MIME-Version: 1.0 Content-Type: multipart/related; boundary="----=_NextPart_01C66F82.9AD28F80" This document is a Single File Web Page, also known as a Web Archive file. If you are seeing this message, your browser or editor doesn't support Web Archive files. Please download a browser that supports Web Archive, such as Microsoft Internet Explorer. ------=_NextPart_01C66F82.9AD28F80 Content-Location: file:///C:/A55AE274/Uncorrupt.htm Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset="us-ascii" Uncorruption, Education, and Health

‘Uncorruption’, Education, and Health as Complementary Public Goods

(James Stodder and Peter Schrot= h, August 2005)

Journal of Economic Literature codes: K42, H41, O17

Key Words: Corruption and Growth, Public Goods, Cost Complementarities, Social Network= s, Health, Education

Total Word Count, includi= ng all notes: 9,995

 

’Uncorruption’, Education, and Health as Complementary Public Goods

 

ABSTRACT: In previous studies, freedom from corruption and the rule of law have b= een shown to promote public health and education.  Using international panel data, we present evidence that the reverse causality also holds:  these public goods are mutually reinforcing.  Anti-corruption incentives in developing countries, which have emphasized penal law, should= be linked to improvement of public goods that meet basic needs.

 

1.       Introduction : ‘Uncorrupt= ion’ as a Public Good

We do not want to know if life improves when Togo becomes Denmark; we want to know if life improves when a poor Togo becomes a richer Tog= o.        =             &nb= sp;           William Easterly (1999)

            &n= bsp;   Why are anti-corruption incentives are so difficult to implement?  A rationally self-interested state – what Mancur Olson (2000) calls a “stationary bandit” – would end widespread corruption in order to maximize the surplus it could extract.  It is elementa= ry to show that decentralized and uncontrolled graft makes a state too “wea= kly corrupt” for its own selfish good, by charging multiple corruption “taxes” that are too high to maximize its total tax revenue.[1]  That such states are too weak to r= eform themselves – too corrupt to rationalize their own corruption – suggests that the sources of corruption may be beyond direct state control.=

            &n= bsp;   Non-corruption is commonly referred to as ‘transparency,’ including by well kn= own groups such as Transparency International.  This may be somewhat misleading, be= cause better transparency is helpful, but neither a necessary nor a sufficient condition for freedom from corruption.  The UK, for example, is indexed as less corrupt than the US by Transparency International’s Corruption Perception Index (www.transparency.org).  Yet government in the US is probably more transparent than in th= e UK, which has an Official Secrets Act.[2]  The English language lacks a syste= m-characterizing word for “non-corruption” – terms like “integrity” seem too personalized.  Let us introduce the neologism of “uncorruption” to mean freedom from corruption – bearing in mind that all words like transparency or corruption are relative terms.  A neologism seems pref= erable to using a well-established word like transparency or integrity, but attemp= ting to restrict its meaning as a term of art.

Much of the new corruption lite= rature (Gupta et al. 1998, 2000; Azfar & Gurgur 2004) s= hows that uncorruption promotes health and education, even after accounting for = its indirect benefit through faster growth.  We show that the reverse is also true:  improvements in health = and education also encourage uncorruption; the causality is reciprocal.  The econometric methodology for sh= owing this is the same instrumental variable approach used by Gupta and others.

            &n= bsp;   Anti-corruption measures throughout the world have tended strongly to emphasize penal law, = only occasionally augmented by training for civil servants, accounting and discl= osure requirements and so forth.  Th= is emphasis continues in the recent United Nations Convention against Corrupti= on (UN 2003), as in previous treaties (OAS 1996, OECD 1998, Council of Europe 1999, SADC 2001, AU 2003), and in legislation from the earliest (in the United States, FCPA 1977, see Schroth 2002= ) to the most recent (notably that in South Africa, RSA 2004).  We are not aware of national or international initiatives to fight corruption by improving the provision of public goods, but the research reported in this paper suggests that such initiatives would be at least appropriate in support of, and possibly essen= tial to the success of, the anti-corruption efforts already underway.=

            &n= bsp;   We begin by considering why uncorruption should be improved by other voluntari= ly provided public goods.  We pro= pose both supply-side and demand-side explanations, namely cost complementarity = and increased resistance to free riding.  We then use two recent World Bank data sets, one developed by William Easterly (1999) in our part 5, the other by Daniel Kaufmann (2003) in our p= art 6, to estimate a model of the co-determination of health, education, and un= corruption.  The two data sets are from differe= nt periods (1980 to 1990 for Easterly; and 1996 to  2002 for Kaufmann) and were complied using different methods.  This probably accounts for most of the differences in our results from the two d= ata sets, but also adds some weight to results that are similar for the two.  We conclude with a discussion of s= ome policy implications of our results.

 

2.       The Parad= ox of Persistent, Immiserizing Corruption

            &n= bsp;   The “new economics of growth” literature shows a strong positive correlation between economic growth and uncorruption (Barro 1997).  Another current, “the new political economy of corruption” literature, has shown that corruption is negatively correlated with growth. (See Mauro 1995; Lopez & Mitra 2000; Gupta et al. = 2000; Reisen & Soto 2001.) Mauro (1997, p. 91) finds that <= /p>

if a given country were to improve its corruption “grade” from 6 out of 10 to 8 out of 10, its investment-GDP ratio would rise by alm= ost 4 percentage points and its annual growth of GDP per capita would rise by almost half a percentage point.

Generally similar conclusions w= ere reached by Burki & Perry (1998), Hines (1995), Wei (2000), and Rose-Ackerman (1978, 1= 997, 1999).

            &n= bsp;   Despite the conclusion these of studies that a government curtailing corruption sho= uld reap large payoffs in growth and welfare, effective anti-corruption incenti= ves are weak to non-existent in much of the developing world (e.g., Schroth 200= 3, 2005).  Why do most countries = find it so difficult to take measures that are in the economic interests of the great majority?  And why is th= is especially true for the poorest of the poor?

 

3.a. Uncorruption as a Public Good: Voluntary a= nd Governmental Provision

        &= nbsp;       Uncorruption is far from unique in being a public good that is produced primarily by voluntary provision within long-term social networks, instead of by direct governmental production, in the way that a government builds, say, a hydro-electric dam. [3]<= /span>  Many public goods are produced, in= whole or in part, by voluntary acts within social networks and communities, for reasons including fairly immediate self-interest, long-term reciprocity, and emotional commitments to values and other human beings. (See Frank & Scitovsky 1992; Axelrod & Cohen 2000.)=   Besides uncorruption, social networks also provide economic security= and insurance, personal security, collective defense, recreation, health care, = and education.  The last two goods= on that list are the focus of this paper.&nbs= p; Health care and education are goods that are public to the extent th= at one individual’s enjoyment of those goods does not always decrease, a= nd may well increase, the amount available to others.  That they are provided largely thro= ugh family and voluntary social networks is obvious, although state and private provision play an increasing role in developed economies.=

The impli= cit social contract to provide these public goods can be organized along what t= he political scientist Margaret Levi (1988) terms “vertical” or “horizontal” dimensions.  The vertical contract is that one is getting back fair value for one’s contributions to the central government, usually by paying taxe= s, but perhaps through other services (as an employee, soldier, elected offici= al, etc.)  A horizontal contract i= s that others similarly situated are contributing their fair share, through volunt= ary provision of the public good, or by just paying their taxes – as lega= lly required, but not necessarily enforced. Obviously the horizontal dimension = is of most importance for voluntarily provided public goods.  But a feeling that a contract is n= ot being respected and that one is being cheated along either dimension, vertically or horizontally, may have spillover effects for continued contributions along both dimensions, and towards other public goods.   Levi’s method is historical-comparative, but recent empirical tests confirm that trust toward others is correlated with voluntary tax compliance (Scholz and Lubell, 1998). 

        &= nbsp;       Since uncorruption is a public good produced both by central governmental provisi= on and by voluntary social networks, its provision should therefore be correla= ted with society’s success in providing other public goods, by either mea= ns, voluntary or governmental.  Th= is paper examines the correlation between corruption and two other such (partially) public goods – health and education.

