BarroEasterlyLevine

B= arro Economic Growth in a Cross Se= ction of Countries. QJE 1991.

98 Countries

1= 960 to 1985.

Dependent variable (t= o start with) is growth rate of real GDP per capita.

S= olow model- unconditional convergence.

y=3Df(k), where y is income per capita, and k is capital per capita (per capita and p= er worker are used interchangeably).

By diminishing margin= al returns to capital, capital should flow to where capital is scarce.

Where capital is scar= ce, starting period income should be lower.&nb= sp;

Hence, “…= a country’s per capita growth rate tends to be inversely related to its starting level = of income per person.” (p.= 407).

This hypothesis ̶= 0;seems to be inconsistent with the cross-country evidence, which indicates that per capita growth rates have little correlation with the starting level of per capita product.” (p. 407-8).

See figure one.<= /o:p>

S= ummers and Heston / Penn World Tables.

http://www.pwt.econ.upenn.edu/<= o:p>

G= rowing role in theory for human capital. Romer = model – human capital as input to research, that= leads to technological progress.

Barro is using cross = country regressions to see if the ‘convergence’ story might hold true conditional on human capital levels.

Regress: growth rate 1960-1985 (average ann= ual)

On:=

GDP in 1960 and its s= quare

GDP in 1970

Secondary education s= tarting period (50 and or 60)

Primary education sta= rting period (50 and or 60)

In a variant student to teacher ratios.

Government consumptio= n as a share of GDP

Revolutions and coups= per year on average

A= ssassinations per million per year.

Over / Under valuation of the currency

Socialist dummy<= /o:p>

Mixed economy dummy

Africa Dummy

Latin America Dummy.

F= inds conditional convergence. “…an increase in per c= apita real GDP by $1,000 lowers the real per capita growth rate …by 0.75 percentage points per year.” P. 414

See figure II on page= 415.

The results suggest t= hat starting period education variables have an impact on ensuing economic grow= th, and this is about the same range as seen for the initial per capita GDP res= ult (around 5%).

See figure III on pag= e 417.

Enrollment is quantit= y of education. Barro tries a qual= ity of education measure, by using the student to teacher ratio (the higher, the l= ower the quality).

Next section consider= s when fertility is an endogenous variable. Net fertility rate as a function of regressors. See figures V and VI.

Private investment as= a ratio of GDP is given the same kind of treatment. See VII and VI= II.

Policy variables: government consumption as a share = of GDP, higher the share the lower the growth. This is not the starting period, b= ut the average over the same time period for which the growth occurred.

“…poor co= untries tend to catch up with rich countries if the poor countries have high human capital per person…but not otherwise” (p. 437).=

What does the high en= rollment rate measure capture?

Does this mean that investment in education will be needed as a necessary condition for growth = to occur?

T= he policy measures lead to the conclusion that policy may influence growth. <= o:p>

C= ausality issues.

E= ndogeneity issues.

Does higher growth ma= ke fewer assassinations happen?

Do fewer assassinatio= ns make growth faster?

D= iscussion of simultaneous equations.

What are the key assu= mptions in an OLS regression?

1) _<=
![endif]>

2) _<=
![endif]>

3) _<=
![endif]> are indepen= dent

4) _<=
![endif]> are indepen= dent, _<=
![endif]> =

Since cov(x,u)=3DE(xu)-E(x)E(U= ), this zero covariance gives us E(xu)=3D0, given (1).<= o:p>

5) We often also assu= me _<=
![endif]>

Present two equation = system.

Illustrate issue.

So we want an instrum= ental variable. <= /p>

If cov(u,x) is not equal to zero, and cov(z,x) is not equal to zero, we can use z as an instrument for x if cov(u,z)=3D0.

We can use E[zu] as a moment condition / regress x on z and use the resulting estimate of ‘x-hat’ in a second stage (2 stage least squares).

Tropics, Germs, and Crops: Easterly and Levine, J= ME 50:3:39.

G= eography / Environment hypothesis.=

Environment directly influences the quality of land, labor and production technologies. (p. 5)

S= oils. <= span class=3DGramE>Diseases. Landlocked. Distance from larger markets. Impact on both the input bundle available for the production function and the production function itself.

1)&n= bsp; Soils have = low fertility and are fragile.

2)&n= bsp; High preval= ence of crop pests and parasites

3)&n= bsp; High plant respiration and lower rate of net photosynthesis

4)&n= bsp; High evapor= ation and unstable water supplies

5)&n= bsp; Lack of sui= table seasons for temperate crops

6)&n= bsp; Ecological conditions favoring human diseases

7)&n= bsp; Lack of coal deposits

8)&n= bsp; High transp= ort costs.

[drawing on Sachs and Bloom and Sachs}

Diamond points out the germs and crops have a historical pattern that might be important. =

Eurasia is the center of domestication for both grains and livestock.

This lead to productivity improvements, possibility of urbanization,…

It also lead humans to be in long term contact with livestock, thus building up resistance (or dying) from exposure to livestock – human diseases. =

Diamond also notes that these livestock and grains spread ‘east west’ rather than ‘north south’. Sahara, tse tse.

Olson cited noting of the 56 heaviest seeded wild grasses on earth, 33 in western= Eurasia, 4 in SSA, 2 in SA. Grains are key= .

I= nstitution view.

Environment’s m= ain impact runs through institutions.

What type of crops ca= n be produced? Different crops lea= d to different social / political outcomes.&nbs= p; Wheat leads to good outcomes, sugar to bad outcomes by the nature of= the production of the crop.

See top of page 10

Settlers from Western Europe came with institutions. Places where western Europeans wen= t (and did not die) inherited institutions developed elsewhere. They went where their crops worked= .

Colonial argument: in places with bad diseases, settl= ers established extractive institutions. ‘Indirect rule’.

Institutions created = by colonialists endured after independence.&n= bsp;

O= ne other finding – among colonized places, those that had higher urbanization / population density in 1500 have lower incomes and institutional quality tod= ay. T= his suggests it is more than just “tropics”.

In places where they = intended to stay, they developed institutions to protect private property and limit = the power of the state.

Policy view: economic policies and institutions reflect current knowledge and political forces. History does not play a large role= .

Sound macroeconomic p= olicies, openness to international trade, and the absence= of capital account controls will tend to foster long-run growth (p. 11)

Turn to section 3 to = look at correlations.

Note that in contrast= to Barro, we are looking at GDP per capita levels, not growth.

When we get to instit= utions, sidetrack to www.govindicators.or= g

4.&n= bsp; Results.

1)&n= bsp; Do endowmen= ts explain variation in GDP per capita?

2)&n= bsp; Second, do endowments explain variation in institutional development?

3)&n= bsp; Do endowmen= ts explain variation in GDP per capita beyond the impact on institutions?=

4)&n= bsp; Do policies impact GDP per capita after the rest is taken into account?

OLS Table 2 answers q= uestion 1 “YES”

OLS table 3 results a= nswers question 2 “YES”.

Now we need to answer question 3. We need to use instrumental variables. First stage, estimate institutions index. Then plug in estimated index in a regression of the GDP level.

Results suggest that endowments influence institutions, and institutions influence GNP. There is not evidence of a direct = link from endowments to growth once the impact of endowments on institutions is taken into account.

4= .4, move on to include policies. Simultaneity issues here as well &= #8211; do good policies lead to growth or does growth lead to good policies?<= /o:p>