finland Cameroon Denmark Kenya Australia Portugal Mongolia Portugal Morocco Kingdom Norway Jordan Japan Iran, Sri Norway I.R. Lanka Sri Lanka Bulgaria Cyprus Fiji Fiji Russian Australia Panama Chile Namibia Slovak Luxembourg Federation Sri Lanka Republic Peru Estonia Japan Czech Italy Indonesia Uruguay Thailand El Salvador Sweden Denmark Spain Republic Finland New Zealand Australia Philippines South Burundi South Sweden Singapore Iran, Africa Gabon I.R. Norway Africa Fiji India Hungary Tanzania Burundi Uruguay Finland South Peru Africa Paraguay Jordan Gabon Togo Ukraine Bahrain, Kingdom e(gdppc growth X) Kuwait Korea, Republic Egypt Lesotho Poland Thailand Jordan Panama Korea, Singapore Republic Korea, Republic Chile Pakistan Bahrain, Mauritius Thailand Slovak Luxembourg Kingdom Latvia Cameroon Republic Bahrain, Uganda Kingdom Morocco Iran, I.R. Thailand Poland Mexico Bahrain, Singapore Kingdom Tunisia Lithuania Indonesia Malawi Morocco Italy Guatemala Mexico Guatemala Hungary South Egypt Denmark Uruguay Kingdom Africa India Thailand Colombia Japan Luxembourg Israel Finland Tunisia Chile Kingdom Kingdom Spain Portugal Hungary Kenya Panama Germany Spain Mauritius Costa France Kenya Mauritius Germany Egypt Austria Finland Luxembourg Mauritius Belgium Italy Norway Rica Tunisia Spain Cyprus Romania Israel Italy Cyprus Croatia Egypt Canada Sweden Belgium Sweden Lithuania Estonia Russian Federation Cyprus Austria France Korea, Italy IndiaMexico Panama South Bolivia Bolivia Portugal Austria Republic Africa Denmark Germany Australia Austria Chile Pakistan Portugal Germany Morocco Russian Paraguay Singapore Germany Netherlands Israel Chile Kingdom Cyprus Swaziland Netherlands Iran, Canada I.R. Sri Slovenia Lanka Kenya Panama Portugal Bulgaria Australia Finland France Tunisia Kingdom Norway Japan Sri Jordan Lanka Iran, I.R. Malawi Federation Namibia Cyprus Norway Uruguay Japan Luxembourg El Salvador Denmark Italy Sri Mongolia Denmark Australia Fiji Peru Fiji Sri Lanka Lanka Spain Australia Denmark Sweden Indonesia Slovak Cameroon Republic IndiaBurundi South Philippines Africa Finland Czech Estonia Republic Thailand New Sweden South Tanzania Africa Gabon Iran, Singapore Zealand I.R. Norway Burundi Uruguay Fiji Finland South Hungary Africa Peru Paraguay Togo Jordan Ukraine Bahrain, Kingdom Gabon e(cap spend*highconst X) coef = , (robust) se = , t = e(cap spend*lowconst X) coef = , (robust) se = , t =.64 Funded by Deficit e(gdppc growth X) High Constraints Korea, Republic Egypt Kuwait Lesotho Poland Thailand Korea, Singapore Jordan Panama Republic Chile Korea, Republic Mauritius Pakistan Thailand Cameroon Latvia Slovak Uganda Bahrain, Republic Bahrain, Luxembourg Iran, MoroccoKingdom Kingdom I.R. Lithuania Indonesia Malawi Thailand Singapore Tunisia Denmark South Israel Israel Finland Italy Kingdom Russian Bahrain, Uruguay Mexico Poland Morocco Croatia Guatemala Guatemala Sweden Egypt Africa IndiaSwaziland Sweden Kingdom Belgium Thailand Kingdom Netherlands Norway Kenya Germany Netherlands Netherlands Germany France Austria Canada Luxembourg Mexico Egypt Federation Kingdom Cyprus Estonia Tunisia Chile Romania Italy Luxembourg Japan Finland Egypt Spain France Italy Bolivia Kenya Israel India Pakistan Norway South Denmark Germany Slovenia Panama Australia Netherlands Germany Africa Bolivia Denmark Finland Germany Singapore Kingdom France Tunisia Austria Austria Kingdom Hungary Belgium Mauritius Colombia Lithuania Cyprus Tunisia Portugal Korea, Costa Republic Spain Chile Spain Spain Iran, Cyprus Cyprus Portugal I.R. Canada