Ronan Lyons | Personal Website
Ronan Lyons | Personal Website

October 2012

Uganda-gate: Debunking two myths about sub-Saharan Africa

Earlier in the week, news broke that some of Irish overseas development assistance to Uganda had been misappropriated by the Prime Minister’s office. In a Father Ted-style development, it seems that up to €4m of Irish aid ended up in the Prime Minister’s account. A poll taken the next day on had, depressingly, almost half of those voting no to the question “Should Ireland distribute foreign aid while it is in a bailout?”

What was particularly depressing to those who believe that the return on taxpayers’ money when spent on things like malaria nets and primary schooling in sub-Saharan Africa is huge was the nature of the discussion that followed under the poll. Here are some choice comments [if it looks like a typo, there’s an implied (sic.)!]:

  • “I going to be blunt here… Not a single euro more should we give. It is disgraceful pumping money into these warlord run countries when we are in a bailout program.”
  • “But the current situation is simply unsustainable. We are borrowing billions of euro, at a five per cent interest rate, just to hand it over to corrupt foreign heads of state. Madness.”
  • “Do we want to be seen giving money to people who are clearly living in the stone age??I don’t want my money going to these people. I want it going to my people, Irish people, for health,education, job creation and infrastructure!!!”
  • “Africa is mineral rich but kept in poor by foreign bankers and Corporations who suck each country dry and prop up evil corrupt regimes.”

I used to work in Irish Aid, as it happens, although only briefly, so I won’t go into the ins and outs of something I’m not current on. However, there are a number of misconceptions about sub-Saharan Africa revealed in these comments that I think would be useful and relatively straightforward to dispel.

I’ll take two: the belief that the continent is beyond fixing (aka that Western so-called aid is keeping these countries poor); and that the political system in sub-Saharan African is corrupt and/or authoritarian.

Myth #1: Africa’s rulers are corrupt war-mongering dictators

Sub-Saharan Africa is home to forty-plus countries (the birth of Namibia, Eritrea and South Sudan, among others, mean the number is not constant). Thanks to the Polity Project, which quantitatively assesses the democratic/authoritarian nature of every government from 1800 on, it is possible to see whether Africa remains home to dictators or instead whether it is a hotbed of democracy.

The Polity score ranks every country from -10 (full-blow dictatorship) to +10 (full-blown democracy), with special indicators for where government has collapsed (due to foreign occupation or civil war). The graph below shows for 1975, 1995 and 2011, the percentage of countries in Africa that were autocratic (a score of -6 or less), autocratic-leaning (-1 to -5), democratic-leaning (zero to 5) and democratic (+6 or more), as well as the worst off, those without a government.

Proportion of sub-Saharan countries by regime type, 1975-2011

What’s pretty clear is that the trend is a positive one. In 1975, six out of every seven countries in sub-Saharan Africa was a dictatorship. Now, there are only two autocratic countries on the continent, Swaziland and Eritrea.

In 1975, there were only two islands of democracy, literally in the case of Mauritius, 500 miles off the coast of Madagascar (the other being Botswana). Now there are 19 democracies and 11 other countries that could be described as democratic-leaning.

No-one is for a minute arguing that the political system in Africa is perfect – democracies can be just as corrupt as autocracies in certain circumstances – but the idea that there is some sort of warlord class of dictator still ruling over the world’s second most populous continent is ridiculously uninformed.

Myth #2: The African economy is a basket case with no hope

The other clear implication that one gets from reading comments such as those on’s poll is that Africa is an economic no-hoper, as poor today as it has ever been. This is a tough one for NGOs and Governments to counteract. Make the case too strongly that there have been returns on the investment of aid and people start to question whether it is needed anymore.

But I think those involved in the aid industry do have a case to answer for in not showing the progress that has been made. The graph below shows average growth rates and also income per head in Africa at the end of each five-year period (population-weighted). The source is the IMF World Economic Outlook, the primary repository of comparable international statistics over time.

Average income and economic growth (% annual) in sub-Saharan Africa, 1980-2017

Average annual income in Africa has risen from less than $900 in 1980 to $2,500 now and is set to increase to over $3,100 in the next five years. Even with growth slowing to about 5% per annum, by the late 2020s, the average increase in an African’s income in three years will be greater than their entire income was in the late 1970s.

Which brings us on to growth. GDP growth in sub-Saharan Africa averaged 2.5% in the twenty years to 2000, barely enough to cover population growth. Since then, population growth has slowed while economic growth has accelerated. The average rate of growth in the period 2000-2017 is expected to be 6% – with the final few years actually slightly slower than the period 2000-2012.

