Thursday, 29 March 2018

Housing: the happy self-delusion of ‘no shortage’

Posted by Paul Cheshire

The assertion that there is no actual shortage of houses seems to be gaining, if not traction, then at least supporters. Ian Mulheirn, of the consultants Oxford Economics, originated it. Matthew Parris on 10 Feb in The Spectator took up the cause of no housing shortage. It is true house prices have more or less doubled in real terms in every decade since the 1950s and continued to rise well ahead of inflation until this year. We know the young are getting squeezed out of owning houses. Similarly we know that the ONS measure of affordability shows houses are twice as unaffordable as in 1998 and are now less affordable relative to incomes than at the height of the 2007 boom. The measure may be imperfect but it is transparent - just the ratio of the median house price to median income.

The currency the ‘no shortage’ assertion has gained seems to be less the result of the persuasiveness of the evidence for it than the fact that it is a comforting narrative, appealing both to politicians and the CPRE/NIMBY brigade alike. It allows people to claim nothing uncomfortable (or effective) needs to be done about the crisis of housing affordability. It appeals to the NIMBY/CPRE brigade because the ‘solution’ doesn’t require us to build any more houses.

But the problem is that this claim of no ‘shortage of houses’ is based on no understanding of how housing markets work or even how one might usefully define a ‘shortage’. A shortage can only be usefully defined in terms of the balance of supply and demand. Basic economics is enough to give at least a hint that if prices are persistently rising in the long term, as house prices have, supply is less than demand. One of the additional paradoxes of our shortage, however, is that the constraints on supply imposed particularly by our planning system, cause prices to be more volatile too. So when demand does fall, all the adjustment is via price.

The evidence cited in support of the ‘no shortage’ assertion is that there were more houses per household in 2016 than there were in 1971. True: but there were more doctors per person in 2015 than in 1971 too. Far more. According to the World Bank the number of doctors per thousand people increased from 1 in 1971 to 2.8 2015. By comparison houses to households hardly increased at all: from 1 to 1.02 (England).

No one is asserting there is a surplus of doctors. As people get richer they demand more health care; that also happens as they get older. The ability of doctors to treat illness has greatly improved. It takes more doctors to treat cancer patients now than in 1971, partly because treatment can do so much more. The rising ratio of doctors to people reflects rising prosperity, the aging population and technical progress complementary to the demand for doctors. Much the same is true of houses. One of the inconvenient facts about the demand for houses is not just that as people get richer they demand more housing space but they do as they get older. Even after adjusting for income, education and other relevant factors, older people demand more housing space. As car ownership has grown people demand more space around their houses.

Not only do people buy bigger houses as they get richer, a few buy second houses. A second home is no more ‘needed’ than is a private pension or a new outfit. But incomes are not equally distributed and that is how markets work. Also, of course, since the real price of houses has risen so rapidly over the past two generations (and since 2007, other assets have performed so badly) houses are, increasingly, pensions; not just shelter. They may be bought to let using equity accumulated by older owner occupiers or just be a nest egg, even if vacant.

Houses are not barrels of Brent Crude – all the same. They all vary and one attribute on which they vary is space: both internal living space and space in gardens. Also location is a critical attribute of houses. As we have pointed out before houses in Barnsley – though buildable on brownfields – are not substitutes for houses in Oxford – where the high productivity and high paying jobs are.

As I and LSE colleagues recently showed, more restrictive local planning in fact increases the number of vacant houses as well as increasing commuting distances for workers. Because houses all vary, ‘house hunting’ involves search - for the best set of attributes you can afford, where you want to live. More restrictive local planning increases local house prices, creating an incentive to keep them occupied.

But tighter local planning also makes it more difficult to adapt the characteristics of the housing stock to what people want and where they want to live. The result is house hunting becomes less efficient, so more houses are empty. These two forces work against each other but when you carefully analyse the data over the past 30 years, it is clear that impaired search dominates. Local vacancies are 23 percent higher if local planning restrictiveness increases by one standard deviation. That is not all. Because it makes finding a suitable house locally more difficult it also increases the average distance people have to travel to work. The same increase in local restrictiveness causes a 6.1 percent rise in commuting distances.

If, in the 25 years to 2016, we had built in England at the rate we built in the 25 years to 1991, we would have built 2.2m more houses; that is we would have built 63 percent more houses than we did. New houses have also been smaller. The result is an aging stock of increasingly cramped housing. In 1967 62.1 percent of English houses were less than 50 years old: in 2015 that had shrunk to 38.8 percent: not much more than the proportion that were less than 25 years old in 1945 – despite almost no building during the 5 years of WWII. English houses are akin to Cuban cars: they are still in use but they are clapped out and polluting.