Italy Rica Hungary Mauritius Sri Lanka Luxembourg Panama Malawi Mexico Cameroon Paraguay Kenya Chile Mongolia Australia Australia Portugal Norway Jordan Iran, JapanI.R. Sri Lanka Russian Morocco Portugal Panama Namibia Bulgaria Cyprus Sri Lanka Peru Federation El Denmark Italy Uruguay Indonesia Japan Luxembourg Slovak Sri Fiji Republic Lanka Fiji Thailand Estonia Denmark Salvador Sweden Czech Republic Spain Australia Philippines Finland India Burundi South New Singapore Zealand Iran, Africa Sweden South Gabon I.R. Africa Fiji Norway Tanzania Burundi Finland Uruguay Hungary Peru South Africa Paraguay Jordan Togo Gabon Ukraine Bahrain, Kingdom e(cap spend*highconst X) coef = , (robust) se = , t = 1.33 e(gdppc growth X) Low Constraints Korea, Republic KuwaitEgypt Singapore Poland Korea, Republic Chile Panama ThailandLesotho Korea, Republic Pakistan Mauritius Jordan Bahrain, Thailand Cameroon Slovak Singapore Bahrain, Kingdom Poland Luxembourg Latvia Kingdom Republic Tunisia Indonesia Bahrain, Chile Guatemala Guatemala Russian Egypt Italy South Egypt India Africa Hungary Denmark Israel Federation Kingdom Thailand Lithuania Kingdom Uruguay Luxembourg Belgium Thailand Colombia Kingdom Tunisia Mexico Kingdom Israel Finland Croatia Uganda Kenya Mauritius Panama India Kenya France France Spain Swaziland Sweden Spain Netherlands Singapore Chile South Costa Germany Portugal Hungary Austria France Tunisia Finland Belgium Italy Cyprus Mexico Malawi Iran, I.R. Luxembourg Sweden Norway Bolivia Pakistan Mexico Africa Rica Italy Chile Germany Mauritius Spain Italy Korea, Luxembourg Mauritius Cyprus Cyprus Japan Egypt Lithuania Estonia Kingdom Canada Netherlands Romania Portugal Austria Israel Republic Austria Austria Portugal Cyprus Netherlands Sri Denmark Germany Lanka Canada Paraguay Bulgaria Russian Panama Kingdom Uruguay Morocco Sri Mongolia Lanka Federation Namibia Germany Portugal Portugal France Tunisia Bolivia Finland Germany Jordan Panama Slovenia Australia Iran, I.R. El Salvador Luxembourg Cameroon Kenya Cyprus Norway Sri Norway Lanka Japan Denmark Fiji Iran, I.R. Fiji ItalyPeru Sri Australia Lanka Australia Malawi Singapore India Burundi Denmark Denmark South Africa Australia Finland Sweden Japan South Africa Sweden Philippines Spain Indonesia Slovak New Zealand Czech Estonia Republic Republic Thailand Gabon Norway Burundi Tanzania Fiji Iran, I.R. South Hungary Uruguay Africa Finland Paraguay Peru Togo Jordan Ukraine Bahrain, Kingdom Morocco Morocco Gabon e(cap spend*lowconst X) coef = , (robust) se = , t =

8 Figure 3: Added variable plots: the capital spending-growth nexus across accountability levels ( voice used as a proxy) Funded by Current Spending High accountability Low accountability e(gdppc growth X) Korea, Republic EgyptKuwait Singapore Lesotho Thailand Panama Jordan Korea, Republic Poland Bahrain, Korea, Slovak Bahrain, Bahrain, Iran, Kingdom Chile Cameroon Republic Colombia Kingdom Kingdom I.R. Luxembourg Mauritius Croatia Russian Lithuania Federation Pakistan Thailand Latvia Singapore Indonesia Tunisia Uganda Morocco Sri Lanka Cyprus Thailand Denmark Malawi Israel Italy Poland Finland Kingdom Uruguay Mexico Morocco Egypt Guatemala Italy Thailand Sweden Luxembourg Kingdom Egypt Tunisia Mexico Belgium Japan Romania Luxembourg Estonia Lithuania Cyprus Chile Mauritius South Guatemala Israel Belgium India Africa Swaziland Sweden Egypt Canada Kingdom Spain Netherlands Germany Netherlands Tunisia Sri Spain Kingdom France Austria Kenya France Austria Hungary Iran, Bolivia Austria Finland Portugal