So again, no-one is arguing that Africa is perfect or even that there are no chronic situations in Africa. But those who assert that money put into Africa is money wasted, because ‘clearly the continent is a basket case (possibly under the thumb of the West)’ again make their case devoid of all evidence.

Indeed, what’s ironic is that those who believe these countries are in the Stone Age are the ones themselves who have rather archaic opinions of what is and is not happening in Africa.

The first house price index based on the house price register

Over the weekend, the Sunday Independent featured some analysis I’ve done with the help of the team at on the Residential Property Price Register (RPPR). The full report is available here (PDF) – in this post, I’ll give a quick overview of what we did and what we found.

What is it?

The aim of the exercise was to produce a property price index for the residential market in Ireland, one based on transactions price (as per CSO but not the existing asking price index) and for all properties, not just mortgage-backed ones (as per the existing index, but not the CSO).

Many, such as NAMAWineLake, greeted the RPPR as the end to all existing property market reports but unfortunately, it’s not quite that simple. The Register certainly gives us valuable information on the number of transactions, by county and by month – and ultimately, it is volume, not value, where we will see a property market “recovery”.

The Register contains no information, however, on property type or size. Without this information, the only vaguely reliable statistic on trends in property prices is the median price. And even that is of limited use, in a market with an anaemic level of transactions, as I pointed out earlier in the month.

Luckily, where it is possible to connect up addresses in the RPPR with addresses in very large datasets of properties, such as the archives, then the attributes needed for a house price index – in particular location, type and size – can be added in and standard methods applied to develop a house price index for Ireland.

How was it done?

The first part of the exercise was matching up the transactions with properties in the archives. Fortunately – given I’ve no idea how to do this efficiently – I was not involved in this task, which was led by Paul, Head of Development at Daft. They managed to identify a property type (detached, semi-d, terraced, bungalow or apartment) for just under 20,000 transactions. Adding in information on bedrooms and bathrooms reduced the sample to about 13,500 transactions.

For these, we have all the information needed to conduct hedonic (i.e. like-for-like) analysis. How representative is this sample? The bad news is that we cannot know for certain – if we knew more about the properties excluded, we wouldn’t have to exclude them. The good news, however, is that for the main heading we do know (county), they look to be representative. The proportion of all transactions that made it in the sample for analysis varies only from 28% in Leinster (ex-Dublin) to 23% (in Connacht-Ulster), with the figure for urban areas and Munster being 25%.

The model applied breaks down each property’s price into five components: time of transaction (by quarter), regional market (six within Dublin, the other cities are each one, and each county is another regional market), number of bedrooms, number of bathrooms (relative to bedrooms) and the property type.

The full model interacts the bedroom and type variables, as well as the quarterly trend, with broad region (Dublin, Other cities, Leinster, Munster, Connacht-Ulster). This is to allow property differentials to vary around the country and also to allow the trend to vary by region. For technophiles, the method is a pooled semi-log hedonic regression. The model explains over two-thirds of the variation in house prices (the R-squared is 68.4%).

One other methodological point is warranted. The dataset is a messy one, with outliers due to errors and to extreme or unusual properties.  To combat this, a relatively standard two-step procedure is used. At the first stage, the model is run and then the model’s prediction is compared to the actual price. Where the actual price is more than twice or less than half the predicted price, clearly other factors are at work for this property. These are then excluded. It’s worth noting that the number of properties excluded by this method (roughly 7%) is below the similar proportion in the CSO index.

What did you find?

Now… on to the goodies. The first graph above shows the average property price for a number of methods. The national average is calculated with each ‘regional market’ having a weight equal to its Census 2011 share. Three datasets are used: the full RPPR (50,000 properties; median only), the subsample of the RPPR used here (13,500 properties; both median and the results from the hedonic exercise described above), and the listings 2010-2012 dataset (200,000 properties; the existing methodology (a variant of the above), the methodology described above applied to this dataset; and the median).

There are a couple of points worth noting. Firstly, while the two median series based on RPPR data point to a small increase in prices in Q3 2012, this increase disappears using the like-for-like method (as I suggested it might). Secondly, the average price in all three series is above that in all three RPPR series, and this is especially the case when the model described above is applied to listings. This suggests that in general asking prices are 10% above transaction prices. (This latter fact is probably due to the fact the full listings model can include a variety of sub-county controls, something the limited transactions in the RPPR cannot do.)

The second graph (above) shows the quarter-on-quarter change in RPPR transactions and in property listings. Again, I took two things from this graph. The first is that by and large, they tell the same story – when list prices fell more sharply, so too did transaction prices. The difference at the end is certainly not typical and perhaps more noteworthy for that.