There are two other important points. The no housing shortage assertion tells us to look at figures for ‘net additions to the housing stock’ not at those for completions – houses actually built. Politicians increasingly do the same. For example, housing forecasts and targets in the latest London Plan are all in terms of ‘net additions’. On the face of it this sounds plausible but there are good reasons why the number of new houses actually built gives a far more reliable and useful figure. We have comparable data for a very long time – at least since 1946. So one can track house building over the long term. You can count houses but ‘net additions’ include conversions and changes of use as well as taking account of demolitions.

In the old days we used to knock down unfit houses, so ‘net additions’ were less than completions. But as the price of houses has gone up and up, instead of knocking them down we spruce them up, often subdividing them into two smaller units. Thus one house becomes two. In other words the worse the shortage of housing, the higher will be net additions relative to actual building. So measuring changes in the shortage of houses by comparing the number of households and the stock of homes over time will definitionally tend to underestimate the ‘shortage’. There may be ‘homeunits’ but, reflecting the growing shortage, they are increasingly small and less fit for purpose.

A similar argument applies to the denominator in the chosen measure of ‘shortage’: households. The increasing unaffordability of housing has generated an increase in young people living with, or returning to, their parents. Young couples put off having a family and live in a parent’s front room. Household formation is itself a function of houses prices and as they go up in real terms, household formation rates fall. Again Ian Mulheirn’s measure definitionally underestimates the shortage.

There is certainly a housing shortage in the sense that we have not been building enough houses to satisfy demand for at least 25 years. That there are fractionally more homeunits per household is irrelevant: incomes have risen so therefore demand has too – substantially; population has aged so (paradoxically to some) demand for housing has increased; the stock of houses, reflecting the falling rate of building, is aging and decreasing in average size; and because of the shortage, the formation of new households has been choked off.

Sorry but the politicians cannot just sit back comfortably. There is a housing shortage and it is causing a crisis of housing affordability. The only way to resolve the problem is to do something radical and uncomfortable. Build more houses.

Tuesday, 20 March 2018

Who benefits from neighbourhoods designated as conservation areas?

Homeowners and people nearby benefit, though the implications for society are less clear, writes Gabriel Ahlfeldt

Opinions on conservation areas are split. Proponents would argue that conservation areas protect the visual appearance of historic neighbourhoods, by preventing owners from making changes that would be detrimental to character. Opponents would counter that this form of protection, in practice, means a severe restriction of property rights and, as a result, owners cannot adapt their homes to changing needs. For example, it is difficult to expand the living area after having children, e.g. by means of an attic extension, or to improve the energy efficiency by having new PVC windows. There is also a concern that such restrictions make the property less attractive to potential buyers, depreciating its market value. By the same logic, however, it can also be argued that the prospect of neighbourhood stability that comes with conservation area designation adds to the value of a location and increases the market value of properties located in the area.

Theoretically, there could be a trade-off between the desire to preserve cultural heritage for future generations and current homeowners’ interests. However, whether the policy makes properties more or less attractive to homeowners today is an empirical issue that cannot be answered based on theoretical considerations alone.

In a recent article, we analyse the causes and consequences of conservation area designation. To structure the problem and ensure that the empirical analysis is as transparent as possible, we develop a new formal political economy model whose predictions we then take to the data. The key idea is that residents with a taste for architectural distinctiveness stand to benefit from preservation, since it solves a coordination problem. The problem arises where individual homeowners make undesirable alterations to their properties or fail to keep them well-maintained. While removing a troublesome tree here or installing double-glazing windows there may make sense for an individual homeowner, such changes quickly destroy the character of the neighbourhood if everyone makes them. Therefore, if local homeowners like the character of their neighbourhood, it is worthwhile imposing strict regulation and maintenance obligations in the form of a conservation areas.

Our model predicts that if homeowners are able to game the system, then neighbourhoods will be more likely to be designated in areas where residents have a greater preference for architectural amenities. Moreover, in terms of timing, designation is likely to happen in gentrifying neighbourhoods because the inflow of wealthy heritage-affine residents will make it more likely that the perceived benefits of designation exceed the costs of restricted property rights.

Because conservation areas are designated as soon as they become worthwhile from the perspective of local owners (i.e. when the benefits have risen to equal the costs), the model predicts that no effect on prices of properties inside a conservation area will be associated with the incidence of a designation. However, there could be a positive price response outside the designated conservation area. This is because homeowners just outside a conservation area enjoy the benefits when passing through or looking at a protected area, without facing the cost of property rights restrictions.

The empirical results are in line with the predictions from our theoretical model. We find that an increase in the number of affluent residents, and residents who hold a university degree, significantly increases the chances of an area being given conservation status. Concretely, a 1 per cent increase in the degree share is associated with an 11 per cent increase in the designated land share – a strong effect.

Using more than a million property transactions from Nationwide Building society, we also find that the designation of conservation status has no immediate effect on properties prices inside conservation areas, but there is sometimes a positive price effect just outside.

In analysing price effects, we control for various property characteristics (size, number of bathrooms, type of heating system, etc.) and compare the price trend inside to-be-designated conservation areas to other areas with similar characteristics that do not experience a change in designation status (a control group). Figure 1 demonstrates that there is no significant discontinuity in the price trend at the time of designation (year zero). If anything, there is an increased price growth over time.

