Against the survival of the prettiest

It is a truth universally acknowledged that many buildings are ugly. Every day, we encounter offices, apartment blocks, retail sheds, McMansions, warehouses and hospitals that are unpleasant to look at. But here is what seems to me to be a much stranger fact: most of us struggle to identify an ugly building built before about 1930, and acutely so before 1830. In my experience, this is often true even for people who like the best recent buildings as much as they like the best old ones and who think we have good reasons not to build in old styles today. Thoughtful friends of contemporary design normally enjoy and respect old towns, while agreeing that quite a few recent places are terribly designed.

This is plausibly one reason why older homes tend to command a price premium, even after controlling for factors like location – a remarkable fact that distinguishes housing from virtually all other goods, whose value falls over time. These trends were presumably part of what Milan Kundera referred to with his famous phrase ‘the uglification of the world’.

If pressed, we can probably identify some borderline cases of ugly old buildings. Plenty of old buildings are clumsy, crude imitations of elite metropolitan styles. There are decrepit old buildings, which are sad to look at without being intrinsically ugly. Some people find highly ornate styles hard to stomach, like German Rococo, Churrigueresque or the temple architecture of Southern India. Old functional buildings, like mills, prisons and fortresses, can be bleak. Perhaps some of Hawksmoor’s and Vanbrugh’s buildings partake of ugliness of a kind. But there is an unmistakable quality of barrel-scraping about all this. Maybe Seaton Delaval, Nikkō Tōshō-gū and the Palazzo Pitti are ugly; but if they are, they are surely less ugly than countless buildings within walking distance of most of those reading this, and one wonders if they are really ugly in the same sense at all.

The claim that people tend to feel this way is, confessedly, based only on anecdotal evidence. But as anecdotal evidence goes, I think I have a lot of it. In 2019, I worked as a research assistant on the British government’s Building Better, Building Beautiful Commission. The Commission interviewed hundreds of people, and the unfavourable comparison of historic and many (not all) contemporary developments recurred constantly, quite regardless of the interviewee’s professional background or architectural orientation. What people did sometimes disagree about, though, is why this is.

The obvious explanation for why new buildings are uglier than old ones is that we build more ugly buildings today than we used to. But there is also a seemingly more sophisticated explanation, which draws on the idea of a survivorship bias.

The canonical story of a survivorship bias occurred during the Second World War, when the US military was analysing the damage its bombers had sustained from enemy fire. The statistician Abraham Wald pointed out that the Air Force’s records only included damage to bombers that had survived, meaning that all damage to the more vital parts of the planes had gone unrecorded. Paradoxically, therefore, the parts of planes which most needed armouring turned out to be exactly the ones with least recorded damage.

The survivorship bias theory about buildings starts from the premiss that people are less likely to demolish beautiful buildings than ugly ones, all else being equal. This will be partly because people like beauty, but also because the property owners who could afford to invest in beauty could also afford to invest in build quality. Thus, as the stock built in a given period ages, its uglier members will be demolished at a greater rate than its beautiful ones, leading to the proportion of beautiful buildings rising over time.

Eventually a situation might be reached in which only a handful of the most beautiful buildings have survived, generating an illusion that the buildings of that time were uniformly extremely beautiful. This is perhaps the situation with medieval build stock, of which only a tiny and completely unrepresentative sample remains today.

This is an elegant theory, and in my experience it attracts a lot of clever people. It also has the obvious advantage of removing the pressure to believe that we mysteriously got worse at making attractive buildings, at the same time as we were getting better at doing so many other things. Given what a strange and counterintuitive idea this is, this is no small virtue.

I think the survivorship theory probably does accurately model some real patterns, as I will go on to explain. But I do not believe that it can be the main explanation for the scarcity of old ugly buildings. In fact, I think we have a huge amount of easily accessible evidence for this, which is sufficient to disprove the theory fairly decisively.