3.b. &nbs= p; Cost Complementarities and Network-Provided Public Goods<= /span>

        &= nbsp;       If health care and education are both products of and contributors to the “social capital” accumulated within social networks, then their provision should not only be correl= ated with the provision of other public goods, such as uncorruption, but also he= lp determine the levels of their provision.  This is the argume= nt about causality that we address in the econometric methods of within-country panel estimates, in the next section.

        &= nbsp;       Consider first a simple model of cost-com= plementarity in the provision of Uncorruption (U) and Health outcomes (H); for the latte= r, we could as well substitute education (E).=   If H is a p= ublic good, its demand is found by the summ= ing prices in the demands of individuals 1 and 2:

Ph1 =3D a - bH,=        

  &nbs= p;            &= nbsp;           &nbs= p;    =3D>        &= nbsp;  Ph =3D Ph1 + Ph2   =3D   (a+c) - (b+d)H.  (1)

Ph2 =3D c - dH  &nb= sp;      

 

The demand for U is formed similarly:

 

           &nb= sp;    Pu1 =3D e - fU,  

  &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;       =3D>        &= nbsp;  Pu =3D Pu1 + Pu2      =3D    (e+g) = - (f+h)U.        &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;   (2)

  &nbs= p;             = Pu2 =3D g - hU   <= /span>

  &nbs= p;             =

H and U a= re cost complements in the sense that the more one is produced, the cheaper it= is to produce the other.  Simple = examples of total cost functions TC(H) and TC(U) with complementarities and Marginal Cost functions MC(H) and MC(U) would be:

            &n= bsp;   TC(H)  =3D j + kH + mH2 - nHU  =3D>   MC(H) = =3D (k -nU) + 2mH,              =             &nb= sp;            =             &nb= sp;            =             &nb= sp;   (3)

and=

        &= nbsp;       TC(U)  =3D q + rH + sH2 - tHU<= span style=3D'mso-tab-count:1'>        &= nbsp;       =3D>   MC(U) =3D (r -tH) + 2sU.        &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;  (4)

(Please Insert Figure 1 about here)

 

For param= eters such as those shown above, the relation between the provision of health care (H) and Uncorruption (U) is:  &nb= sp;  

  • If there is free riding, but with full provisio= n of H, then U will be undervalued by society.  The demand intercept shifts f= rom e + g down to e and we move from the efficient equilibrium at E1, down to E2.  The cost= of production is still correct.
  • If H falls, but with no free riding (so we have both layers of demand), this pushes U provision from E1 to E3 R= 11; i.e., from a marginal cost curve with an intercept of r-tH to a higher one at r.  Uncorruption is now being cor= rectly valued by demand, but has become too costly to provide adequately.
  • If there are both free riding and non-provision= of the complementary good H, uncorruption will fall even further, to a po= int such as E4.  If free riding or missing cost complements are severe enough, provision may disappear altogether.  In Figure 1, if the cost inte= rcept were greater than the demand intercept, r > e, then Uncorruption would then be too little valued, relativ= e to its cost, to be worth supplying at all.  

Of course, a graph such as Figure 1 could also be drawn for the reciprocal eff= ects of U (Uncorruption) upon H (Health).  Also, when under-provision of a public good is driven mostly by free-riding, as at E2, then th= e good will be under-priced.  = If under-provision is driven by missing cost complements, as at E3, then the good will be over= -priced.  An analysis of supply and demand conditions for several public goods within different economies would be required to disentangle the roots of under-provision, and we will not attem= pt such an analysis in this paper.   

It is important to say here that the present paper presents no evidence= to distinguish demand from supply factors.&nb= sp; However, whether the main problem is demand-side free riding or supp= ly-side cost complements, it should be clear that either source of under-provision can create cascading effects of under-provision:  <= o:p>

·         An under-provided cost complement can force lower le= vels of provision for a second level of other public goods. Those on the second level may be cost complementary to public goods on a third level, etc.  Simultaneous feedback will then fo= rce lower provision for the “original” under-provided goods. <= /o:p>

·         There may be “moral contagion” between under-provided public goods, even without cost complementarity.  The dissolution of commitments to reciprocity in one public good could make it more difficult to sustain such reciprocity for other public goods.  This dissolution of reciprocity can be motivated by a rupture with implicit social contracts, along Levi’s horizontal and/or vertical dimensions. 

4.  Variation Within (as Opp= osed to Between) Countries: Income v= ersus Uncorruption

        &= nbsp;       With the exception of the Reisen & Soto (2001) study, the papers cited above= are based on cross-sectional = data tests.  To test the emerging consensus on growth, Easterly (1999) constructed a large international panel data set on income growth and standard indicia of progress under developmen= t, including uncorruption, health, and education.   Testing for within-country effects, he found that development indicators such uncorruption, rule of la= w, schooling, and life expectancy did not improve with income.

If within-country effects of income growth are at odds with the cross-sectional evidence, as in Easterly (1999), there are arguments that the cross-section= al evidence remains more fundamental.  Since the time periods for the data sets used in this paper are medium-term – 10 years for the Easterly data, and 5 years for the Kaufmann data (2003)  – = they are unlikely to capture all of the socioeconomic changes wrought by sustain= ed income growth.  It may be that= any “corruption promoting” effects of growth are strictly short-term.  Easterly’s = growth “pessimism” would then be seen as a mild qualification to the m= ore orthodox pro-growth optimism: the negative backwash of growth a minor frict= ion on the virtuous longterm upward spiral of income growth.  These issues will not be resolved = by this paper, but only by longer data sets.

        &= nbsp;       Part 5 of the present paper builds upon Easterly’s findings. We begin with= his panel data set, to show how corruption interacts with other quality-of-life variables, such as health.  Te= sting for within country variation, we show that improvements in health and liter= acy can improve uncorruption, even after the negative effect of income upon uncorru= ption is taken into account.  The following Part 6 will address the Kaufmann (2003) data set.  Although income in this set appear= s to weakly promote uncorruption , the same anti-corruption effects of health and literacy will still  be shown.=

5. Estimates on Easterly Data:  Health Care and Corruption

            &n= bsp;   Easterly shows in his Tables 1 and 2 (we do not repeat his regressions) that:

1)       Higher income improves life expectancy between countries, but harms it within countries.  Higher income tends to improve inf= ant mortality, both between and within countries.

2)       Income growth improves uncorruption and rule of law between, but harms them within, countries. 

            &n= bsp;   In the present paper, we show that higher life expectancy improves both uncorr= uption and rule of law, both within and= between countries. Because income pushes health car= e up but uncorruption down, at least in the Easterly data set, the interesting question is their joint e= ffect on uncorruption.  Here we foll= ow the standard practice of the literature (Pritchett & Summers 1996), and mod= el income as more determinative of health care than vice versa. A more complete model would test for co-determination here, and our estimation procedure in= no way rules this out.

        &= nbsp;       We begin by applying instrumental variables to our health care variable of life expectancy.  The first step is to model the determination of life expectancy as dependent on both Income (measured here= as real per capita GDP) and other instrumental variables: length of paved highways, and cars per capita.  That ordinary least squares (OLS) regression is shown in Table 1.

(Please Insert Table 1 about here.)<= span style=3D'font-size:10.0pt;line-height:200%;layout-grid-mode:line'>

There is = likely multicollinearity or functional dependency between variables in Table 1: e.= g., between Cars per capita and Highway density.  If we cared about specific coeffic= ient estimates, this would be problematic.  However, here we are interested only in getting a “fitted̶= 1; form of life expectancy; multicollinearity is of no concern.