Figure 1: Relative prices in to-be-designated conservation areas

Notes: Figure shows the difference in ln prices adjusted for property attributes between areas to be designated (in year zero) and matched areas that do not experience a change in designation status. Dots denote differences by year, solid black lines are the linear predictions from a difference-in-difference regression model allowing for an impact of the policy on levels in trends, and dashed black lines are the 95% confidence intervals.

Do these results support conservation area policies? Our findings suggest that the regulation helps homeowners solve a collective action problem, preventing some owners or landlords to freeride on others who invest in the historic housing stock. This means that the policy generally works in favour and not against the interests of local homeowners. It is not that current homeowners pay the price for preserving some heritage for future generations.

Yet, the implications for society as a whole are less clear. There are benefits, e.g. to commuters or tourists living in other areas, and costs, e.g. due to limited supply of new housing and affordability problems for first-time buyers, to residents living outside conservation. Our results suggest that the system essentially delegates the decision of whether or not to designate a neighbourhood as a conservation area to local homeowners. While this is good for homeowners in the area, the downside is that the costs and benefits to residents living outside such areas are ignored.

This blog post is based on the author’s paper Game of Zones: The Political Economy of Conservation Areas, co-authored with Kristoffer Moeller, Sevrin Waights and Nicolai Wendland, The Economic Journal, October 2017.

Blog reposted from LSE Business Review 

Monday, 12 March 2018

Empty homes, longer commutes: the unintended consequences of more restrictive local planning

Posted by Paul Cheshire, Christian Hilber, and Hans Koster

We have argued for a very long time that the fundamental problem with housing in Britain is a lack of supply: we have been underbuilding for two generations. Updating the simple estimate one of us made in 2014, the shortfall in building in England just since 1994 has gone up from about 2m homes to 2.5m. And we go on building the wrong sort of houses in the wrong places. We built more than twice as many per person in low-demand areas like Doncaster and Barnsley over the past 15 years than in Oxford and Cambridge.

In the pursuit of 'urban density' and 'building on Brownfields' we build too many cramped flats and maisonettes in less attractive cities or city neighbourhoods but almost no family friendly homes with gardens within reach of high paying jobs. We are spending £18bn on CrossRail but once it gets over the Green Belt boundary can build no houses. There is a price to pay for building on Brownfields and not allocating enough land: a crisis of affordability and a hugely inequitable transfer of housing assets to the rich and the old. Our housing crisis is a long-term crisis of supply: an endemic lack of supply interacting with rising demand.

One of the many arguments used for allocating less land for housing is 'all those vacant homes'. Even one of the least restrictive English Regions, the East Midlands, argued in their Regional Spatial Strategy that they could allocate less land because they assumed they would reduce housing vacancies by 0.5 percentage points (that is by about 12.5% of the long term average). Islington Council moved to use the planning system to tackle the 'scandal of empty homes' in 2014.

A logically equivalent assertion was made by Lord Rogers, the long standing advocate of urban density, in arguing against allowing offices to be converted to housing to help with London's housing supply: '…why should we rush to convert office blocks when we already have three-quarters of a million homes in England lying empty…?.

The trouble with interventions in the housing market is that however well-intentioned, they often generate all sorts of unintended consequences. Markets respond by generating new and sometimes perverse incentives. Reflecting this, one of our most recent research findings, just published in the Journal of Public Economics, is that more restrictive local planning actually has the net effect of increasing the proportion of vacant homes.

Having fewer empty houses is in itself a good thing. We have a shortage of houses and using the stock more intensively is a way of increasing efficiency. That is just how cut price airlines operate: they keep their seats full and their aircraft in the air. But they do not just assume planes will spend more of their lives in the air and seats will be fuller. They have an analysis of how to achieve this. 

Unless we understand why houses are vacant, we cannot rationally hope to reduce the numbers that are empty just by being more restrictive. It would come as no surprise to economists to observe that in well-functioning labour markets there was unemployment. Workers search for jobs and employers seek (better) qualified workers. Attempting to regulate unemployment away makes no sense. Vacant houses are equivalent to unemployed workers so it makes no more sense to try to 'regulate' vacancies away. That does not, of course, mean that we should not have policies to try to minimise their number (what those policies should be is material for another blog).

What really happens if, by tightening planning restrictiveness (saying no to more development proposals) a Local Authority makes housing even scarcer? Well on the one hand it will make housing in its area more expensive. This will increase the incentive to occupy housing, so reduce vacancies. 