Spain Lanka Cyprus Cyprus Luxembourg Italy Norway Netherlands Germany Denmark Germany Portugal South India Singapore Germany Israel Kenya Fiji Spain Netherlands Costa Sri Australia Panama Slovenia Canada Tunisia Austria Lanka Italy I.R. Mauritius Korea, Panama Iran, Fiji Rica Pakistan Germany France Norway Bolivia Mexico Jordan Republic Finland Kingdom Africa Malawi Panama Hungary Denmark Cameroon Paraguay Portugal Australia Portugal Namibia Norway Japan I.R. Chile Russian Cyprus Federation Kenya Morocco Mongolia Australia Peru Luxembourg Bulgaria Denmark Sweden Italy Indonesia Japan Uruguay Chile Slovak Singapore Thailand Estonia Republic El Czech Denmark Salvador Spain New Finland Fiji Republic Australia South Philippines Burundi Gabon Zealand Sweden Iran, Africa I.R. India Norway South Africa Tanzania Burundi Uruguay Finland South PeruAfrica Gabon Paraguay Jordan TogoHungary Ukraine Bahrain, Kingdom e(gdppc growth X) Kuwait Egypt Korea, Republic Singapore Thailand Lesotho Jordan Panama Korea, Poland Republic Bahrain, Korea, CroatiaBahrain, Colombia Pakistan Mauritius Slovak Kingdom Thailand Republic Luxembourg Republic Bahrain, Iran, Kingdom I.R. Latvia Cameroon Chile Singapore Kingdom Tunisia EgyptMexico Mexico Italy Thailand Sri Russian Egypt Kingdom Lanka Lithuania Federation Indonesia Uganda Morocco Guatemala Denmark Israel Finland Poland Malawi Morocco Tunisia Thailand Romania Kingdom Uruguay Italy India Guatemala South Egypt Belgium Luxembourg Tunisia Kingdom Belgium Africa Japan Israel Sweden Cyprus Lithuania Estonia Sri Sweden Portugal France Spain Austria Finland Lanka Netherlands Germany Mauritius Spain Sri Norway Kenya Germany France Luxembourg Austria Cyprus Cyprus Spain Chile Netherlands Kingdom Italy Italy India Panama Panama Mexico Singapore Kenya Costa Austria Fiji Mauritius Portugal Bolivia Austria Netherlands Lanka Russian Panama Germany South Bolivia France Tunisia Denmark Germany Mauritius Canada Hungary Sri Cyprus Lanka Australia Germany Jordan Israel Slovenia Swaziland Iran, Canada I.R. Fiji Korea, Rica Federation Namibia Pakistan Kingdom Africa Republic Iran, I.R. Finland Hungary Norway Chile Japan Malawi Morocco Paraguay Portugal Chile Kenya Mongolia Cyprus Australia Denmark Uruguay Luxembourg Denmark Italy Peru Australia Cameroon El Salvador Denmark Japan Fiji Singapore Sweden Bulgaria India Burundi South Sweden Finland Australia Spain Czech Slovak Indonesia Estonia Republic Republic Thailand Philippines New Gabon Africa Zealand South Tanzania Africa Norway Iran, I.R. Burundi Uruguay Finland South Peru Africa Paraguay Togo Jordan Hungary Ukraine Bahrain, Kingdom Gabon e(cap spend*highvaccount X) coef = , (robust) se = , t = e(cap spend*lowvaccount X) coef = , (robust) se = , t =.6 Funded by Revenue High accountability Low accountability e(gdppc growth X) Egypt Kuwait Korea, Singapore Republic Lesotho Thailand Panama Jordan Korea, Bahrain, Bahrain, Poland Republic Iran, Kingdom Kingdom I.R. Korea, Cameroon Slovak Bahrain, Luxembourg Mauritius Croatia Chile Russian Colombia Republic Lithuania Kingdom Federation Pakistan Thailand Latvia Singapore Indonesia Tunisia Uganda Morocco Sri Lanka Cyprus Denmark Malawi Thailand Italy Poland Kingdom Finland Mexico Guatemala Israel Thailand Italy Mexico Uruguay Morocco Egypt Sweden Luxembourg Kingdom Tunisia Japan Egypt Romania Luxembourg Estonia Lithuania Cyprus Belgium Israel South Netherlands Guatemala India Belgium Egypt Chile Mauritius Cyprus Cyprus Africa Kenya Germany Germany Swaziland Kingdom France Canada Sweden Spain Netherlands