        &= nbsp;       Next, we show the estimates of the between-country and within-country effects of = life expectancy upon uncorruption.  The between- and within-country equations are shown here as (1a) and (1b), rela= ting to the Table 2, columns a and b, respectively:

      &n= bsp;         Uncorruptioni  =3D Constant  + b(Life Expectancyi )<= span style=3D'mso-spacerun:yes'>   + (Incomei, Other Variablesi )β  +  ε i<= span style=3D'mso-tab-count:1'>        &= nbsp;           &nbs= p;         (1a)<= /span>

      &n= bsp;         Uncorruptionit =3D Constanti + b(Life Expectancyit )  + (Incomeit, Other Vari= ablesit)β +  ε i        &= nbsp;           &nbs= p;          (1b)

where i =3D (1, 2, …, C),= the number of individual countries, and t =3D (1, 2, ..T), the number of time p= eriods observed.  In the portion of t= he Easterly data used below, the number of countries is 103, and the number of times periods is at most 2 (1980 and 1990).  (The Kaufmann data, to be examined= in part 6 of this paper, are more recent.)

(Please Inse= rt Table 2 about here.)

 

        &= nbsp;       The number of included exogenous variables in the structural form regressions of Table 2 is smaller than the total number of exogenous variables in the redu= ced form regression of Table 1, so the necessary conditions for the identificat= ion of the coefficients estimated in Table 2 are met.  These conditions are met by all the structural form estimates in this paper.

        &= nbsp;       In the within-country estimates of Table 2b, the coefficient on Life Expectanc= y is positive and significant, while that on the Income term is short of statist= ical significance, and has turned negative. Comparing this with the between-coun= try estimates Table 2a, note that the positive effect of Life Expectancy upon U= ncorruption is seriously understated by these estimates.  The coefficient on Life Expectancy= in Table 2b is almost ten times as high as it is in 2a, as well as now being significant.  The coefficient on per-capita Inco= me in Table 2b is not only of the opposite sign from that in Table 2a, it is also= now much more significant.  The Ha= usman test shows that the null hypothesis of a Random Effects (RE) specification = for the intercept showing between-country variation – rather than the Fix= ed Effects (FE) estimates shown here – is rejected at a level of less th= an two percent.

In the within-country estimates of the effect of income alone on uncorruption (not shown here), income showed a negative correlation with uncorruption –= as has been well-stressed by Easterly (1999).=   The present estimates strengthen Easterly’s basic conclusion, = by adding the “omitted variable” of life expectancy.   Because life expectancy is positively correlated with uncorruption, while income is negatively correla= ted with both uncorruption and life expectancy, a biasing of the results can be expected to hold when life expectancy is omitted.  Income’s negative effect on uncorruption is understated when i= t is “flying solo” in Table 2a, giving it some undeserved credit for= the positive effect of life expecta= ncy, as seen in Table 2b. 

Along wit= h Uncorruption, Easterly’s data set codes for the “Rule of Law,” both assembled by the IRIS Country Risk Database (University of Maryl= and).  It might be thought that Uncorrupt= ion and Rule of Law would be virtually the same, and the two series are indeed highly correlated, but there are interesting exceptions that make the difference worth studying.[4]  The equations on Rule of Law have functional forms similar to (1a), (1b); equations (2a) and (2b) correspond = to regressions in Tables 3a and 3b below:

      &n= bsp;         Rule Lawi  =3D Constant<= span style=3D'mso-spacerun:yes'>  + b(Life Expectancyi )<= span style=3D'mso-spacerun:yes'>   + (Incomei, Other Variablesi )β  +  ε i<= span style=3D'mso-tab-count:1'>        &= nbsp;           &nbs= p;                (2a)<= /span>

      &n= bsp;         Rule Lawit =3D Constanti + b(Life Expectancyit = )  + (Incomeit, Other Vari= ablesit)β +  ε i        &= nbsp;           &nbs= p;                 (2b)<= /span>

 

(Please Inse= rt Table 3 about here.)

&nbs= p;

        &= nbsp;       In regression 3(b) above, as compared to the earlier 2(b), the coefficients showing the negative effect of Income and the positive effect of Life Expectancy, both on Rule of Law, are of similar magnitudes, although more significant than they were on Uncorruption.  This is reflected in the elasticity calculations to follow.

        &= nbsp;       Interpreting these results in terms of elasticities, we can take the regressions in Tabl= es 2b and 3b, and evaluate them at the sample means for Uncorruption, Rule of = Law, Life Expectancy and Per Capita Income (3.301, 2.893= , 63.08 years, and $4,857, respectively): <= o:p>

 (Please Insert Table 4 about here.)

 

        &= nbsp;       These results are striking:  a 1% in= crease in Life Expectancy yields almost a 3% improvement in Uncorruption, and near= ly a 2% improvement in the Rule of Law.  These elasticities are approximately 9 times and 3.5 times as great, respectively, as those from Income – and in the opposite direction.  From these estimates, it is clear = that improving Life Expectancy is a reasonable way to fight corruption, much bet= ter than increasing Income itself.[5]  Bear in mind that the changes in L= ife Expectancy were measured over a short time horizon – about a decade in these Easterly data. 

  &nb= sp;            = We have m= odeled Uncorruption and Health as jointly determinative, so the next task is to estimate the effect of Uncorruption upon Health.  (As previously noted, there is a l= arge body of research pointing to a significant and positive correlation, both between and within countries.)  We estimating a reduced form of Uncorruption in Table 5.  This fitted form allows us to esti= mate the effects of Uncorruption on Life Expectancy in Table 6.

(Please Inse= rt Tables 5 and 6 about here.)

 

Rule of L= aw variable was also estimated in a fitted form, but was not shown to be a significant determinant of Life Expectancy.  Thus we find that Life Expectancy = has a positive influence on both Uncorruption and Rule of Law.  Transparency appears, from the regression in Table 6b, to have a positive influence back upon Life Expecta= ncy.  

Rule of L= aw can also be shown to have a significant effect upon Life Expectancy. (Not surprisingly, Rule of Law and Uncorruption are mutual supporting – re= gressions not shown).  The fitted form o= f Rule of Law is estimated in Table 7. 

 

(Please Insert Table 7 about here.)

<= o:p> 

The within-country form of the regression, below in Table 8b, makes it clear that Rule of Law significantly <= /p>

 

promotes Life Expectancy.

 

(Please Inse= rt Table 8 about here.)

 

These co-determinations are summarized in the following schematic figure:

 

(Please Inse= rt Figure 2 about here.)

 

        &= nbsp;       Using the same Easterly data, we also examined the relationship between Uncorrupt= ion, Rule of Law, and Average Years of Schooling.   We did not find statistically significant results here. The results were suggestive, however, in that Yea= rs of Schooling (in a fitted reduced form) was positively associated with Rule= of Law, but significant only at the 15 percent level (thus the ‘broken arrow’ shown in Figure 3).  We next use a more recent data set to show the same association as highly significant, using a variable on educational outcomes, literacy. 

(Please Inse= rt Figure 3 about here.)

 

6. Estimates on Kaufmann Data:  Health Care and Corruption

        &= nbsp;       A more recent Governance and Regional Capacity data set compiled by Daniel Kaufmann (2003, 2003a, 2002; see also 2004), principal researcher behind the development of these new World Bank indexes on governance, shows a mixed pattern of correlations between Income growth on the one hand, and Uncorrup= tion or Rule of Law on the other.  = The data cover the late 1990s and early 2000s, the time period during which many initiatives criminalizing international bribery were adopted.  (It may be that the effect of such= laws follows a period of delay and therefore is not yet reflected, or fully reflected, in the data.)  Sepa= rating the world into the (not perfectly exclusive) groupings of Industrial East A= sia, Developing East Asia, OECD, Middle East, Eastern Europe, Latin America, Sou= th Asia, Sub-Saharan Africa, and the Former Soviet Union, Kaufmann finds a more mixed pattern of correlations between uncorruption and growth –positi= ve correlations in most of the world (growth falling along with uncorruption in much of Latin America and the Former Soviet Union), but negative correlatio= ns in Asia (rising growth and falling uncorruption).

        &= nbsp;       As will be seen in our analysis of the Kaufmann data, the correlations between Income and Uncorruption or Rule of Law are often positive or non-significan= t, in contrast to the negative correlations in the Easterly data.  (In the Conclusion, we discuss some possible reasons for this divergence.)

&nbs= p;

6.b. Kaufmann Data, Life Expectancy, Uncorrupti= on, and Rule of Law.