Unfortunately more restrictive planning also makes it harder to modify homes or build new ones in different places or with different features to adapt the characteristics of the housing stock to the constantly changing patterns of demand. Jobs grow in a locality, so demand for houses there increases; the local school gets better so the demand for family sized homes increases; people buy a car, so want parking, a garage; they have fewer children or separate, so they want smaller homes - the list is potentially endless. The result of this is that in more restrictive locations people wanting a home find it more difficult to match their preferences to what is available. So they have to search longer or further afield. The result of that is there are more empty houses in the more tightly regulated places and more people living and commuting from the less regulated places further afield. Both this 'mismatch' and the price effects work at the same time but in opposite directions. So which dominates is an empirical question.

In our paper we go to great lengths to deal with problems of reverse causation and endogeneity. We have 30 years of data for 350 English Local Authorities and our results show with substantial reliability that the net effect of more local restrictiveness is not just to increase the proportion of empty homes but to increase it substantially. A one standard deviation increase in local restrictiveness causes the local vacancy rate to increase by nearly a quarter. That is not all. Because it makes finding a suitable house locally more difficult, it also increases the average distance people have to travel to work. The same increase in local restrictiveness causes a 6.1 percent rise in commuting distances.

So attempting to regulate housing vacancies away by allocating less land or being more restrictive with respect to new building or adaptation of existing structures, in fact increases the proportion of local homes that are empty as well making people who work in the area commute further. The absolute opposite of what the advocates of the policy want to achieve.

It is the mismatch between the preferences of households and the housing stock on offer that leads, other things equal, to higher vacancy rates in the more regulated - typically more desirable - places. Such constraints will likely cause a significant welfare loss. This is because too much housing stays empty in the most regulated, most desirable and, by implication, most productive places with the strongest demand and highest valuations for living space. So people are induced to commute further, while living in the "wrong" places.

The policy lesson would seem to be that planners should not allocate less land for development on the grounds that there are empty houses; nor should they make it more difficult to build or adapt houses. Rather they should encourage more flexibility with the number, location and type of houses.

There is moreover a nice irony for advocates of the 'compact city'. The most common policy to attempt to implement this ideal is to impose growth boundaries, making land scarcer, and implemented via Green Belts in Britain. Such policies also imply more restrictiveness with respect to adaptations of the existing stock or new construction in the areas in the Green Belts or beyond the containment boundary. Aiming for a compact city, in other words, makes planning policy more restrictive. Our results show this will have exactly the opposite to the intended effect because average commuting distances will lengthen as residents search further afield for housing they can afford and more closely matches their preferences.

[This post first appeared in CEP's CentrePiece, Spring 2018]

Thursday, 19 October 2017

Distributive Politics Inside the City?

Do local politicians target their voters when making policy decisions? In other words, did your mayor build that park next door to please her voters? This question has been discussed by economists and political scientists for decades, and belongs to a field of enquiry we call distributive politics. Answering this question is as important as understanding the effects of the policies themselves. Why? Because policies and public investment decisions are not created in a vacuum. To state the (painfully) obvious, politicians have motivations of their own – like everyone else – and it is sometimes these motivations, and not some loosely defined “greater good”, that determine policy.
Together with Luca Repetto, we have recently revisited this issue by taking a new approach and looking at the allocation of investments inside cities. Our question is straightforward: do Spanish mayors target their voters with local investment?  
Understanding the determinants of national policies and national investment allocations is of course important. Hence, it is not surprising that most studies in the academic literature have focused on national level allocations and their determinants. But roughly half of public investment is carried out by local governments (OECD) and we know little about how electoral factors shape those allocations. Moreover, the spatial extent and the policy levers of local governments are different from those of their national counterparts. So, government behaviour could be different too! In our paper, we try to find out whether this is the case.  
The main challenge when taking this question to the local level arises from a data problem. Transfers to and between local authorities are recorded in national and regional budgets, but allocations within those authorities are usually not easily accessible for research. In our paper, we overcome this problem by exploiting data from Plan E, a large stimulus program applied by Zapatero’s socialist government in Spain between late 2008 and 2011. This program transferred roughly 13 billion euros to Spanish municipalities in an attempt to kick-start the economy. Local governments had essentially full discretion to allocate investment projects within their boundaries and jumped on the resources immediately. Over 57,000 municipal investment projects where carried out under Plan E. These where usually parks, plazas, and basic service infrastructure, all of which are likely to have spatially localized benefits. And here comes the special treat. As an unusual present for future researchers, the national government required municipalities to geo-locate all projects.
By combining data on these projects with polling station data for Spanish municipal and national elections, we are able to study whether Spanish mayors allocate more Plan E spending to areas of strong electoral support. An illustration of the data we use in the project can be seen in Figure 1.
Figure 1

Notes: Points correspond to Plan E projects located in the municipality of Madrid. Census areas are coloured in red if the socialist party PSOE received the majority of votes in the 2007municipal election, with the intensity of the shade varying with the vote share. Similarly, blue areas correspond to areas where the right-wing PP obtained the majority of votes.