Secondly, both series point to much tougher market conditions in 2011 than in either 2010 or 2012. The average quarterly fall in prices in 2011 was 5.1% (in the RPPR series, 4.5% in list prices), compared to 3% before/after. Put another way, prices in 2012 Q3 were just 2.3% lower than in Q1, whereas twelve months previously they had fallen 10% in the same period. (For reference, they fell by 6.3% during the same period in 2010.)

The RPPR index is shown in the table below, alongside the existing asking price index and the CSO index, each of which has been rebased so that the average for 2010 is set to 100. The correlation between the RPPR index and the CSO index is 99.8%, while the correlation with the asking price index is 99.7%.

So what?

Those whose time is tight might be wondering what’s new in all this. In one sense, there’s not an awful lot of “new news” in this. Prices are close to stable but are falling gently still… but not at anything like the speed we saw in 2011 – “but sure we knew that from the CSO and reports anyway”.

The value in adding an RPPR index (as the Report will hopefully be doing from now on) is not that it will reveal necessarily anything hugely different from other sources. Rather, it is another signal from the market and together all these signals will paint a picture. This will come as a disappointment both to those hoping prices had actually been rising secretly and to those hoping they’d been falling sharply still. But to the rest of us, the fact that all market signals are saying roughly the same thing is welcome news: it will make it all the easier to know when the property crash is over.

Reports of our death are greatly exaggerated: existing house price reports respond

There now follows a statement from Continuity EPR (Existing Property Reports), which was delivered to me anonymously in the dark of night.

Hello. We are Continuity EPR. We represent existing property price reports. We would like to respond to NAMAWineLake’s recent obituary for us. We are not dead… Indeed nothing could be further from the truth!

It was very certainly interesting and rewarding to read NAMAWineLake’s assessment of our contribution over the last few years. However, NAMAWineLake – while incredibly strong on the legal and journalistic side of things – seems to be unaware of the economics and statistics that underpin solid property market reports like ours.

Take, for instance, this quotation “With the launch yesterday of a property price register in Ireland, we are now close to drawing a line under the relevance of the … indices which prevailed for so long” – which is quickly followed by a less than charitable “Good Riddance!”. There are three core reasons that, like us or lump us, existing property price reports are not only not dead but every bit as relevant now as they ever have been.

A register, not an index

Firstly, the new Property Price Register is not an index and not designed to be one. It is a register. As notes, “It is important to note that the Register is not intended as a ‘Property Price Index’. The details made available on the property register are limited to price, address and date of sale and do not include such details as property size or number of rooms.”

A NAMAWineLake partisan might say that it doesn’t matter and we can use averages or standard deviations or some such to throw off the shackles of existing indices and embrace the new register. As star economist and all-round nice guy Ronan Lyons has already outlined, even using something that sounds sensible like the median, which is unlikely to be skewed by errors or outliers is fraught with issues of interpretation and confidence.

Does anyone reading this believe that prices in Dublin rose 14% in the third quarter? That is what the PPR median would have you believe!

Quantity as well as quality

“But,” the NAMAWineLake fundamentalists cry, “surely someone somewhere will do something!” They write: “by January 2012 there should be some entrepreneurial move to create an index based on settled prices”.

It’s funny that they should mention this, because who will be best placed to compile such an index? That’s right, those already familiar with the data pitfalls and methodologies required. Unfortunately for NAMAWineLake, they may find that come January 2013, rather than celebrate the demise of existing reports, they are forced to report an expanded set of statistics from them.

Even aside from issues of individual errors, there is a major drawback associated with the Property Price Register in current market conditions. That drawback is volume. The greater the volume, the more reliable the findings from a rigorous property market report.

With just 400 transactions this year so far in all of Connacht and Ulster excluding Galway city, it will not be possible for even the best data wizards to ensure inclusion of structural differences in house prices between, say, the coastline north of Sligo town and the coastline west of it. This is something that at least one member of Continuity EPR, the Report, does every quarter, while another of our group, the CSO Index, can currently only compile indices for Dublin and the rest of the country.

So the quantity is there – while NAMAWineLake themselves argue that the quality is going to get better. Preliminary figures suggest a 96% correlation between asking prices and closing prices – more of that anon – but that may get even better. NAMAWineLake writes, remarkably in our very own obituary: “as a result of the transparency revealed in the property price register[,] asking prices should be more realistic… and buyer/seller expectations should be more closely matched.”

Remember 2006!

“But,” a NAMAWineLake fundamentalist might cry, “surely at some point in the future this problem will go away!” And indeed, it is presumably a question of when, rather than if, the volume of property sales is roughly similar to the volume of property listings. Surely then, we become defunct?

