More on agriculture productivity

I've written before about the productivity of farming compared to other industries in different countries. In that post I was using a very crude measure of comparative productivity, which was just the fraction of GDP produced by the agriculture sector divided by the sector's relative size. We can do a little better with the use of the World Development Indicators database. The database gives us two valuable indices: value added per agriculture worker and value added per worker in any given country. Knowing the percent of labor force employed in agriculture we can then calculate value added per non-agriculture worker (it's simple arithmetic) and compare the two. In other words, we can calculate directly how much stuff that an agriculture worker is producing on average, as percent of the amount of stuff produced on average by a non-agriculture worker.

So that's what I did for 89 countries of the world, taking each variable of interest as an average between 2000 and 2008. Below are two graphs with the results. (You can click on each graph for a larger image.) The first one shows per worker value added of agriculture plotted against GDP per capita. In Poland, per worker value added is about $2,400. What this means is that, in a year, an average Polish farmer produces stuff that's worth about twenty-four hundred dollars. Yes, you read that right. $2,400. The second graph shows per worker productivity of agriculture in relative terms, i.e. as a percentage of value added per non-farm worker, plotted against the size of employment in agriculture. In Poland, an average farmer produces about 7% of value produced by an average non-farm worker. Yes, you read that right. Seven percent. Seven over one hundred. Zero point oh seven. And the thing to keep in mind is that all this is happening in a situation where agriculture is hugely subsidized. There are direct cash transfers from both the Polish government and the EU. There are price controls in terms of state-guaranteed minimal sale prices of certain agriculture products. And last but not least, farmers in Poland are not required to pay income taxes. Yes, you read that right. Polish farmers are exempt from income taxes.

The second graph is interesting in one more respect, I think. One particular data point jumps out as a huge outlier: Romania. For its relatively low level of income and large size of the agro sector, Romanian agriculture is impossibly productive. I wonder what's going on there. I'll probably be coming back to this stuff at some point. Here are the graphs:

And if you want to see exactly where the measures and the graphs come from, R code is below the fold.

# pick WDI series, custom names for each series, start year and end year
wdi.vars <- c("SL.GDP.PCAP.EM.KD","EA.PRD.AGRI.KD",
"SL.AGR.EMPL.ZS","NY.GDP.PCAP.CD")
mynames  <- c("gdp.worker","gdp.agrowk","pct.agrowk","gdp.cap")
between  <- c(2000,2008)

# load WDI data
if (require("WDI")==FALSE) {
install.packages("WDI")
library(WDI)
}
wdi.load <- WDI(country="all",indicator=wdi.vars,
start=between[1],end=between[2])

# create a data frame with country-level averages
wdiStat <- function(x,y,z) {
tapply(x,y,z,na.rm=TRUE)
}
A <- apply(wdi.load[,4:ncol(wdi.load)],2,wdiStat,y=wdi.load$country,z=mean)
A <- data.frame(A)
colnames(A) <- mynames

# remove observations with missing values
A <- subset(A,(apply(apply(A,2,is.na),1,sum))==0)
# remove any observations that are not countries
A <- subset(A,(substr(rownames(A),1,4)%in%c("Euro","OECD","High","East",
"Lati","Uppe"))==FALSE)
# remove countries with extremely small populations
A <- subset(A,(rownames(A)%in%c("Barbados","Cyprus","Estonia","Iceland",
"St.Lucia","Latvia","Luxembourg","Malta"))==FALSE)

# create new variables
A$pct.agrowk <- A$pct.agrowk/100
# convert indicator "SL.GDP.PCAP.EM.KD" from 1990 $$ to 2000 $$
A$gdp.worker <- A$gdp.worker*1.32
# gdp.nonagr is value added per non-agriculture worker
gdp.nonagr <- (A$gdp.worker-(A$pct.agrowk*A$gdp.agrowk))/(1-A$pct.agrowk)
agro.share <- A$gdp.agrowk/A$gdp.worker
agro.comp  <- A$gdp.agrowk/gdp.nonagr
burden     <- (gdp.nonagr-A$gdp.worker)/gdp.nonagr
gdp2008    <- as.numeric(wdiStat(wdi.load[,ncol(wdi.load)]
[wdi.load$country%in%rownames(A)],wdi.load$country
[wdi.load$country%in%rownames(A)],max))

# graphs
X <- gdp2008
Y <- A$gdp.agrowk
cutpt <- ((Y > 4000) | (X > 20000))
plot(X,Y,xlim=c(0,70000),
main="Value Added Per Worker in Agriculture and GDP Per Capita",
xlab="2008 GDP Per Capita (Constant 2000 USD)",
ylab="Value Added Per Agriculture Worker (Constant 2000 USD)")
text(x=X[cutpt==TRUE],y=Y[cutpt==TRUE],
labels=rownames(A)[cutpt==TRUE],pos=4,cex=0.8,col="darkgreen")
text(x=X[rownames(A)=="Poland"],y=Y[rownames(A)=="Poland"],
labels=rownames(A)[rownames(A)=="Poland"],pos=4,cex=0.8,col="red")

X <- A$pct.agrowk*100
Y <- agro.comp*100
cutpt2 <- (gdp2008 >= 10000)
plot(X,Y,ylim=c(0,70),xlim=c(0,max(X)),
main="Agriculture: Per Worker Productivity versus Industry Size",
xlab="Percent Employed in Agriculture",
ylab="Agro Worker Output as Percent of Non-Agro Worker Output",cex.main=1.3)
text(x=X[cutpt2==TRUE],y=Y[cutpt2==TRUE],labels=rownames(A)[cutpt2==TRUE],
pos=4,cex=0.6,col="blue")
text(x=X[cutpt2==FALSE],y=Y[cutpt2==FALSE],labels=rownames(A)[cutpt2==FALSE],
pos=4,cex=0.6,col="red")
text(x=60,y=70,"Countries with GDP Per Capita > $10,000",
col="blue",cex=0.8)
text(x=60,y=67,"Countries with GDP Per Capita < $10,000",
col="red",cex=0.8)

Created by Pretty R at inside-R.org