The first kind of evidence is that old images fail to show the large numbers of ugly buildings that the survivorship theory predicts. Before the later nineteenth century, admittedly, these are only drawings or paintings, like those of Canaletto, Bernardo Bellotto and Eduard Gärtner. These show many decayed buildings but few ugly ones; but maybe they were inaccurate or unrepresentative.

From about 1850, however, we have large numbers of photographs, which again show few ugly buildings. It may be objected that there will be biases in the buildings that were chosen to be photographed. This is no doubt true, but much less so of aerial photography, beginning with this photo of Boston. We have a lot of evidence of this kind: 95,000 photographs of early-twentieth-century Britain are available here, covering much of urban Britain at the time, and many such photographs were of course also taken in other affluent countries. Again, we do not find the predicted swathes of lost ugly buildings.

It should be acknowledged that these pictures are taken from a distance, where some defects will be invisible. It is also presumably true they will still evince some selection biases, like a preference for photographing city centres and cathedral towns. In any case, however, they offer little support to the survivorship theory.

The second kind of evidence is that there are classes of buildings of which a large proportion has survived, thus allowing relatively little opportunity for a survivorship bias to emerge. The most obvious and important category of which this is true is the build stock of the late nineteenth and early twentieth centuries. As can be seen from this map, a great part of nineteenth-century Paris has survived; this map shows the same thing for Amsterdam and other Dutch towns. These maps show that the survivorship rate is pretty high in urban Russia too.

I do not know of quite such neat resources for British cities, but by switching between maps on this database (or this one for London) it is obvious enough that the same pattern obtains: only a minority of Britain’s Victorian and Edwardian build stock has been demolished, and many Victorian and Edwardian neighbourhoods remain largely unaltered. So it just cannot be true that there is a huge survivorship bias for Victorian and Edwardian buildings, because so many of them have survived.

It is worth taking a closer look at the exceptions to this high survival rate, which fall into three main categories. The first is war damage – about five percent of the 1939 build stock in Britain, more in Germany, less in most of Western Europe, nothing in the USA or the neutral states. The second is commercial centres like the City of London, where many of the Victorian buildings have been replaced by taller ones. The third is neighbourhoods that were demolished in ‘slum clearance’ of the postwar period.

There are rather complicated relationships between buildings’ beauty and whether they were destroyed in World War II. Sometimes beautiful cities were intentionally selected for destruction, as in the ‘Baedeker Raids’. Occasionally they were intentionally spared, as in the case of Kyoto. More generally, bombing tended to be targetted against larger settlements, city centres, densely populated neighbourhoods, and industrial areas. It is hard to say what the overall survivorship effect would be. But it is in any case obvious that this cannot be a general explanation for the paucity of ugly old buildings, since the absolute quantity of destruction was too small, and since there are still very few ugly old buildings in the unbombed cities of Portugal, Switzerland or the United States.

The relatively high loss of stock in business centres is also complex. It is plausible that there is a tendency for businesses to preserve particularly spectacular or important nineteenth-century buildings (perhaps due to conservation laws) while allowing unremarkable ones to be replaced. On the other hand, downtowns saw especially high levels of architectural investment in the nineteenth century, precisely because they were normally already high-value areas: many therefore ended up with concentrations of magnificent Victorian commercial architecture. So I suspect there are two conflicting tendencies here: splendid individual buildings are less likely to be destroyed, but splendid neighbourhoods are more likely to be.

Working out which of these trends is stronger would be quite complicated. But it is also not really necessary, because it is obvious from other evidence that Victorian business centres are not the places where the missing ugly buildings are hiding. Downtowns are among the most heavily illustrated and photographed of places, so we really do have a decent idea of what most buildings in, say, central London in the 1920s looked like (consult the fifty-three volumes of the Survey of London, or alternatively search ‘London’ here). It just isn’t true that they are wildly misrepresented by those of their number that have survived.