Austria France Austria Iran, Italy Netherlands Germany Norway Denmark Australia Portugal Bolivia Hungary Austria Finland Tunisia Spain Sri Portugal Spain Kingdom Fiji Netherlands Costa Sri Canada Spain Panama Slovenia Lanka South India Singapore Germany Kenya Israel Africa Germany Pakistan Denmark Finland France Norway Mexico Austria Italy I.R. Luxembourg Lanka Mauritius Korea, Panama Fiji Tunisia Iran, Bolivia Japan Republic Jordan Rica I.R. Chile Panama Hungary Kingdom Malawi Namibia Norway Cyprus Russian Federation Cameroon Paraguay Australia Portugal Kenya Morocco Mongolia Australia Luxembourg Peru Chile Slovak Bulgaria Republic Denmark Sweden Denmark Italy Indonesia Japan Uruguay Estonia El Czech Singapore Thailand Salvador Spain Finland New Fiji Republic Zealand Australia Philippines South Burundi Sweden Iran, Gabon Africa I.R. IndiaNorway South Africa Tanzania Burundi Uruguay Finland South Peru Africa Gabon Paraguay Jordan TogoHungary Ukraine Bahrain, Kingdom e(gdppc growth X) Kuwait Egypt Korea, Republic Singapore Thailand Lesotho Jordan Panama Korea, Poland Republic Croatia Korea, Bahrain, Pakistan Colombia Mauritius Republic Thailand Slovak Kingdom Bahrain, Kingdom Iran, I.R. Chile Latvia Cameroon Republic Luxembourg Bahrain, Singapore Kingdom Thailand Sri Tunisia Italy Kingdom Lanka Mexico Mexico Guatemala Egypt Denmark South Thailand Tunisia Kingdom Uruguay Israel Finland Poland Egypt Lithuania Russian Indonesia Federation Uganda Morocco Malawi Morocco Italy Africa India Guatemala Egypt Luxembourg Japan Kingdom Spain Portugal Germany Spain Mauritius France Kenya Germany Austria Finland Norway Tunisia Romania Chile Belgium Belgium Israel Sweden Cyprus Lithuania Estonia Sri Sweden Canada Spain Lanka Luxembourg Mauritius Austria Cyprus France Mauritius Cyprus Italy Italy India Mexico Panama Kenya Costa Austria Sri Bolivia Portugal Austria Kingdom Netherlands Netherlands Germany Denmark Australia Rica Lanka Canada Hungary Iran, Sri Swaziland Lanka Cyprus I.R. Fiji Russian Pakistan Portugal Germany South Singapore Korea, Bolivia Germany Africa Panama Kingdom Federation Morocco Paraguay Portugal Australia Namibia Finland France Israel Republic Slovenia Fiji Chile Mongolia Tunisia Iran, I.R. Norway Japan Jordan Hungary Norway Malawi Denmark Cyprus Kenya Uruguay Chile Luxembourg El Salvador Japan Denmark Italy Peru Australia Cameroon Denmark Spain Fiji Czech Singapore Sweden Slovak Bulgaria Indonesia Estonia Republic Republic Thailand India Australia Burundi South Philippines Finland Sweden Gabon New AfricaZealand Norway Iran, I.R. South Tanzania Africa Burundi Uruguay Finland South Peru Africa Paraguay Togo Jordan Hungary Ukraine Bahrain, Kingdom Gabon e(cap spend*highvaccount X) coef = , (robust) se = , t = e(cap spend*lowvaccount X) coef = , (robust) se = , t =.7 Funded by Deficit High accountability Low accountability e(gdppc growth X) Korea, Republic Egypt Kuwait Singapore Lesotho Thailand Panama Jordan Poland Korea, Republic Korea, Cameroon Slovak Bahrain, Bahrain, Iran, Pakistan Thailand Latvia Republic Republic Kingdom Luxembourg Colombia Chile Croatia Kingdom Mauritius I.R. Bahrain, Russian Singapore Indonesia Tunisia Uganda Sri Morocco Lanka Lithuania Kingdom Federation Denmark Malawi ThailandEgypt Cyprus Finland Kingdom Israel Poland Italy Uruguay Mexico Guatemala South India Israel Luxembourg Guatemala Africa Swaziland Canada Kingdom Belgium Thailand Sweden Mexico Egypt Kingdom Italy Morocco Tunisia Luxembourg Romania Japan Belgium Estonia Lithuania Cyprus Sweden Netherlands Germany Netherlands