This week, for the first time in my life, I bought a copy of the Irish Farmers Journal. My eye was drawn to an advertisement in another newspaper noting that this week’s edition would have a special supplement on Agricultural Land Prices, broken down by county. Being a sucker for a good data set I bought the journal and examined the results.
The headline figure was that while property/housing continues to lose value in most parts of the country (apart from parts of Dublin), land prices per acre are increasing nationally for the first time in six years. Of course that increase is spatially varied with some counties losing and other counties gaining value at different rates. The map below (Figure 1) shows the national pattern at county level with the largest losses in adjusted price per acre in Louth and Armagh and the largest increases in counties such as Dublin, Meath, Wicklow, Kilkenny and Tipperary.
Figure 1. National Pattern of Agricultural Land Price in 2012 (Source: IFJ, March 9th, 2013)
What was to me more interesting, as a class of spatial data geek, were the lessons one could draw from this data source in terms of spatial data provenance, scale & sampling and the production of time-series data which tells us all something more broadly about how data is produced and that also functions as a simple lesson in critical cartography.
Spatial Data Provenance: The word provenance is taken from the antiques business and refers to the small bits of paperwork (a bill of sale, a letter, an auctioneer’s note) that authenticate the object being sold. In many ways this is what we also call metadata, data about data. If we ask the question of this data set what do we find out? Firstly the data is based on sales of all farms or land parcels, 1,218 in total, in the year 2012. The source of the data was from publicly available data (gathered from auctioneers and selling agents who honestly reported the sale values) and contained further information (data variables) on; the size of farm, total sale price, whether sold at public auction, tender or by private treaty, whether residential or non-residential farms and whether finally disposed of under the hammer at auction, or withdrawn for private sale later on. In Kilkenny for example (Figure 2), one can see the county specific statistics summarised in text and (carto)graphical form.
Figure 2. Summary Data on Agricultural Land Sales in Kilkenny in 2012 (Source: IFJ, March 9th, 2013)
What we can see is that Kilkenny had the highest land price in the country after Dublin but that this was based on a specific number of sales, 48 and included 30 farms of less than 40 acres, 16 between 40 and 99 acres and two above 100 acres. The average overall price per acre was €13,203 and was an increase from the previous year of almost 25%. The figure at the bottom of the page noted a ‘weighted’ average price (a word that should act as a red rag to a bull for critical geographers interested in spatial data). The editor of the supplement, Joanne Fox, provides a very useful summary on page 30 of the report on how that weighted average was produced which was a global figure based on the total value divided by the total number of acres within Kilkenny. This produces a slight variant to the raw average, €13,031, and essentially prioritised the sale of bigger farms.
Scale & Sampling: Of course being a geographer we would like to have access to this data at an individual farm level to perhaps overlay the averages costs against background soil, geology or land use classification layers but sometimes one is grateful for any data one can lay one’s hands on. It might also be interesting to know whether the farms the data was based on were suburban, semi- or deep rural? But the presence of some sort of meaningful individual sample, aggregated up to a spatial scale that is readily understandable and easy to interpret, is a valuable step and given the data is collected by a specialist newspaper that shows some understanding of the power and value of spatial information, we should be thankful. What is also useful for geography students and others to understand is that the different proportions of land sold within each county each year, shaped by the number and size of ‘sold farms’, will naturally affect prices (Figure 3 below). In another year, more sales and variable farm sizes may dramatically affect the average land price and make year on year comparability difficult. But of course that’s also the power of multiple statistical data, to iron out that variation and pick out longer term trends. I haven’t studied it in depth but it might be interesting to look also at the relationships between agricultural land availability and average price and to perhaps track this back over time.
Figure 3. Comparative Sampling from the Survey, 2012 (Source: IFJ, March 9th, 2013)
Time-Series: Given the Irish Farmers Journal have been publishing this data since 2007 I’m sorry I didn’t spot it before. Figures 4a and 4b below compares the national trends for the different years that the survey has been published and confirms that pattern of slow decline, a levelling off and a recent rise. This graphic can also be compared with the patterns for individual counties such as Kilkenny to provide a form of multilevel modelling in spatial terms, within which one can see a very similar pattern, with a small but necessary tweak of the Y-axis and its price range.
Figure 4. Comparative Trends 2007-2012 for a) Ireland and b) Kilkenny (Source: IFJ, March 9th, 2013)
I’m not completely sure what it is I want to say here other than to alert students to be aware of interesting spatial data sets emerging as it were, out of left field. My intent I guess, is also to try and promote a critical attention to the nature of spatial data, where it comes from, how its calculated, how geographically detailed it is, whether its available for different and comparable time-periods and how its reported and interpreted. In a country suffering a long bout of bad economic news, the headline figure for the report tentatively wonders whether this marks the ‘Signs of a recovery’. Perhaps this is too narrow or small a data set to draw such conclusions from but it does provide the slightest of glimmers in a dark-for-too-long sky.