No Distributive Politics Inside the City? Our Analysis
Comparing allocations in cities ruled by different parties is tricky, because our cities are likely to be different in many dimensions. To deal with these confounding factors, we implement a close election regression discontinuity design. We compare municipalities where the socialist party (PSOE) barely won the mayoralty with municipalities where it barely lost. We then look at whether areas within these municipalities where PSOE had strong electoral support receive more resources under a socialist mayor.
The main results are illustrated in Figure 2. The horizontal axis represents the vote share distance to a PSOE majority. The vertical axis represents one of our measures of PSOE partisan alignment in the allocation of Plan E projects. You can think of it as the city level covariance between PSOE support and the amount of spending. The graph indicates that there is no difference in this measure on either side of the discontinuity. We interpret this as evidence that there is no partisan bias in the allocation of resources to Spanish municipalities. Whichever use mayors do of this money, they do not use it to invest in neighbourhoods where their voters live.

Figure 2

Note: The vertical axis shows different measures of bias in the allocation of Plan E projects towards PSOE voters. The horizontal axis shows the PSOE winning margin, defined as the vote share distance to a seat majority change. Dots are averages in 1% bins of the winning margin. The lines are local linear regression estimates.

It is tempting to extrapolate from this result and conclude that distributive politics do not play an important role within cities. Perhaps investments benefit a broader group of voters, as people moving beyond their residence and its surroundings enjoy the benefits of municipal investments throughout the city. Or maybe local politicians lack the sophistication of their national counterparts.[1] But we must be cautious. The targeting of supporters is not the only prediction in theories of distributive politics. There is a decades-long debate between political scientists on whether politicians target their supporters or, rather, target swing voters; voters who are likely to switch sides if policy is favourable to them. If the latter theory were correct, then distributive politics could still play an important role, albeit one that is invisible to us. To be sure, we will need more research in this area.
Despite these caveats, we continue to think our findings are good news. While there seems to be a good deal of partisan bias in allocation of national resources to cities, our results show this phenomenon appears to be absent within these cities. The optimism of the will may lead us to think that politicians could be targeting citizen based on their needs and not on their party affiliation. And perhaps that is good. Alternatively, the pessimism of the intellect may us to think that the game is being played in some other margin. We can’t say for sure, but we’ll try to revisit this issue in the future. I’ll keep you posted.

Carozzi, Felipe and Luca Repetto, "Distributive Politics inside the City? The Political Economy of Spain's Plan E", CESifo Working Paper No. 6628, August 2017 (newest version)
Carozzi, Felipe and Repetto, Luca , “Distributive politics inside the city? The political economy of Spain's Plan E”. SERC Discussion Papers, SERCDP0212. Spatial Economics Research Centre, London School of Economics and Political Science, London, UK. February 2017

[1] We discuss these and other possibilities in the last version of our working paper. In general, evidence for these alternative mechanisms is not very convincing.

Tuesday, 29 August 2017

What’s happened to rents in London and the UK after the Brexit referendum? Evidence from a new rental index.

Everyone talks about house prices – either rising or falling – as a vital indicator of local and national economic health, and a range of indices frequently make the headlines. Rents get less robust attention, largely due to a scarcity of good timely information on the current state of the market. This is clearly a problem, since now nearly 20% of households are private renters in England, and nearly 30% in London, a figure that has nearly doubled from what it was 10 years go (Survey of English Housing). It’s also a gap in our information about the economy, because rents are likely to be more responsive than sale prices to current conditions, local and national, and so a better barometer of pressures in the market. Turnover in the sales market is much slower, since people keep owner occupied homes for longer and properties can spend a long time on the market. In general prices reflect people’s expectations about long-run trends and will be sluggish to adjust to unanticipated shocks.

Last year I spent a bit of my time developing new rental indices for HomeLet based on their private rent data to try to address this omission. The index aims to carefully adjust rents for changes in the types of properties being rented out and their location, just like the leading house price indices. Fortune telling based on price indices can be a futile exercise, but these new rental indices are starting to reveal some fresh information and interesting patterns.

Figure 1 plots a range of different trends spanning June 2014 (the start date for reliable rent data) to July 2017. The trend in blue is the publicly available monthly Nationwide House Price Index for the UK. The red line is a version of the new HomeLet index of private sector rents, based on new rental agreements for the UK. At the bottom, in green, is the ONS ‘experimental’ rents index.

Figure 1: Rents and price indices for the UK

There are some intriguing features here. Unsurprisingly, all the indices agree rents and prices have risen in general over this period. Broadly speaking rents for new agreements and house prices follow similar patterns, and, up until June 2016 followed each other reasonable closely. These new rents exhibit a marked cyclicality, rising over the spring up to July and then tailing off a bit in the summer and autumn. The fall-off in rents in June 2016 looks dramatically deeper at the end of 2016 and beginning of 2017 – more on which later.

The broad similarity of the patterns in rents and prices is quite reassuring for people who like their economics simple, because the fundamentals of housing should be similar for both renters and owners. To an approximation we would expect prices to roughly reflect the value of the rents the property could generate in the future (plus expected capital gains, less depreciation, maintenance etc). 