Norway Germany France Kenya Austria France Spain Egypt Italy Finland Bolivia Austria Tunisia Spain Chile Mauritius Spain Sri Kingdom Lanka Cyprus Cyprus Iran, Hungary Netherlands Denmark Germany Fiji I.R. Korea, Netherlands Sri Slovenia Australia Austria Lanka Norway India Kenya South Singapore Germany Israel Costa Panama Republic Portugal Spain Canada Portugal Italy Luxembourg Mauritius Panama Fiji France Pakistan Finland Germany Kingdom Africa Tunisia Mexico Iran, Bolivia Jordan Japan Rica Malawi Denmark Cameroon Paraguay Australia Kenya Morocco Portugal Mongolia Australia Portugal Norway Hungary I.R. Chile Russian Panama Namibia Federation Cyprus Italy Denmark Indonesia Uruguay Chile Luxembourg Peru Bulgaria Slovak Republic Japan Denmark Sweden El Czech Singapore Thailand Estonia Salvador Spain Fiji Republic India Australia Sweden Philippines Burundi South Finland New Gabon Zealand Iran, Africa I.R. South Africa NorwayTanzania Burundi Finland Uruguay Paraguay Peru South Jordan Africa Gabon Togo Hungary Ukraine Bahrain, Kingdom e(gdppc growth X) Egypt Singapore Kuwait Korea, Republic Thailand Lesotho Panama Poland Croatia Singapore Bahrain, Kingdom Pakistan Korea, Mauritius Bahrain, Republic Korea, Republic Jordan Russian Federation Tunisia Colombia Thailand Slovak Cameroon Chile Latvia Kingdom Republic Egypt Indonesia Luxembourg Bahrain, Kingdom Iran, I.R. Egypt Guatemala Denmark Italy Kingdom Guatemala India South Israel Luxembourg Thailand Belgium Kingdom Africa Tunisia Uruguay Mexico Finland Poland Thailand Lithuania Israel Kingdom Sweden Kenya Singapore Panama India Mauritius France Kenya Netherlands Spain France Spain Swaziland Tunisia Chile Belgium Sweden Egypt Italy Luxembourg Romania Japan Cyprus Mexico Sri Lanka Uganda Malawi Lithuania Estonia Morocco Austria Finland Germany Fiji South Mexico Costa Portugal Norway Bolivia France Netherlands Cyprus Italy Kingdom Italy Africa Netherlands Austria Sri Luxembourg Rica Russian Paraguay Federation Pakistan Panama Kingdom Chile Morocco Mongolia Namibia Germany France Tunisia Bolivia Israel Lanka Jordan Denmark Germany Mauritius Germany Mauritius Cyprus Canada Spain Hungary Austria Sri Lanka Morocco Panama Slovenia Canada Portugal Cyprus Sri Australia Fiji Iran, Lanka I.R. Finland Hungary Germany Norway Korea, Chile Portugal India Singapore El Salvador Uruguay Luxembourg Cameroon Cyprus Kenya Italy Peru Denmark Norway Republic Japan Iran, I.R. Australia Denmark Denmark Bulgaria Malawi Burundi FijiSweden Japan Indonesia South Australia Sweden Philippines Gabon Africa New Finland Czech Spain Slovak Estonia Republic Republic Thailand Zealand South Africa Norway Burundi Uruguay Tanzania Iran, I.R. South Africa Paraguay Peru Finland Togo Jordan Ukraine Hungary Gabon Bahrain, Kingdom e(cap spend*highvaccount X) coef = , (robust) se = , t = e(cap spend*lowvaccount X) coef = , (robust) se = , t =

9 Table 1: OLS Results (with standard errors clustered by country) Regressors (1) (2) (3) (4) (5) (6) Cap spend*highacc 2.898** 3.538*** 2.908** 3.519*** * (2.021) (2.693) (2.054) (2.723) (1.326) (1.713) Cap spend*lowacc (0.474) (0.602) (0.641) (0.704) (-1.381) (-0.959) Cur spend*highacc *** *** (0.068) (-0.122) (-2.793) (-2.791) Cur spend*lowacc *** ** (0.447) (-0.002) (-3.454) (-2.586) Revenue*Highacc *** 1.042*** (0.068) (-0.122) (2.798) (2.737) Revenue*Lowacc *** 1.451*** (0.447) (-0.002) (3.735) (2.770) Deficit*Highacc *** *** *** *** (-2.793) (-2.791) (-2.798) (-2.737) Deficit*Lowacc *** ** *** *** (-3.454) (-2.586) (-3.735) (-2.770) Initial GDP p.c *** *** *** *** *** *** (-3.477) (-3.784) (-3.477) (-3.784) (-3.477) (-3.784) Initial Schooling 1.136* 1.266** 1.136* 1.266** 1.136* 1.266** (1.811) (2.222) (1.811) (2.222) (1.811) (2.222) Private inv/gdp 1.100*** 1.137*** 1.100*** 1.137*** 1.100*** 1.137*** (4.245) (4.375) (4.245) (4.375) (4.245) (4.375) Pop growth ** ** ** ** ** ** (-2.514) (-2.304) (-2.514) (-2.304) (-2.514) (-2.304) Financing source Cur spend Cur spend Revenue Revenue Deficit Deficit Accountability proxy Const Voice Const Voice Const Voice Observations No. countries Adjusted R-squared Wald Cap Spend Wald Cur Spend Notes: The dependent variable is the GDP per capita growth rate over 8 years. Pooled OLS estimations, with robust standard errors clustered by country (which take within-country serial correlation into account) in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. Wald, Cap spend (Cur spend) tests the equality coefficients on capital (current) spending across different accountability levels. 9

10 4. Estimation results with democracy proxies Table 2: Role government accountability using democracy/autocracy" as a proxy Interactions with accountability level With accountability and income levels With accountability and enforcement levels Regressors (1) (2) (3) (4) (5) (6) (7) (8) (9) Capital spend*highacc 7.092*** 7.597*** 6.286** (2.329) (2.266) (2.565) Capital spend*lowacc 2.865* 3.640** (1.699) (1.649) (1.357) Capital spend*highinc*highacc 6.872*** 7.171*** 5.921** (2.441) (2.405) (2.562) Capital spend*highinc*lowacc (2.579) (2.390) (2.496) Capital spend*lowinc*highacc 6.866* 7.165** (3.513) (3.317) (3.823) Capital spend*lowinc*lowacc ** (1.357) (1.315) (1.241) Capital spend*highenf*highacc 6.166** 6.603** 5.155* (2.747) (2.718) (2.989) Capital spend*highenf*lowacc * (1.797) (1.715) (1.701) Capital spend*lowenf*highacc 5.859** 6.296*** 4.849* (2.531) (2.379) (2.644) Capital spend*lowenf*lowacc * (1.694) (1.656) (1.441) Current spend*highacc 0.505** * (0.248) (0.639) (0.312) (0.583) (0.292) (0.602) Current spend*lowacc 0.775*** ** 0.738** *** 0.655** ** (0.290) (0.852) (0.323) (0.737) (0.309) (0.752) Revenue*Highacc 0.505** 1.312** * ** (0.248) (0.658) (0.312) (0.665) (0.292) (0.603) Revenue*Lowacc 0.775*** 2.615*** 0.738** 2.806*** 0.655** 2.249*** (0.290) (0.847) (0.323) (0.813) (0.309) (0.760) Budget deficit*highacc ** * * ** (0.639) (0.658) (0.583) (0.665) (0.602) (0.603) Budget deficit*lowacc ** *** *** *** ** *** (0.852) (0.847) (0.737) (0.813) (0.752) (0.760) 10

11 Regressors (1) (2) (3) (4) (5) (6) (7) (8) (9) Initial GDP p.c. (log) *** *** *** * * * ** ** ** (3.259) (3.259) (3.259) (5.120) (5.120) (5.120) (4.218) (4.218) (4.218) Initial level schooling (1.519) (1.519) (1.519) (1.219) (1.219) (1.219) (1.400) (1.400) (1.400) Private investment/gdp 1.606*** 1.606*** 1.606*** 1.414** 1.414** 1.414** 1.650*** 1.650*** 1.650*** (0.557) (0.557) (0.557) (0.581) (0.581) (0.581) (0.443) (0.443) (0.443) Population growth ** ** ** ** ** ** ** ** ** (2.919) (2.919) (2.919) (2.361) (2.361) (2.361) (2.663) (2.663) (2.663) Financing source Cur spend Revenue Deficit Cur spend Revenue Deficit Cur spend Revenue Deficit Observations No. countries No. instruments Arellano-Bond AR(1), p-value Arellano-Bond AR(2), p-value Hansen, p-value Diff Hansen 1, p-value Diff Hansen 2, p-value Wald Cap spend, p-value Wald Cur spend, p-value Wald Cap spend, Highacc, Income Wald Cap spend, Lowacc, Income Wald Cap spend, Highacc, Enforcement Wald Cap spend, Lowacc, Enforcement Notes: The dependent variable is the GDP per capita growth rate over 8 