        &= nbsp;       The Kaufmann data, however, do show a highly similar pattern of correlation between Life Expectancy, Uncorruption, and Rule of Law.  The following Figure 4 (Kaufmann data) and Figure 2 (Easterly data) are identical – both show the same pattern of co-determinations.

 

(Please Inse= rt Figure 4 about here.)

&nbs= p;

        &= nbsp;       We begin by estimating a fitted form of Life Expectancy, using World Bank data over the same period as the Kaufmann (2003) data (for the years 1996, 1998,= and 2000), as shown in Table 9.

(Please Insert Table 9 about here.)<= span style=3D'font-size:10.0pt;line-height:200%;layout-grid-mode:line'>

        &= nbsp;       Note that there is certain to be widespread multi-collinearity in the above regression, as for example between the closely related terms on Income (measured here as per-capita GDP, adjusted by Purchasing Power Parity) or between Public Health Expenditure and Population Growth.  Our purpose here is to derive a fi= tted form of Life Expectancy, so collinearity is of no concern.  Other specifications of Life Expec= tancy were tried, but the above was found to give the best results in the estimat= ions to follow.

        &= nbsp;       Next, we use the fitted form of Life Expectancy to do panel data estimations on U= ncorruption and Rule of Law, defined similarly as in the Easterly data set, with differ= ent normalizations of coding.

(Please Insert Table 10 about here.)=

        &= nbsp;       From the Within Country regression on Uncorruption (Table 10b), one can see that= the income coefficient is not significant, while that on Life Expectancy is of = full significance.  The P-Value on = the Durbin-Watson statistic is in the approximate form, and given as a range.  The probability that the null-hypo= thesis (that there is no serial-correlation) is correct is thus at least 8.6 perce= nt, and at most 45.2 percent, and therefore should not be rejected. =

The regre= ssion on Rule of Law will also show the correlation with Life Expectancy as highly significant:

(Please Insert Table 11 about here.)=

        &= nbsp;       In the Within Country estimation (Table 11b), the income term is almost at statistical significance, as it was in the previous regression on Uncorrupt= ion.  (In contrast to the Easterly data,= the effect of Income upon Rule of Law here appears to be weakly positive.) The = Life Expectancy term, however, is extremely significant.  

        &= nbsp;       We now turn to reciprocal causality – the determination of Uncorruption = and Rule of Law upon Life Expectancy.  We begin, as before, by estimating the fitted form of Uncorruption, = as shown in Table 12.  In Table 12, we use the fi= tted form of this variable to estimate the effect of Uncorruption upon Life Expectancy. =

(Please Insert Table 12 about here.)=

From just= these Between Country estimates, it is clearly it would appear that the effect of= Uncorruption upon Life Expectancy is insignificant.&nbs= p; The Within Country estimates, however, show a much stronger effect:<= o:p>

        &= nbsp;       Table 12b shows Income negatively but not significantly associated with Life Expectancy on a Within Country basis.  This contrasts with the significantly negative income effects in the earlier Easterly Within-Country estimates. Uncorruption is here significant= ly and positively associated with Life Expectancy on a Within-Country basis.  This high t-statistic is somewhat surprising, given the short time frame of the analysis, which is over just three years (1996 through 1998) for most countries, and at most five years (1996 through 2000) for any country. 

We now tu= rn to estimating the reciprocal effect of Rule of Law upon Life Expectancy.  We use the exogenous variables to generate a fitted form of Rule of Law, as shown in Table 13.

(Please I= nsert Tables 13 and 14 about here.)

        &= nbsp;       We will not develop a fully simultaneous model of the joint determination of t= he two governance variables (Uncorruption and Rule of Law) and the public good= s of Health and Education, but certain basic correlations can be noted.  Both Rule of Law and Uncorruption = have just been shown to promote Life Expectancy, as previous literature confirms.  Another interesting question concerns their joint influence.&n= bsp; This turns out to be difficult to estimate, because these two govern= ance variables are highly correlated, and because of serial correlation.  However, using an auto-regressive = form, it can be shown that, while both governance variables are positively correl= ated with Life Expectancy, the Rule of Law is much more so, and more significant= ly, as the Within-Country regressions in Tables 12b and 14b suggest. 

        &= nbsp;       We conclude this section with a comparison of the elasticity effects of Income= and Life Expectancy upon our governance variables.  Table 15 (Kaufmann data) is struct= ured the same as Table 4 (Easterly data).  Table 4 shows Income as having negative coefficients, but is similar= in that Life Expectancy has far greater absolute elasticities.  In Table 15, the relative elastici= ties from Life Expectancy are greater still – an order of magnitude higher than those from Income – and their coefficients are much more significant.

(Please I= nsert Table 15 about here.)

&nbs= p;

        &= nbsp;       Next we turn to the Kaufmann data (2003) on Illiteracy, a measure of educational effectiveness, and their influence upon our governance variables, Uncorrupt= ion and Rule of Law.  We begin by estimating the fitted form of Illiteracy, as shown in Table 16.  Using this fitted form, we estimat= e the effect of Illiteracy upon Uncorruption, as shown in Table 17.

(Please I= nsert Tables 16 and 17 about here.)

 

        &= nbsp;       Here we see that, unlike in the regressions on the Easterly data, the fitted for= m of Illiteracy in the Fixed Effects estimates is highly significant.  The negative sign on the Population Density term (or, in an alternative specification, on an Urban Population t= erm) might be explained by the increased economic value of literacy in an urban environment.  Interestingly, t= he positive sign on Female Labor Force suggests that the diminished child care aspect of female labor may outweigh the increased economic integration of women. 

The posit= ive effect from the proportion of Female Population in Table 17b was unexpected, but appeared to be robust to alternative specifications.  (Nor were the other coefficients m= uch changed if Female Population was dropped.)=   It is worth asking what links might exist between the Rule of Law and the personal security of women, female health, female infanticide, or other factors that could influence the overall sex ratio.  We have no model to offer, however= , and will not pursue the question here.

        &= nbsp;       We turn next to an estimation of the effects of Illiteracy upon Rule of Law, in Table 18. 

(Please I= nsert Tables 18 and 19 about here.)

        &= nbsp;       The negative correlation between Illiteracy and both Uncorruption and Rule of L= aw is schematized in Figure 5 below.  Compare this pattern to the schematic of a similar relationship in Figure 3, based on the Easterly data, which showed Years of Schooling as positively correlated with Rule of Law – although not quite at the le= vel of statistical significance – and failed to show any effect upon Unco= rruption.  In the Kaufmann data, by contrast, Illiteracy is negatively correlated (because it is a bad) with both Uncorru= ption and Rule of Law.   <= /o:p>

        &= nbsp;       Another interesting similarly with the Easterly data is that it was not possible to show two-way arrows of causation between Uncorruption or Rule of Law, on the one hand, and Illiteracy, on the other.&nb= sp; Several reasonable specifications were attempted, using Illiteracy as the dependent variable and the fitted forms of Uncorruption or Rule of Law = as one of several independent variables.  This is not to claim tha= t no reciprocal evidence exists, however; it must be borne in mind that the time series here is short, at most five years.&= nbsp;

        &= nbsp;       As in the previous Tables 4 and 15, we next calculate the elasticities of Inco= me and the public good upon our two governance variables, as shown in Table 19= .

&nbs= p;

  &nb= sp;            = The very = high elasticities on Uncorruption are probably an artifact of the mean value of = Uncorruption over the sample being near zero.  More important here are the relative elasticities.   Here we see, in comparison t= o the previous Tables 4 and 15, that elasticities on the public good term (here Illiteracy, a public bad) are of similar magnitude to those on Income.  Thus, the question of the most eff= ective way to improve governance – reducing Illiteracy or raising Income = 211; will depend much more on the relative costs of such programs.

        &= nbsp;       The effect of Illiteracy upon Rule of Law in Table 17b has a coefficient of sim= ilar size but much greater significance, as compared to its effect upon Uncorrup= tion in Tables 16b and 16d (in the Within-Country estimates).  (This is shown in Figure 5 below w= ith a bold arrow between Illiteracy and= Rule of Law.)  This is qualitatively similar to the greater relative significance of the influence on Rule of Law found with the earlier Easterly data, as was shown in Figure 3. 