Cast your minds back to the second quarter of 2006. Ireland was getting ready for yet another Celtic Tiger summer and the ESRI-PTSB report showed annual house price growth of 16%, up from 13% in the first quarter. Yet along came the Report, which showed annual growth in asking prices slowing dramatically, from 14% in early 2006 to 6% in mid-2006. Indeed, it showed no change in effectively no change in asking prices between late 2005 and mid-2006.

It happened again in early 2007, when transactions-based measures showed house prices rising but the Sunday Times led with “Ireland’s Property Market R.I.P.”, based on the latest Report.

The point is: we did not come into existence because there was no other measure of house prices. The Department of the Environment produced an average house price series for decades before we came along. And throughout the last ten years, there has been either the ESRI-PTSB (a former member of our group) or the CSO index (a current member) tracking prices based on transactions.

The key point is that asking prices capture sellers’ expectations and this is generally a lead indicator for the market. The importance of indices based on asking prices will probably increase, rather than diminish, as Ireland (eventually) heads back into its next real estate boom.

Different Strokes

So, if we are so good, why do we vary so much? Why, every quarter, do the CSO, and myhome indices show different quarter-on-quarter percentages?

The first thing is to take a step back. Have a look at these indices from 2006 to now. Are they really telling vastly different stories? Not at all. Journalists would be well served by focusing less on the quarter-on-quarter changes and more on the bigger picture, such as annual or from-peak changes.

But why specifically do they vary from quarter to quarter? Remember that none of these metrics is equivalent. None of them are meant to be the same. They all use roughly the same methodology (hedonic regressions) but on different datasets. The CSO dataset uses all mortgage transactions, the two limitations being it cannot include cash transactions and that it is lagged by the length of time it takes between agreeing a price and drawing the mortgage down officially, which is typically at least a month. (Note that the Property Price Register also suffers from this timeliness drawback!)

The myhome index is based on a snapshot of all properties on the market at any one time. It captures a mix of past and present expectations on the part of sellers. The drawbacks here are its limited market share, especially outside Dublin, and the consequence of its design, i.e. that a property whose list price was set in 2009 but is still on the market is given the same weight as one whose price was set yesterday.

The index uses only those properties newly listed or that have changed their price in a particular quarter. This leaves rich datasets – as many in one quarter on average as there are transactions in total in the Property Price Register from the start of 2010 – but excludes all but the most recently set expectations.

We look forward to NAMAWineLake’s extended coverage of our reports come January and forgive them for their premature obituary!


P.R. O’Perty, Continuity EPR.


First results from the property price register – the IMF effect and rising prices?

As I mentioned in my earlier post today, Ireland has –at long long last – a public property price register. I also mentioned that it is a tricky business to extract any meaningful signals about the market as a whole from a database that has no structured information about location (other than county), never mind the property’s attributes, such as size, type, number of bedrooms, bathrooms, etc.

“Tricky” is not the same as “impossible”, though, so in this post, I’ve assembled a few facts and figures about the market, based on the 50,000 market-price transactions that have taken place since January 1 2010.

The IMF effect

The key use of the property price register, when talking about analysing the market as a whole, is the information it contains on the volume of transactions. The picture that emerges is a tale of two halves. The first half is the what happened in 2011 compared to 2010, the IMF effect as the economy in general and property market in particular adjusted to life in bail-out Ireland.

The impact on volumes was clear. The number of transactions in the first half of 2011, at just over 7,000, was 20% lower than the same period in 2010, when it was just over 8,800. Dublin in particular was hard hit by the uncertainty, with the number of transactions down 31%, from 3,100 to 2,100.

The first half of 2012 has been a period of regaining lost ground, with the total for the six months to June 2012 of 8,740 very close to the 8,800 two years previously. Dublin has been driving the higher volume of transactions, with over half of the extra 1,700 transactions occurring in the capital, but in general the increases in 2012 just offset the losses of 2011.

An overview of the volume of transactions by broad region is shown in the graph below.

Trading volume, by regional property market, 2010-2012

A first look at prices

Using a property price register to say anything at all about what has happened prices in the market as a whole is a game fraught with dangers. The most solid statistic is the median (or typical) price, i.e. the one with as many transactions cheaper than it as more expensive. It is quite different to the mean (or average) price, which will be skewed if one property sells for €100m, compared to €10m.