years. System GMM estimations for dynamic panel data models. Constant and time dummies are not shown for brevity. All explanatory variables were treated as endogenous except for initial GDP p.c. and initial schooling years, which were treated as predetermined. Orthogonal deviation was used to transform variables. Only one lag was used as an internal instrument to reduce the number instruments. Robust standard errors are in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. Diff Hansen 1 tests the exogeneity the instruments used in the level part ( the system) as a whole. Diff Hansen 2 tests the exogeneity the lagged level output used as an instrument in the level part. Wald, Cap spend (Cur spend) tests the equality coefficients on capital (current) spending across different accountability levels. Wald, Cap spend, Highacc (Lowacc), Income tests the equality coefficients on capital spending across different income levels for countries with accountable (unaccountable) governments. Wald, Cap spend, Highacc (Lowacc), Enforcement tests the equality coefficients on capital spending across different law enforcement levels for countries with accountable (unaccountable) governments. 11

12 Table 3: Role government accountability using XPOLITY (Vreeland s correction) as a proxy Interactions with accountability level With accountability and income levels With accountability and enforcement levels Regressors (1) (2) (3) (4) (5) (6) (7) (8) (9) Capital spend*highacc 7.008*** 7.508*** 6.162** (2.350) (2.277) (2.573) Capital spend*lowacc 2.851* 3.622** (1.682) (1.622) (1.341) Capital spend*highinc*highacc 6.895*** 7.204*** 5.902** (2.433) (2.392) (2.557) Capital spend*highinc*lowacc (2.534) (2.346) (2.457) Capital spend*lowinc*highacc 6.522* 6.831** (3.449) (3.244) (3.735) Capital spend*lowinc*lowacc ** (1.327) (1.286) (1.219) Capital spend*highenf*highacc 6.236** 6.671** 5.205* (2.821) (2.798) (3.076) Capital spend*highenf*lowacc * (1.759) (1.669) (1.673) Capital spend*lowenf*highacc 5.713** 6.147** 4.681* (2.489) (2.329) (2.602) Capital spend*lowenf*lowacc * (1.638) (1.588) (1.386) Current spend*highacc 0.500** * * (0.244) (0.627) (0.314) (0.568) (0.288) (0.598) Current spend*lowacc 0.772*** ** 0.747** *** 0.656** ** (0.289) (0.850) (0.322) (0.739) (0.306) (0.750) Revenue*Highacc 0.500** 1.347** ** ** (0.244) (0.654) (0.314) (0.648) (0.288) (0.600) Revenue*Lowacc 0.772*** 2.589*** 0.747** 2.802*** 0.656** 2.234*** (0.289) (0.840) (0.322) (0.818) (0.306) (0.755) Budget deficit*highacc ** * ** * ** (0.627) (0.654) (0.568) (0.648) (0.598) (0.600) Budget deficit*lowacc ** *** *** *** ** *** (0.850) (0.840) (0.739) (0.818) (0.750) (0.755) 12

13 Regressors (1) (2) (3) (4) (5) (6) (7) (8) (9) Initial GDP p.c. (log) *** *** *** * * * ** ** ** (3.233) (3.233) (3.233) (5.118) (5.118) (5.118) (4.230) (4.230) (4.230) Initial level schooling (1.515) (1.515) (1.515) (1.205) (1.205) (1.205) (1.387) (1.387) (1.387) Private investment/gdp 1.595*** 1.595*** 1.595*** 1.420** 1.420** 1.420** 1.642*** 1.642*** 1.642*** (0.553) (0.553) (0.553) (0.583) (0.583) (0.583) (0.444) (0.444) (0.444) Population growth ** ** ** ** ** ** ** ** ** (2.894) (2.894) (2.894) (2.342) (2.342) (2.342) (2.620) (2.620) (2.620) Financing source Cur spend Revenue Deficit Cur spend Revenue Deficit Cur spend Revenue Deficit Observations No. countries No. instruments Arellano-Bond AR(1), p-value Arellano-Bond AR(2), p-value Hansen, p-value Diff Hansen 1, p-value Diff Hansen 2, p-value Wald Cap spend, p-value Wald Cur spend, p-value Wald Cap spend, Highacc, Income