(Please Insert Table 20 about here.)

(Please Insert Figure 5 about here.)

4.  Summary and Conclusions

On the ba= sis of this study of two rather different data sets, taken together with the previously published results of other researchers, we conclude:<= /span>

(a)     &nbs= p;     that criminalizing bribery, even augmented by other changes in the law, such as accounting and disclosure requirements, should be comgined with a policy th= at includes improvement of the provision of appropriate public goods, and=

(b)     &nbs= p;     that an appropriate direction for additional responses is suggested by the evide= nce that Uncorruption and the Rule of Law are public goods that are complementa= ry with more traditional public goods of Health Care and Education.=

This latt= er conclusion is supported by within-country estimates, based on extensive pan= el data, demonstrating not only correlations, but a significant determining influence of Life Expe= ctancy and Illiteracy upon Uncorruption and Rule of Law.  This result does not contradict, b= ut is in addition to, the reverse causality established by earlier literature, wh= ich is further supported in our own estimations on Uncorruption’s reciprocal influence upon and from Life Expectancy (see Figures 2 and 4 above).

        &= nbsp;       Thus, there is a positive determining influence for health care and education upon corruption.  Yet, this is not = the same as proving a direct causal link:  both improvements might be caused by some third term.  How might better public health imp= rove Uncorruption and the Rule of Law?  There is evidence that good health is highly associated with strong community and fa= mily networks (Castles 1994, Cox et al. 1997).  Health is a good most = people prize highly, so that those who have the wealth to do so invest much leisure time and disposable income in promoting it.  The same social networks that prom= ote good health may also promote greater long-term relationships of trust, whic= h is the social basis of Uncorruption and the Rule of Law.

        &= nbsp;       Mutual aid delivered through community health care may also allow developing count= ries to leverage traditional forms of social capital used to promote honesty and= uncorruption.  The example of Grameen Bank has be= en widely noted, and is built upon the informal ROSCA networks found throughout the developing world (Besley et al.= 1993). Literacy and numeracy can only aid such leverage. 

        &= nbsp;       The findi= ngs of this paper would support raising the level of basic public goods like educa= tion and health, but no direct evidence as to the best means of doing so – whether through public, private, or voluntary/community provision:

  • If the correlation between uncorruption and pub= lic good provision is due to their both being provided by traditional community networks, then the strengthening of those networks may be ca= lled for.  Thus, traditional v= alues and networks in developing countries can be seen, --- instead of roadblocks to uncorruption and efficient public good provision -- as t= he best foundation for both (Besley et al. 1993, De Soto 2000, Schroth & Sharma 2003).
  • If, on the other hand, the uncorruption-public = good link is due to basic cost complementarities between them, then this may strengthen the case for further large-scale provision with decreasing-average-costs – whether from the private or the publi= c-sector.
  • Finally, of course, it is likely that there are complex interactions between both sorts of causation – from the voluntary provision side of uncorruption and related public goods, and from large-scale private and public-sector provision.

&nbs= p;

A program= of both (1) strengthening social network provisions, but also (2) leveraging t= hem through the public sector and the private market -- becomes even more important in = light of the controversy and contradictory evidence on whether income growth enco= urages or discourages corruption (Easterly 1999, Kaufmann 2003).  Capitalist modernization has often degraded traditional values and networks.&= nbsp; To the extent it has been so biased, it may also frustrate the establishment of more transparent institutions, and thus hinder its own mat= uration.

A closing thought on why social networks should not be ignored: modern capitalism in Western Europe and in Japan was not built upon the simple destruction of traditional social networks.   Rather, what made these soci= eties so successful was their ability to mobilize these traditional forms and transform them toward capitalism (Polanyi 1962, Stodder 1995, Rosser & Rosser 1999).  Polanyi, a stud= ent of Weber, stressed that capitalist social relations in Europe rest upon an underpinning of solid “bou= rgeois values” (in the non-Marxist and approving sense of that phrase= ), built up over centuries of medieval trade and town life.   Without such a broad social = basis, the institutions of a well-functioning capitalism are impossible.  Development programs must not only punish corruption; they must also nurture the social basis of uncorruption.=   That social basis, traditional nor= ms of honesty and reciprocity, also support the vast informal sector.<= /span>

            &n= bsp;   Improving the provision of public goods does not appear to be inconsistent in any way with the better publicized anti-corruption strategies and seems at least li= kely to improve their prospects for success.&nb= sp; Perhaps the most important difference, however, is cost:  to put it cynically, it costs very little to amend the penal code, particularly if no additional resources are allocated to investigation and prosecution.  However, unlike increased prosecut= ion, improved education and public health are good in themselves, whether or not corruption is reduced, so that our results may be interpreted as adding additional weight to arguments for public policies favoring allocation of resources to these areas.

 

APPENDIX: Data Sources

The Easterly Data are available at the World Bank working pa= per website, www.worldbank.org/r= esearch/peg/,  Working Paper Number 17.=

The Kaufmann = Data on Governance (w= ww.worldbank.org/wbi/governance/) also uses World Bank World Develop= ment Indicators (www.worldbank.org/= data/wdi2004/index.htm).

 

REFERENCES

African= Union (AU 2003), African Union Convention on Preventing and Combating Corruption, opened for signature 11 July 2003.

Axelrod, Robert, & Michael Cohen (2000), Harnessing Complexity: Organizational Implications of a Scientific Frontier, New York: Free Press.

Azfar, = Omar, & Tugrul Gurgur (2004), “Does Corruption Affect Health and Educat= ion Outcomes in the Phil= ippines?” Working Paper, IRIS Center and Department of Economics, University of Maryl= and.

Barro, = Robert (1997), Determinants of Economic Gr= owth: A Cross-Country Empirical Study, Cambridge (MA): MIT Press.

Besley, Timothy, Stephen Coate & Glenn Loury (1993), “The Economics of Rotating Savings and Credit Associations,” American Economic Review, September 83(4):792–810.

Bhargava, A., L. Franzini & W. Narendanath= an (1982), “Serial Correlation and the Fixed Effects Model,” Review of Economic Studies, 49<= /b>:533-549.

Bialos, Jeffrey P. & Gregory Husisian (1997), The Foreign Corr= upt Practices Act:  Coping with Corruption in Transitional Economies, Dobbs Ferry, NY:  Oceana.

Burki, = Shahi & Guillermo Perry (1998), Beyon= d the Washington Consensus: Institutions Matter Viewp= oints,” Washington, DC:&n= bsp; World Bank.

Castles= , Ian (1994), How Australians Use Their T= ime, Australian Bureau of Statistics, Catalog No. 4153.0, Canberra: Australian Government Printing Service.

Charap, Joshua & Christian Harm (1999), “Institutionalized Corruption and= the Kleptocratic State,” IMF Working Paper WP/99/91, www.imf.org/external/­pubs/­ft/­wp/­1999/&= shy;wp9991.pdf.

Council= of Europe (1999), Criminal Law Convention on Corruption, 27 Jan. 1999, Europ. T.S. No.173.<= /p>

Cox, Do= nald, Emmanuel Jimenez & Wlodek = Okrasa (1997), “Family Safety Nets and Economic Transition: Worker Household= s in Poland,” Review of Income and Wealth, Ju= ne: 191-209.

De Soto, Hernando (2000), The Mystery of Cap= ital: Why Capitalism Triumphs in the West and Fails Everywhere Else, New York: Basic = Books.

Easterl= y, William (1999), “Life During Growth: A Compendium of Political, Social and Environmental Indicators of What Gets Better and What Gets Worse from L= ow to High Income,” Journal of Economic Growth, 4(3):239-276.

Easterl= y, William & Ross Levine (1997), “Africa’s Growth Tragedy: Policies and Ethnic Divisions,” Quarterly Journal = of Economics, 112(4):1203-50.=

Engle, = Robert F., & David F. Hendry (1983), “Exogeneity,” Econometrica, 51(2): 277-304.