Even so, it is often said that prices are rising much faster than rents in the UK, perhaps something to do with constraints on development making land scarce and pushing up property prices in anticipation of future capital gains, or some other form of speculative bubble or incentives that favoured buying, such as “Help to Buy”. As it turns out from this - admittedly short - time series of data, that wasn’t actually true nationally between mid-2014 and mid-2016 if we compare new rents with prices.

One reason underlying the perception that prices rise faster than rents might be that the main point of reference has been the ONS rental index – the green line – which indeed suggests rents have climbed in a sluggish plod, showing a 7% rise since June 2014. Prices, on the other hand, increased by over 12%. These comparisons – if they represent the long run picture, rather than a short term blip - would mean that renting a home is an ever-increasingly cheaper option than buying, or else the expected percentage capital gain on housing sales is ever-increasing. Neither of these possibilities seems very plausible.

But this comparison is misleading. The ONS index is based on valuation data and reflects average rents in the stock of private rental accommodation. This ONS index is based on data on rents for occupants of all lengths of tenure – those that have just begun a rental agreement, and those that have lived in a property for many years. This gives a poor indication of what a landlord can ask and what a tenant can expect to pay when setting up a new rental agreement. If new rents are rising. but landlords are less willing to raise rents for existing long tenure tenants than new tenants, then the rent in the stock will rise more slowly. Although this behaviour can be hard to rationalise, it is a commonly observed feature of the rental market.

I we look at new tenancy agreements, as in the HomeLet data, things look more in line with what simple theory would suggest. But something dramatic seems to have happened in mid-2016. New rents fell sharply and stagnated right up until May 2017, while prices continued to rise, with just a brief dip early in 2017. What could have caused this drop in rents? While it is impossible to attribute causality with a simple time series like this, some more insights emerge if we split out London from the rest – see Figure 2.

The blue line is the Nationwide quarterly house price index for London (unfortunately they do not publish a rest-of-UK index). The solid red line is the HomeLet rental index for London. The dashed red line is the HomeLet rental index for the rest of the UK. The green lines show the corresponding trends in the ONS rental index.

Here it is clear that the trends in prices and rents in London departed company some time back in 2015. But what is more striking is that London rents nose-dived in mid-2016 falling about 4 % in the May-May year-on-year comparison. Indeed it is this fall in London that explains the drop in the national index in Figure 1. The fall over the summer in the rest of the UK is not much different from what it was in 2015. Rents have recovered in the last few months, but only to about where they were this time last year. 

Figure 2: Rents and price indices for London and Rest of UK

House prices haven’t fallen (much, yet) in London – at least according the Nationwide index. But they have flattened off a lot over the period when rents started to tumble.

What can explain this sharp fall in rents in London and why haven’t prices done the same? There are many candidate explanations, which surely have something to do with a fall in demand relative to supply of rental accommodation.

The Brexit vote might seem like an obvious candidate. But it is a dubious explanation on its own. Although there is some recent evidence that net migration decreased over this period (mainly due to EU-8 emigration), there are still many more workers, both EU and non-EU, entering the country than leaving (net migration between March 2016 and March 2017 was +246,000). Based on recent years of data, London accounts for about 40% of net international migration which means an additional 100,000 foreign workers arrived in London between March 2016 and March 2017. By the way, if you wonder how these international workers are accommodated, some relevant facts to consider are that net international migration into London in 2015-16 was 126,000, net migration from London to other UK regions was 93,000 and the number of new dwellings completed was 24,000.

The longer run economic outlook and exchange rate changes might be relevant factors too, but would expect house prices to be more responsive than rents if people were anticipating a future economic downturn or were concerned about exchange rate risks to their investments.

So perhaps it is less to do with demand for living in London, and more to do with supply of rental accommodation and incentives for ownership versus renting. An associate of mine who is an expert in property valuation and knows the London market intimately, believes this is at least part of the story. Non-resident owners in London are letting their properties rather than selling, given the low interest rates and the low cost of holding property, which is has been increasing the supply of rental accommodation and pushing down rents.

Whatever the explanation turns out to be for these patterns, this new index promises some unique insights into the rental market that have previously been quite obscured by lack of timely data. Deeper analysis is obviously required to properly understand the causes.

Declaration of interest: I was given some remuneration for developing the code to estimate the HomeLet index, but have no other interests in the company.

How the index works:

Simply looking at average rents in sample in new rental agreements is potentially misleading if the composition of the sample – in terms of the location, size, type and quality or properties being let – is changing over time. For example, in recent years the share of homes that are privately rented relative to owner occupied has increased but these changes are not evenly spread geographically. So changes in average rents will reflect both changes in the rent you can expect for letting a property in a given location, and changes in the number of properties being let in different parts of the country or different parts of a city. The changing patterns of rental locations and changes in the type of property being let can lead to misleading long run trends and to short run volatility in average rents.