Wald Cap spend, Lowacc, Income Wald Cap spend, Highacc, Enforcement Wald Cap spend, Lowacc, Enforcement Notes: The dependent variable is the GDP per capita growth rate over 8 years. System GMM estimations for dynamic panel data models. Constant and time dummies are not shown for brevity. All explanatory variables were treated as endogenous except for initial GDP p.c. and initial schooling years, which were treated as predetermined. Orthogonal deviation was used to transform variables. Only one lag was used as an internal instrument to reduce the number instruments. Robust standard errors are in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. Diff Hansen 1 tests the exogeneity the instruments used in the level part ( the system) as a whole. Diff Hansen 2 tests the exogeneity the lagged level output used as an instrument in the level part. Wald, Cap spend (Cur spend) tests the equality coefficients on capital (current) spending across different accountability levels. Wald, Cap spend, Highacc (Lowacc), Income tests the equality coefficients on capital spending across different income levels for countries with accountable (unaccountable) governments. Wald, Cap spend, Highacc (Lowacc), Enforcement tests the equality coefficients on capital spending across different law enforcement levels for countries with accountable (unaccountable) governments. 13

14 5. Exploiting time variations in institutional proxies Table 4: Interaction between public spending and government accountability Regressors (1) (2) (3) (4) (5) (6) Capital spend/gdp (1.804) (1.743) (1.810) (1.651) (1.570) (1.630) Capital spend/gdp*account 1.101** 0.645* 1.172** 0.639** 0.985** 0.572* (0.468) (0.324) (0.447) (0.310) (0.450) (0.303) Current spend/gdp (0.944) (0.779) (1.355) (1.151) Current spend/gdp*account (0.155) (0.087) (0.246) (0.137) Revenue/GDP (0.944) (0.779) (1.443) (1.199) Revenue/GDP*Account (0.155) (0.087) (0.270) (0.150) Budget deficit/gdp (1.355) (1.151) (1.443) (1.199) Budget Deficit*Account (0.246) (0.137) (0.270) (0.150) Account(ability) (4.411) (2.349) (4.411) (2.349) (4.411) (2.349) Initial GDP p.c. (log) *** *** *** *** *** *** (2.837) (2.809) (2.837) (2.809) (2.837) (2.809) Initial level schooling 2.864** 2.343* 2.864** 2.343* 2.864** 2.343* (1.413) (1.228) (1.413) (1.228) (1.413) (1.228) Private investment/gdp 1.556*** 1.656*** 1.556*** 1.656*** 1.556*** 1.656*** (0.447) (0.427) (0.447) (0.427) (0.447) (0.427) Population growth ** *** ** *** ** *** (2.936) (2.486) (2.936) (2.486) (2.936) (2.486) Financing source Cur spend Cur spend Revenue Revenue Deficit Deficit Accountability proxy Const Democ Const Democ Const Democ Observations No. countries No. instruments Arellano-Bond AR(1), p-value Arellano-Bond AR(2), p-value Hansen, p-value Diff Hansen 1, p-value Diff Hansen 2, p-value Notes: The dependent variable is the GDP per capita growth rate over 8 years. System GMM estimations for dynamic panel data models. Constant and time dummies are not shown for brevity. All explanatory variables were treated as endogenous except for initial GDP p.c. and initial schooling years, which were treated as predetermined. Orthogonal deviation was used to transform variables. Only one lag was used as an internal instrument to reduce the number instruments. Robust standard errors are in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. Diff Hansen 1 tests the exogeneity the instruments used in the level part ( the system) as a whole. Diff Hansen 2 tests the exogeneity the lagged level output used as an instrument in the level part. 14