Foreign Corrupt Practices Act of 1977 (FCPA 1977), Pub. L. 95-213, 91 Stat. 1494, codified as amended at 15 U.S.C. §§78m(b), (d)(1), (g)-(h), 78dd-1, 78dd-2, 78dd-3, 78ff.

Frank, Ro= bert H. & Tibor Scitovsky (1992), The Joyless Economy: The Psychology of Human Satisfaction, Oxford:

        &= nbsp;       Oxford University Press.      &= nbsp;           &nbs= p;    

Friedman, Milton (1982), Capitalism and Freedom,= Chicago: University of Chicago Press.

Gupta, Sanjeev, Hamid Davoodi & Erwin Tiongson (2000), “Corruption and t= he Provision of Health Care and Education Services,” IMF Working Paper, Fiscal Affairs Department,

www.i= mf.org/external/pubs/cat/longres.cfm?sk=3D3652.0  

Gupta, Sanjeev, Hamid Davoodi & Rosa Alonso-Terme (1998), “Does Corrupti= on Affect Income Inequality and Poverty?,” IMF Working Paper, Fiscal Aff= airs Department,

ww= w.imf.org/­external/pubs/cat/longres.cfm?sk=3D2629.0   

Heal, Geoffrey & P.S. DasGupta (1980), Economic Theory and Exhaustible Resources, Cambridge: Cambridge= University Press.

Hines, Jr., James R. (1995) “Forbidden Payment: Foreign Bribery and American Business After 1977,= ” NBER

            &n= bsp;   Working Paper 5266.

Kaufman= n, Daniel (2004), “Corruption, Governance and Security:  Challenges for the Rich Countries = and the World,” in World Bank, Global Competitiveness Report 2004/2005= , Washington, D.C.:=   World Bank, available at www.worldbank.org/wbi/governance/pubs/gcr20= 04.html.

_______________ (2003), Index on Global Governance and Regional Capac= ity, World Bank Institute (WBI): http://www.worldbank.org/= wbi/governance/.   

_______________ (2003a), “Rethinking Governance: Empirical Less= ons Challenge Orthodoxy,” World Bank Discussion Paper (March).=

Kaufmann, Daniel & Aart Kraay (2002), “Growth without Governance,” Economia, 3(= 1): 169–229, http://= www.worldbank.org/wbi/governance/pubs/growthgov.htm.    

Levi, Margaret (1988) Of Rule= and Revenue, Berkeley, California: University of California Press.

Lopez, Ramon, & Siddhartha Mitra (2000), “Corruption, Pollution, and the Kuznets Environment Curve,” Journal of Environmental Economics and Management 40:137-150.

Mauro, = Paolo (1995), “Corruption and Growth,” Quarterly Journal of Econom= ics, 110(3):681-712.

_______= ____ (1997), “The Effects of Corruption on Growth, Investment, and Governm= ent Expenditure:  A Cross-Country Analysis,” in Elliot, Kimberly Ann (1997), ed., Corruption and the Global Economy, Washington, D.C.= : Institute for International Economics.

Organiz= ation for Economic Cooperation and Development (OECD 1998), Convention on Combating Bribery of Foreign Public Officials in International Business Transactions, OECD/DAFFE/IME/BR(97)16/FINAL (18 Dec. 1997),= 37 ILM 1 (1998), S. Treaty Doc. No. 105-43.

Organiz= ation of American States (OAS 1996), Inter-= American Convention Against Corruption, OEA/Ser. K/XXXIV.1, CICOR/doc. 14/96 rev. 2 = (29 Mar. 1996), = 35 ILM 724 (1996).

Olson, = Mancur (2000), Power and Prosperity: Outgr= owing Communist and Capitalist Dictatorships, New York: Basic Books.

Pritchett, Lant, & Larry Su= mmers (1996), “Wealthier Is Healthier,” Journal of Human Resources= , 31(4):841-868.

Polanyi, Karl (1962), The Great Transformation, Boston: Beacon Pre= ss.

Reisen, Helmut, & Marcelo Soto (2001), “Which Types of Capital Inflows Fo= ster Country Growth?” Internationa= l Finance, 4(1):1-14.

Republic of South Africa (2004), Prevention and Combating of Corrupt Activities A= ct 12 of 2004.

Rose-Ac= kerman, Susan (1999), Corruption and Govern= ment: Causes, Consequences and Reform, Cambridge (UK): Cambridge University Press.

_______= ____________ (1997), “The Political Economy of Corruption,” in Kimberly, Ann Elliott, ed., Corruption and the Gl= obal Economy, Washington, D.C.:=   Institute for International Economics.

_______= ____________ (1978), Corruption: A Study in Poli= tical Economy, New York: Academic Press.

Rosser,= J. Barkley, Jr. & Marina Rosser (1999), “The New Traditional Economy: A New Perspective for Comparative Economics?” International Journal of Social Economics, 2= 6(6):763-778.

Sachs, Jeffrey D., & Andrew Warner (1997), “Fundamental Sou= rces of Long-run Growth,” American Economic Review, Papers and Proceedi= ngs, 87(2):184-88.

Scholz, John T. and Mark Lubell (1998) “Trust and Taxpaying: Testing the Heuristic Approach to Collective Action,” American Journal of Political Science<= /i>, 42:398—417.

Schroth, Peter W. (2005), “The African Union Convention on Preventing and Combating Corruption,” forthcoming in Journal of African Law, = 51:___.

_______= ________ (2003), “National and International Constitutional Law Aspects of Afr= ican Treaties and Laws Against Corruption,” Transnational Law and Contemporary Problems, 13:83-108.

_______= ________ (2002), “The U= nited States and the International Bribery Conventions,” American Journal of Comparative Law, 50 (supp.):593-622.

Schroth, Peter W., & Ana Da= niela Bostan (2004), “International Constitutional Law and Anti-Corruption Meas= ures in the European Union’s Accession Negotiations:  Romania in Comparative Perspective,” forthcoming in American Journal of Comparative Law, 52:___.

Schroth, Peter W. & Preeti Sharma (2003), “Transnational Law and Technology as Potential Forces Against Corruption in Africa,” Management Decision, 41:296-303.

Southern African Development Community (SADC, 2001), Protocol Against Corruption<= /i>, opened for signature 14 Aug. 2001, available at http://­www.sadc.int/­index.php?lang=3D­english&­path= =3D­legal/­protocols/­&page=3D­p_corruption.

Stodder, James (1995), “The Evolution of Complexity in primitive Exchange,” Journal of Compara= tive Economics, Vol. 20; “Theory,” Issue 1, February, pp. 1-31 a= nd “Empirical Tests,” Issue 2, April, pp. 190-210.

____________ (1993), “Ex-Communist Economics: the Logic of Capi= tal Reform,” New York Economic Review, 24(4).

United Nations (UN 2003), United Nations Convention Against Corruption, G.A. Res. = A/­RES/­58/4, 31 Oct. 2003, opened for signature 9 Dec. 2003.

Wei, Shang-Jin (1997), “Why is Corru= ption So Much More Taxing Than Tax? Arbitrariness Kills,” NBER Working Paper No. w6255.

_______= _____ (2000), “How Taxing Is Corruption on International Investors?” = Review of Economics & Statistics,<= /i> 82(1).