The ONS rental price index addresses this problem by estimating rents for a fixed ‘basket’ of properties – in much the same way as a retail price index looks at a basket of consumer goods, or house price indices such as those produced by the Nationwide and Halifax adjust for changes in the types of houses being sold. These are called ‘hedonic’ house price indices. The disadvantage of the ONS index for many users is that it estimates rents in the stock, rather than the price of newly agreed rents.

The new HomeLet rental index combines elements of the ‘hedonic’ approach, with some ‘smoothing’ of the series over time. The source data contains information on 20,000 new tenancy agreements each month, across the UK. The index is estimated by looking at estimated changes in rents from one month to the next for lets within the same small geographical area (e.g. postcode sector). The method applies statistical techniques (regression) to adjust for changes over time in the types of property being let (like the ONS and Nationwide indices), but in addition smooths out short run volatility by using information on trends in rents in recent months, rather than only a single month (it uses local cubic polynomial smoothing). The index therefore offers a big step forward by providing the first index of new rental prices that is properly adjusted for changes over time in the characteristics of rental properties.

Thursday, 27 July 2017

The Local Economic Impacts of Brexit

I've been working with colleagues at the Centre for Economic Performance (Swati Dhingra and Steve Machin) and the Centre for Cities (Naomi Clayton) to take a first look at the local economic impacts of Brexit. You can read the more technical CEP piece here and the less technical CfC piece here.

The research looks at the difference in predicted effects across all Local Authority Areas and across Primary Urban Areas under a 'soft' and a 'hard' Brexit scenario (the former involves zero tariffs, but increased non-tariff barriers with the EU, the latter involves non-zero tariffs and even higher non-tariff barriers). It also provides some initial analysis on whether these predicted impacts are likely to exacerbate or alleviate existing disparities and looks at how the predicted economic impacts of Brexit correlate with voting patterns from the referendum.

The results show that every local authority area is predicted to see a fall in GVA as a result of Brexit. The impacts are predicted to be more negative under ‘hard Brexit’ in every local authority area, as the increase in trade costs would be larger. Cities are predicted to see larger falls in GVA than non-urban areas in both scenarios. The average decrease in GVA under ‘soft Brexit’ is 1.2 per cent in cities compared to 1.1 per cent in non-urban local-authority areas, and 2.3 per cent compared to 2.0 per cent, respectively under ‘hard Brexit’. There is less variation between cities compared to individual local authorities, however, as urban economies tend to have more diverse sectoral profiles.

The report also suggests that in both scenarios, it is richer cities – predominantly in the South of England – which will be hit hardest and most directly by Brexit. This reflects the fact that these cities specialise in financial and business services that are predicted to be hardest hit by the increase in tariff and non-tariff barriers that Brexit could bring. This pattern of results means that the predicted negative impacts are biggest for areas that tended to vote remain.

We aren't yet able to model how those initial impacts will change as the economy adjusts. It's quite possible that the places experiencing the biggest initial shock are not necessarily those that will experience the most negative effects once the economy has adjusted (we draw parallels with the financial crisis).

You can read more about all of these issues in the reports, so let me take the opportunity in this blog post to highlight my personal take on a few important issues:
  • These are predictions relative to trend (i.e. what would have happened if the UK had stayed in the EU).
  • We don’t take a stance on detailed institutional arrangements. Soft Brexit is ‘like a customs union’ – in that we assume zero tariff barriers and external tariff as now; but we take no position on changes in external tariff and who would decide on those.
  • The model estimates medium to long run impact of changes in trade costs. We’re ignoring effects on innovation, immigration, inward investment, etc. We’re also ignoring adjustment. All of this should be clear from the report, but worth highlighting I think.
  • Be careful on over interpreting the sector specific effects – particularly for industries like oil and air services where non-trade factors matter more. The technical paper has some discussion of the issues.
  • The findings in which I have the most confidence are the general trends (1) bigger negative effects for areas specialised in services (setting aside any concerns about the exact sectoral predictions, we know non-tariff barriers and substitution away from UK supply will matter a lot for those sectors); (2) general north south pattern given those sectoral impacts; (3) correlation with vote remain (those predicted to be hardest hit voted to stay); (4) correlation with median wages (rich places hit worse); (5) once the economy adjusts things could look quite difference. 
  • I would avoid getting too hung up on exact rankings. These will change as analysis is refined, other factors added in, etc. (I realise that nothing I say here will stop people from doing this)

The figures in this paper represent a first attempt to look at the Local Authority impacts of the increases in trade barriers associated with Brexit. Further work will be needed to better understand these impacts, to understand the impacts working through other channels, such as migration and investment, and to understand the longer run impacts as the economy adjusts. In short, these figures are far from the last word, but they do provide an initial indication of the way in which the impact of Brexit may be felt differently across the areas of Great Britain.