Slum clearance by contrast almost certainly does generate some important survivorship biases, since the demolished neighbourhoods, like the Gorbals in Glasgow or St Ebbes in Oxford, were typically the poorest areas. But we know from photographs and from surviving neighbourhoods built to identical building regulations that though these buildings were often overcrowded and unsanitary, and though they were relatively austere architecturally, it is just not true that they were particularly ugly. As is often remarked, some neighbourhoods of physically similar buildings have become popular and fashionable since inner urban renewal. Oxford’s Jericho, which narrowly escaped destruction around the same time as St Ebbes, is one example.

The situation with pre-Victorian architecture is rather different. Before about 1850 (in Western Europe), a steadily increasing proportion of the original build stock has been demolished, so survivorship biases become more of an issue. Photographic evidence is also absent, and other pictorial evidence more fragmentary. But there are still cases where a category of buildings survive somewhat randomly, providing something of a controlled experiment. For example, it happened fairly often that towns or neighbourhoods were impoverished or depopulated by war or economic change, thereby largely avoiding redevelopment until the modern conservation movement arose to preserve them. Rothenburg ob der Tauber or the Schnoor of Bremen are examples of this. The buildings of these places survived due to the misfortunes of their inhabitants, not because they were beautiful; yet beautiful they are, with few exceptions.

Another category with a high survivorship rate is church architecture. Because of the Reformation, few English churches were replaced in the sixteenth, seventeenth and eighteenth centuries; by the nineteenth century there was considerable reverence for medieval architecture, so although many churches were extensively restored, relatively few were demolished. England’s 8,000-some medieval churches thus constitute a fairly good sample of the churches that were actually built in the Middle Ages. Some are unremarkable, but it is doubtful that there is even one that could seriously be called ugly. Churches are not of course representative of a period’s buildings, but they do allow for a different controlled comparison, namely: could it similarly be said that there is not one ugly modern English church?

I think there are some kinds of pre-1850 architecture where survivorship bias probably is important, however. The obvious example of this is the housing of the poor, especially the rural poor, of which the majority has been demolished or has perished – a very large majority from before, say, 1600. Although the traditionally bleak view of these buildings has been qualified by recent scholarship, these were certainly pretty bad homes by our standards – poorly insulated, highly flammable, structurally precarious, and easily infested by vermin. Without radical renovation, they surely could not command a price premium today in the way that Georgian and Victorian houses often do. So here we do see survivorship bias playing an important role.

But were these buildings actually ugly? To my mind, the paintings, reconstructions, occasional survivors and continuing traditions of vernacular architecture today do not suggest so. Thatched and timbered cottages are not splendid in the way that Gothic cathedrals are, but they are not generally ugly buildings. So even here, survivorship bias seems to get us only part of the way.

Many readers of this journal are interested in the idea of progress. In many respects, the history of housing conforms approximately to a progressive narrative: the homes we live in are a great deal better built and better insulated than those of our ancestors, and (housing crisis notwithstanding) they are generally much less crowded. But with respect to their aesthetic character, this picture becomes more doubtful. Ugliness seems to be quite common among new buildings, but rare and marginal among old ones. What has emerged here is that although survivorship bias probably does contribute to that to some extent, it is not the main explanation: premodern buildings may on average have been a bit less beautiful than those that have survived, but they still seem to have been ugly far less often than recent buildings are.

The survivorship theory sought to explain the apparent rise of ugliness in terms of a bias in the sample of buildings we are observing. There is another kind of bias theory, which seeks to explain it in terms of a bias in the observer, saying for instance that every generation is disposed to find recent buildings uglier than older ones, and that this is why recent buildings seem so to us. This is a complex and interesting idea, which I am not going to assess on this occasion. Suppose, though, that our eyes are to be trusted. If this is so, strange and eerie truths rise before us: that ugly buildings were once rare, that the ‘uglification of the world’ is real and that it is happening all around us.