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

&n= bsp;

Figure 1:  Complementary Marginal Costs and Public Good Demands: Uncorruption and Health<= /span>

 

 

 

 

 

 

 

 

 

Figure 2.  Within Country Co-Determinations:

Life Expectancy, Uncorruption & Rule of Law= , Easterly Data

 

 

 

 

 

Figure 3.  Within Country Co-Determinations:

Schooling, Uncorruption & Rule of Law, E= asterly Data

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4. Within Country Co-Determinations:

Life Expectancy, Uncorruption & Rule of Law= , Kaufmann Data

 

 

 

 

Figure 5. Within Country Co-Determinations:

Illiterac= y, Uncorruption & Rule of Law, Kaufmann Data<= /p>

= &= nbsp;

= &= nbsp;

= &= nbsp;

= &= nbsp;

= &= nbsp;

= &= nbsp;

= &= nbsp;

= &= nbsp;

= &= nbsp;

= &= nbsp;

= &= nbsp;

= &= nbsp;

= TABLE 1.      Dependent variable: LIFE EXPECTANCY, Easterly Data Set (1980 and 1990);  

Observations = =3D 103, Ordinary Least Squares (OLS) Estimates

 

= Mean dependent variable<= o:p>

63.079

= R-Squared

0.802

= Standard Error Regressio= n

4.750

= Adjusted R-Squared<= /o:p>

0.782

 

Variable

Constant

Coefficient

38.104

T-statistic=

(9.719)

P-value=

[.000]

CALORIES Per Capita

4.62E-03

(1.575)

[.119]

PROTEIN Per Capita

-1.77E-02

(-0.254)

[.800]

RADIOS Per Capita

2.235

(0.933)

[.353]

TVS Per Capita

22.863

(2.546)

[.013]

SANITATION

0.125

(5.397)

[.000]

CO2  Per Capita

-0.644

(-2.375)

[.020]

HIGHWAYS

2.43E-02

(1.011)

[.315]

CARS Per Capita

-28.613

(-2.625)

[.010]

INCOME Per Capita

1.38E-03

(3.735)

[.000]

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

TABLE 2:   Dependent variable: UNCORRUPTION, Easterly<= /u> Data Set (1980 and 1990); O =3D 103, C =3D 67, T =3D 2, <= /p>

    P =3D 36; where O =3D # Observations, C =3D # Cou= ntries, T =3D Maximum # of Time Periods, an= d

    P =3D # of Paired observations (number of countries with both ti= me periods observed).

 

Variation E= xplained= :

= (a) Between-Country

= (b) Within-Country

= Regression Method

= OLS

= Fixed-Effects*

Mean dependent variable

3.3= 96

3.3= 01

Standard Error Regression

1.1= 11

0.7= 60

R-Squared<= span style=3D'font-size:10.0pt'>

0.5= 54

0.9= 43

Adjusted R-Squared

0.5= 26

0.8= 17

Obs= ervations

67<= o:p>

103=

Durbin-Watson Statistic<= /span><= span style=3D'font-size:10.0pt;layout-grid-mode:line'>[6]

 =

2.00

 =

 =

 =

Variables

Coefficients

Coefficients

 =

Constant=

1.033

--

=  

(0.= 466)

--

=  

[.6= 43]

--

= TVS Per Capita

4.023

-3.912

=  

(1.= 884)

(-1= .329)

=  

[.0= 64]

[.1= 93]

= CALORIES Per Capita

8.35E-05

-1.68E-03

=  

(0.= 144)

(-1= .295)

=  

[.8= 86]

[.2= 05]

= INCOME Per Capita

1.15E-04

-2.18E-04

=  

(1.= 67)

(-1= .557)

=  

[.1= 00]

[.1= 29]

= LIFE EXPECTANCY (fitted)

0.016

0.142

 

(0.= 312)

(2.= 329)

 

[.7= 56]

[.0= 26]

 

Note: (t-statistics) in parentheses, [p-values] in square brackets.

* Hausman test of H0: Rand= om vs. Fixed Effects:  CHISQ(4) =3D 16.793,  P-value =3D [.0021]

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

TABLE 3: D= ependent variable: RULE OF LAW, = Easterly Data Set (1980 and 1990);  O =3D 103,  C =3D 67, T =3D 2,

    P =3D 36; where O =3D # Observations, C =3D # Cou= ntries, T =3D Maximum # of Time Periods, an= d

    P =3D # of Paired observations (number of countries with both ti= me periods observed).

 

Variation Explained:

(a) Between-Country

(b) Within-

Country

Regression Method

OLS

Fixed-Effects*

Mean dependent variable<= o:p>

2.944

2.893

Standard Error Regressio= n

1.061

0.509

R-Squared

0.674

0.977

Adjusted R-Squared<= /o:p>

0.647

0.925

Durbin-Watson Statistic

 

2.00

Variabl= es 

Coeffic= ients

Coeffic= ients

 

Constant

1.804

--

 

(0.803)

--

 

[.425]

--

HIGHWAYS

4.33E-03

-3= .44E-03

 

(0.697)

(-0.416)

 

[.488]

[.680]

CARS Per Capita

5.410

4.= 750

 

(1.891)

(1.385)

 

[.063]

[.176]

TVS Per Capita=

3.191

-4= .514

 

(1.209)

(-1.643)

 

[.231]

[.111]

INCOME Per Capita

3.92E-05

-3= .96E-04

 

(0.453)

(-2.242)

 

[.652]

[.032]

LIFE EXPECTANCY (fitted)

-4.96E-03<= /p>

0.103

 

(-0.117)

(2.278)

 

[.907]

[.030]

 

(t-statistics) in parenthe= ses, [p-values] in square brackets.

* Hausman tes= t of H0: Random vs. Fixed Effects:  CHISQ(5) =3D 29.735,  P= -value =3D [.0000]

&n= bsp;

Table 4.              =    Income versus Life Expectancy (Easterly Data):

 &nb= sp;            =    Estimated Elasticities on UNCORRUPTION and RULE OF LAW

 

        Effect of:

Coefficient

Elasticity

       &nbs= p;      Income on Uncorruption   

-2.18 E-04&nb= sp; 0

-0.320

Life Expectancy on Uncorruption<= /p>

0.142 *

2.721

       &nbs= p;      Income on Rule of Law=    

-3.96E-04 *

-0.583

Life Expectancy on Rule of Law

0.103 *

1.969

 

0 - significant at the 15% = level, * - significant at the 5% level

 

 

 

 

 

 

 

 

 

 

 

TABLE 5. <= span style=3D'mso-tab-count:1'>        &= nbsp;   Dependent variable: UNCORRUPTION, Easterly Data Set (1980 and 1990);

   &n= bsp;            = ;            &n= bsp;   Observations =3D 103, Ordinary Least Squares (OLS) Estimates

 

Mean dependent variable

3.301

R-squared

0.598

Standard error regression

1.164

Adjusted R-squared

0.560

 

 

 

 

Variabl= e 

Coeffic= ient

T-statistic

P-value

 

Constant

2.384

2.445

[.016]

CALORIES DAILY Per Capita

-5.68E-04

-0.778

[.439]

PROTEIN DAILY Per Capita

6.40E-03

0.370

[.712]

RADIOS Per Capita

-0.627

-1.052

[.296]

TVS Per Capita=

4.430

1.984

[.050]

SANITATION=

1.06E-02

1.845

[.068]

CO2 Per Capita=

6.50E-02

0.964

[.338]

HIGHWAYS

8.13E-03

1.359

[.178]

CARS Per Capita

1.150

0.425

[.672]

INCOME Per Capita

3.69E-05

0.403

[.688]

 

TABLE 6: <= span style=3D'mso-tab-count:1'>        &= nbsp;  Dependent variable: LIFE EXPECTANCY OF POPULATION,

Easterly Data Set (1980 and 1990);= O =3D 103, C =3D 67, T =3D 2, P =3D 36.

 

Variation Explained:

(a) Between-Country

(b) Within-

Country

Regression Method

OLS

Fixed-Effects*

Mean dependent variable<= o:p>

62.895

63.079

Standard Error Regressio= n

  5.491

   0.824

R-Squared

0.733

0.995

Adjusted R-Squared<= /o:p>

0.711

0.993

Durbin-Watson Statistic[7]

 

2.00

 

 

 =

Variabl= es 

Coeffic= ients

Coeffic= ients

 

Constant

31.371

--

 

(6.683)

--

 

[.000]

--

HIGHWAYS

-0.0205

-3.44E-03

 

(-0.608)

(-0.416)

 

[.546]

[.680]

CALORIES Daily Per Capita

0.00718

4.750

 

(3.096)

(1.385)

 

[.003]

[.176]

RADIOS Per Capita

0.825

-4.514

 

(0.273)

(-1.643)

 

[.786]

[.111]

INCOME Per Capita

0.000149

-3.96E-04

 

(0.371)

(-2.242)

 

[.712]

[.032]

RULE OF LAW (fitted)<= /span>

3.66

0.103

 

(2.214)

(2.278)

 

[.031]

[.030]

(t-statistics)= in parentheses, [p-values] in square brackets.