Monday, 22 May 2017

East Side Story: Historical Pollution and Neighbourhood Sorting

Posted by Stephan Heblich (Bristol + SERC), Alex Trew (St Andrews) and Yanos Zylberberg (Bristol)

“In Manchester [...] prevailing and strongest winds [blow] from the South West. This meant that when the dense sulphurous smoke left Manchester's tall chimneys it usually moved North East, and this was to have a marked effect on the shaping of the city.” 

Stephen Mosley, 2013, The Chimney of the World. 

The East Sides of cities such as London, Vancouver, New York and Paris have historically been the poorest. Some, but not all, have gentrified more recently, and this gentrification has been at the centre of media attention. These observations uncover two questions. Why were these neighbourhoods poor to begin with, and why did some gentrify while others did not? Our answer to this question has been recently discussed in a Guardian article titled “Blowing in the wind: why do so many cities have poor east ends?”

Pollution and sorting 

The explanation lies in the air pollution emitted by the coal-burning factories of the industrial era. During the Industrial Revolution, many areas of Manchester, for example, were covered with layers of soot. Black stains on the pavements and buildings of areas such as Victoria Station or Ancoats in North-East Manchester remained until very recently. With the wrong weather, some areas could be submerged under dark, thick smog. This environmental disamenity made them unpleasant places to live and, in response, those who could afford moved to the neighbourhoods spared by the pollution. This sorting resulted in an unequal distribution of social classes across the city. Since the prevailing winds (The Westerlies) in the northern hemisphere blow from the West to the East, the most visible component of such process ended up being the observed West-East differential in neighbourhood composition.

In Heblich, Trew and Zylberberg (2016), we document this process focusing on the industrial cities of Victorian England. We reconstruct atmospheric pollution and neighbourhood composition at a precise geographic level across English cities using a combination of historic Ordinance Survey maps and individual census records around the year 1881. To locate the main pollution sources, we identify the location of around 5,000 industrial chimneys on these maps and reconstruct the average annual atmospheric pollution within cities using recent atmospheric dispersion models. The chimneys of Victorian England were not very tall compared to modern standards and the coal was burned at a lower temperature. As a result, the distribution of pollution was markedly influenced by topography and wind conditions. It was also more severe -- in cities like Manchester, neighbourhoods such as Ancoats were far more polluted than contemporary Beijing while some others (e.g., Davyhulme) were below the current pollution alert levels. We find that these pollution disparities are reflected in neighbourhood composition. There is a strong connection between the presence of air pollution and the share of low-skilled workers at the end of the nineteenth century. Figure 1 shows that such a correlation was absent before coal became the major energy source at the beginning of the century. The observed effect is substantial: the difference between being in the 10% and 90% most polluted neighbourhoods of Manchester was a difference of about 20 percentage points in the share of low-skilled workers. Most interestingly, the relationship between the presence of historic pollution and the share of low skilled workers in 2011 turns out to be quantitatively comparable to the one observed at the end of the nineteenth century.

Smoke abatement and the persistence of segregation 

The previous result leaves one question unanswered. How could sorting caused by 1880 pollution be visible nowadays almost 100 years after the 1926 Smoke Abatement Act and 50 years after the Clean Air Acts (which quickly and considerably reduced the extent of coal-based pollution within cities)? While the average correlation between (past) air pollution and the share of low-skilled workers is almost as large as before in 2011, this observation masks important differences. In neighbourhoods where past pollution levels were close to the city average (either slightly lower or higher), social segregation disappeared between the Clean Air Acts and today. For instance, cities with low overall levels of pollution, where all neighbourhoods were close to the city average, do not display any West-East differences in outcomes today. By contrast, neighbourhoods where past pollution was well above or well below the city average (mostly present in cities with high average pollution) are locked in their historical equilibrium. These non-linearities in the persistence of neighbourhood composition are shown in Figure 1. This illustrates the existence of ‘tipping dynamics’: past a certain threshold, a poor neighbourhood repels richer residents even when the original disamenity has long waned. While it is beyond the scope of our project to identify the main channels at play, we present indicative evidence that this might be related to (the lack of) work opportunities, school composition (Free-Meal pupils), crime incidence or the quality of the housing stock.

Figure 1: Share of low-skilled workers and pollution (1817, 1881, 1971-2011). 

So what? Let us go back to the original question. Why are the East Sides of post-industrial cities so poor? Our findings suggest that they are poor because of the high levels of pollution during the industrial golden age, the tendency for the smoke to drift toward the East, the associated neighbourhood sorting and the persistence of segregation in very poor districts.

Could these findings inform policies today? First, there are large environmental disamenities in cities of developed economies, and even more pressing challenges in the cities of developing economies. In both cases, disamenities may be unequally spread within cities. The East Side Story shows that any such spatial inequality may anchor neighbourhoods in a persistent equilibrium, and there may be large costs associated with long-term spatial inequalities. Second, many developed economies employ costly urban policies to revive deprived areas, triggering new housing opportunities and business investment. Our findings indicate the importance of a sizable push to bring a neighbourhood past a certain threshold.