The picture that emerges is one probably far removed from the expectations of conspiratorial types who believed that everyone from the CSO to myself was involved in over-stating (or fabricating) a recovery in the property market. If anything, recent reports have understated the extent to which prices have stabilised. The first graph below shows the median price per region per quarter. The stability since the start of the year is notable, particularly when contrasted with free-falling prices in 2011. Also notable – with the exception of Q2 2012 – is the much smaller rate of decline in prices in Dublin than in other regions of the country. Since 2010, prices are down 20% in Dublin but 30% elsewhere.

House prices by region, 2010Q1-2012Q3 (2010=100)

The second graph below condenses this change in conditions since the start of the year. It shows the percentage change in median price between Q3 and Q1, for 2010, 2011 and 2012. Conditions worsened in 2011, with the national median price falling 7% between Q1 and Q3 2011, compared to 6% 12 months previously. By contrast, median prices have risen by 2% between Q1 and Q3 2012. Do not adjust your sets: yes, I wrote “prices have risen”!

Change in prices, Q3 over Q1, by region (2010-2012)

What next?

Suspicious types should be asking themselves why I chose the rate of change between Q1 and Q3, rather than the more obvious Q2 to Q3. Using Q2 instead would suggest a 14% increase in Dublin prices in just three months. Would anyone believe me if I said that?

That highlights the limitations to using the property price register for any sort of price analysis, even the (hopefully) careful median price analysis I’ve done above. The reason why a median price might shift 14% in just three months relates to the limitations of the data: we know nothing about the attributes of the properties in question. So maybe 3-beds formed a bigger chunk of the Q2 market while there were more 4-beds in Q3. Or perhaps conditions improved in South Dublin, and it formed a much bigger chunk in Q3 than in Q2. Ultimately we just don’t know.

It seems such a shame that having all this transaction price information out there, we can’t undertake any meaningful analysis of trends in prices… yet. I’m working with the tech guys at Daft and we’re hoping to match up addresses of property listings and property transactions. That way, we would know lots more about the properties sold, such as exact location, property type, and number of bedrooms and bathrooms. While this would inevitably be just a proportion of all transactions, and a small fraction of all listings, it would still allow the calculation of a substantive mix-adjusted price index of transactions, which would be a companion to the already existing asking price index.

Wish us luck!

Three tips for using Ireland’s property price register

At long long last, Ireland has a property price register! The site went live hours before the revised deadline (end of Q3 2012), and there will hopefully be time in future to quibble about how it took so long for what seems to be no more than “export to .csv” – particularly given how much extra had been done by private citizens, by the time I went to bed less than 12 hours after its launch, from rival sites to apps.

Nonetheless, we should take the opportunity to bask in this wonderful new service available to individuals and households active in the property market. With all the excitement, I thought I might offer three tips for researchers, from someone who spends an unhealthy amount of time with property market data, as I see from thepropertypin and other places that plenty are starting to do their own number-crunching.

Individual analysis, not market analysis

The first tip would be to remind people that the primary contribution of the register is in relation to individual transactions, not market-level analysis. What I mean by this is that someone thinking of buying a property has an area, village or street in mind, and with this tool, they can go on and – using their knowledge of particular properties, including things like size, condition, aspect, etc. – see what prices are like nearby.

The volume of sales

Nonetheless, people will be determined to divine some signals about the market as a whole from the register as currently constituted. So, when it comes to market-level analysis, the only really cast-iron contribution of the register is sales volumes. The register tells you how many properties were sold by month for each county. It is clear that there will be variability in individual county-month pairings, so my own recommendation would be to focus on quarterly data at the provincial level (counting Dublin as its own province).

What about prices?

Ultimately, though, people want to talk about prices, which I get. If you must talk about prices, best to talk about median prices (=MEDIAN in Excel), or perhaps the other quartiles such as 25th and 75th percentiles (one quarter and three quarters the way through the price distribution).

What I am stressing, though, is to avoid talking about mean prices (=AVERAGE in Excel). Even with median prices, the data are going to get dragged this way and that by the mix of locations and types of properties sold in any given quarter, which is why even median prices have greater health warnings than – believe it or not – hedonic or mix-adjusted prices such as the CSO and series.

Mean prices (what we think of when we typically think of averages) will be affected not only by that but also by extreme values and in particular errors in data entry by solicitors, which are not infrequent (having spent a lot of time yesterday wrestling with the data). For example, one property in Limerick sold for €127m [I’m guessing €127,000 is its true price] while others in “leafy” parts of Dublin sold for mere thousands [not hundreds of thousands].

So, use with caution as a market research tool. This is not a substitute for a rigorous index of house prices and, like it or lump it, indices such as the CSO and remain far better tools to analyse price trends in the market as a whole. But that should not take away from the main point, i.e. that this is an absolutely essential tool for those looking to buy and sell. Happy nosy-parkering!