* Hausman test= of H0: RE vs. FE:  CHISQ(5) =3D 38.109,  P-value =3D [.0000]

TABLE 7.  Dependent variable: RULE OF LAW, Easterly Data Set (1980 and 1= 990)

 

 

Observations =3D 103, Ordinary Least Squares (OLS) Estimates

<= o:p> 

 

Mean dependent variable

2.893

R-squared<= o:p>

0.710

 

 

Standard error regression

1.048

Adjusted R-squared

0.682

 

 

 

Variable 

 

Coefficient

 

T-statistic

 

P-value

 

Constant

0.923

1.051

[.296]

CALORIES DAILY Per Capita

1.01E-03

1.541

[.127]

PROTEIN DAILY Per Capita

-0.031

-2.001

[.048]

RADIOS Per Capita

-0.874

-1.629

[.107]

TVS Per Capita

3.519

1.751

[.083]

SANITATION

8.18E-04

0.158

[.874]

CO2 Per Capita

0.085

1.392

[.167]

HIGHWAYS

2.71E-03

0.504

[.616]

CARS Per Capita

5.416

2.221

[.029]

INCOME Per Capita

3.25E-05

0.394

[.695]

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

TABLE 8.         Dependent variable: LIFE EXPECTANCY OF POPULATION, 

Easterly Data Set (1980 and 1990);= O =3D 103, C =3D 67, T =3D 2, P =3D 36

 

Variation Explained:

(a) Between-Country

(b) Within-

Country

Regression Method

OLS

Fixed-Effects*

Mean dependent variable<= o:p>

62.895

63.079

Standard Error Regressio= n

  5.491

   0.836

R-Squared

0.668

0.998

Adjusted R-Squared<= /o:p>

0.663

0.993

Durbin-Watson Statistic[8]

 

2.00

 

Variabl= es 

 

Coeffic= ients

 

Coeffic= ients

Constant

3.318

–

 

12.762

–

 

[.000]

–

 

HIGHWAYS

3.57E-02

-0.028

 

(0.23)

(-2.067)

 

[.819]

[.047]

 

 

 

PROTEIN Daily Per Capita

5.73E-02

0.139

 

(2.818

(2.939)

 

[.007]

[.006]

 

RADIOS Per Capita

3.219

5.265

 

(0.466

(6.038)

 

[.643]

[.000]

 

INCOME Per Capita

4.71E-04

-7.51E-04

 

(0.741

(-3.487)

 

[.462]

[.001]

 

RULE OF LAW (fitted)<= /span>

1.496

0.883

 

(1.491

(1.728)

 

[.141]

[.094

 

(t-statistics)= in parentheses, [p-values] in square brackets.

* Hausman test of H0: RE vs. FE:&nbs= p; CHISQ(5) =3D 41.534,  P= -value =3D [.0000]

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 9. D= ependent variable: LIFE EXPECTANCY OF POPULATION; WDI Data and

   Kaufmann Data Set (19= 96, 1998, and 2000), O =3D 104, C =3D 57,

where  O =3D # Observations,  = C =3D # Countries.

 

Mean of dependent variable

73.236

R-squared

0.878

Standard Error Regression

2.105

Adjusted R-squared

0.866

 

 

 

 

Variable

Coefficient=

t-statistic=

P-value=

Constant

81.172

5.262

[.000]

DEATHS

-1.148

-10.565

[.000]

INCOME (PPP)

-7.68E-05

-1.311

[.193]

PUBLIC HEALTH EXPENDITURE

0.237

1.335

[.185]

CARS

5.76E-03

1.853

[.067]

POPULATION GROWTH

-1.723

-5.313

[.000]

FEMALE POPULATION %

-0.166

-0.523

[.602]

PHONES

1.26E-02

3.930

[.000]

TVS

4.43E-03

2.017

[.047]

URBAN POPULATION %

5.96E-02

3.287

[.001]

 

 

TABLE 10:= Dependent variable: UNCORRUPTION; WDI Data and= Kaufmann Data Set (1996,

       1998= , and 2000), O =3D 104, C =3D 57, T =3D 3; where O=3D#Observations, C=3D#Countries,

         T=3DMax # of Time Periods.

 

Variation Explained:

(a) Between-Country

(b) Within-

Country

Regression Method

OLS

Fixed-Effects*

Mean dependent variable<= o:p>

0.468

0.570

Standard Error Regressio= n

0.890

0.251

R-Squared

0.489

0.980

Adjusted R-Squared<= /o:p>

0.483

0.952

Durbin-Watson Statistic

 

1.888<= /p>

Variabl= es 

Coeffic= ients

Coeffic= ients

Constant

0.79

--

 

(0.348)

--

 

[.730]

--

 

PU= BLIC HEALTH EXPENDITURE

-3.27E-02

7.49E-02

 

(-0.386)

(0.843)

 

[.701]

[.404]

 

UR= BAN POPULATION %

7.89E-03

-0.102

 

(0.785)

(-2.318)

 

[.436]

[.025]

 

IN= COME (PPP)

7.48E-05

2.65E-05

 

(3.852)

(1.527)

 

[.000]

[.134]

 

LIFE EXPECTANCY (fitted)

-2.35E-02

8.25E-02

 

(-0.616)

(2.371)

 

[.540]

[.022]

 

(t-statistics) in parenthe= ses, [p-values] in square brackets.

Hausman test of H0: RE vs. FE:  CHISQ(4) =3D 11.935,  P-value =3D [.0178] 

 

TABLE 11 Dependent variable: RULE OF LAW; WDI Data and Kaufmann Data Set =

        (1996, 1998, and 2000), O =3D 104, C =3D 57, T =3D 3; where O=3D#Observation= s, C=3D#Countries, <= /p>

         T=3DMax # of Time Periods.

 

 

Variation Explained:

(a) Between-Country

(b) Within-

Country

Regression Method

OLS

Fixed-Effects*

Mean dependent variable<= o:p>

0.496

0.580

Standard Error Regressio= n

0.785

0.125

R-Squared

0.490

0.994

Adjusted R-Squared<= /o:p>

0.440

0.986

Durbin-Watson Statistic

 

1.975

Variabl= es 

Coeffic= ients

Coeffic= ients

Constant

-1.691

--

 

(-0.81)

--

 

[.422]

--

TV= S Per Capita

-3.14E-04

-1.71E-03

 

(-0.342)

(-2.799)

 

[.734]

[.008]

CA= RS Per Capita

1.41E-04

1.64E-03

 

(0.104)

(2.037)

 

[.918]

[.048]

UR= BAN POPULATION %

8.38E-03

-4.62E-02

 

(0.921)

(-1.888)

 

[.361]

[.066]

IN= COME Per Capita (PPP)

5.85E-05

1.63E-05

 

(2.539)

(1.665)

 

[.014]

[.103]

LI= FE EXPECTANCY (fitted)

1.26E-02

4.36E-02

 

(0.36)

(2.408)

 

[.720]

[.021]

 

(t-statistics) in parenthe= ses, [p-values] in square brackets.

* Hausman test of H0: RE v= s. FE: CHISQ(5) =3D 11.381,  P-value = =3D [.0443]

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

TABLE 12<= /span>. D= ependent variable: LIFE EXPECTANCY OF POPULATION;

      =   WD= I Data and Kaufmann Data Set (1996, 1998, and 2000), O =3D 104, C =3D 57, T= =3D 3;

         where O=3D#Observations, C=3D#Countries,  T=3DMax # of Time Periods.

 

 

Variation Explained:

(a) Between-Country

(b) Within-

Country

Regression Method

OLS

Fixed-Effects*

Mean dependent variable<= o:p>

72.722

73.236

Standard Error Regressio= n

4.746

0.472

R-Squared

0.523

0.997

Adjusted R-Squared<= /o:p>

0.486

0.993

Durbin-Watson Statistic

 

1.837

 

 

Variabl= es 

Coeffic= ients

Coeffic= ients

Constant

83.264

--

 

(1.954)

--