# Beer vs Coffee. Does Urban Planning Change Consumption?

## Introduction

In September 2016, the media and social networks of the Russian Tyumen city were agitated after blogger Arkadiy Gerschmann criticized the city. He travelled there and visited the new residential neighborhoods and bluntly condemned them in a post. Here’s a shortened quote from it:

There are just 27 identical 17-storey towers in an asphalted field. Would one want to go outdoors in such a place? No, you’d like to buy a thicker door, park your car as close to the entrance as possible, to slip quickly and not see this nightmare. No wonder that the most basic commodity here, according to the market, is alcohol.

Then he shows a photo of beer-on-tap store, which are widespread:

Looking at 2GIS registry, there are really a lot of such stores in residential areas on the fringe. This is true in Tyumen, in Novosibirsk, and in many other big cities in Russia. In the 2010s, new residential buildings have ground floor fitted for commerce, and beer-on-tap shops are almost in every apartment building.

Is there a reason to worry about? I think so, because there’s as much beer commerce as in the degraded neighborhoods, while the residents are definitely not alcoholics or derelicts. These people buy cars (which is less common in Russia), they are able to have mortgage and pay for the estate. Usually, they are young people or adults with small children. How could they consume so much beer?

Beer-on-tap shops in Akatuyskiy block (left) and Zatulinskiy microdistrict (right), Novosibirsk.
Red circles: houses completed in 2010s. Yellow circles: Khruschov period (1957-1969).
Circle size indicates the no. of residents.

The modernist urbanism (which in USSR embodied took shape as these “microdistricts”) was criticized a lot in the West by many specialists like Jan Gehl, Andres Duany, etc.

Alexander Lozhkin, a Novosibirsk-based urban planner, has summarized the results of modernism in the former USSR.

There’s an evident gap between the wish to create quality cityscape we see in the projects, and the results on the ground in the residential neighborhoods and city centers. The actual cityscape is extremely low quality, absolutely incomparable with that of the traditional cities.

Europe attracts far more tourists than any part of the world. It’s the only region where people go not for natural landmarks, but to dive into the cities. Even the iconic modernist capitals, Brasilia and Chandigarh, that were meant to be urban planning showcases, have not attracted pilligrims. Also, these artificially created cities turned out not quite convenient for the locals.

Only extreme adventure bloggers could think of travelling to Soviet microdistricts as tourists. Those who live there usually do so because of no viable alternatives. Even when occasionally a huge amount of money is invested in landscapes (trees, sidewalks, benches), the level of comfort bears little resemblance to that of the old European cities.

Back to beer-on-tap. Arkadiy Gerschmann in his post goes immediately to the same conclusion as the one above, indicating the reasons: aggressive uncomfortable yards and streetscape, lifeless space.

I instead tried to make a hard proof of this thesis, with data.

## Coffee-Driven Center

In Spring 2016 me and Daria Kiselnikova made a research defining the boundaries of the actual center of Novosibirsk city. We tried plotting all companies from registries, but it highlighted the office buildings that hosted lots of small businesses in a rather dull street.

We decided to choose those sectors that contribute to vibrant streets, and then calculate the variety of business, and also the amount of businesses that are located on the frontage of blocks. Architect Dmitri Oschepkov suggested that only coffeeshops would be enough to highlight the real center, and beer-on-tap shops would indicate the real periphery.

Coffeeshops have a special market niche: eating there is too expensive, if possible, and people come there not for food or coffee per sé, but to meet or to stay in public space. That’s why coffeeshops are very sensitive to the centrality of place and to the streets quality. Another niche they take are takeaway drinks, which are popular in central streets and where people go strolling. In Russia, such places are common in the centers and on the fringe, but the latter does not have enough pedestrian traffic to support businesses.

That’s why we decided to also check which places have more coffeeshops and how it correlates with urban planning.

# Problem Statement

All together, we have two thesis to check:

1. Modernist suburbs have more beer shops than on average
2. Central streets and traditionally planned blocks have more coffeeshops

If we simply plot these two categories on a map, we’ll see exactly what is stated, except for it can be a false dependency. An example of such is “the more there are doctors in a city, the more there are deaths”. Of course, one value does not depend on the another, and they both depend on the population.

For instance, many observers ask if beer shops are simply proportional to the population density. To clean the value from population, I divided the amount of beer-on-tap shops by population. To be exact, for every apartment building I calculated the amount of shops in a 500 metres radius, then the number of people in the same radius, and divided the former by the latter. The result is called diffusion.

But still some areas are generally more friendly to retail outlets, hence in some places there can be, say 50% more beer-on-tap shops than usual, just as other kinds of retail will be 50% more frequent, and this should not be a reason to worry.

So I took grocery stores and pharmacies into account (fast-moving consumer goods, or FMCG), and always compared diffusion of beer shops with theirs.

Fundamentally, I’d like to check a more straightforward thesis:

Quality urbanism (streetscape & territory planning) creates a better life quality, which manifests through consumption and stimulates certain industries like coffeeshops. Low-quality urbanism stimulates lower quality life that shows itself in other kind of habits (higher alcohol consumption, microfinancing).

In particular, I wanted to check if modernist districts are more prone to bad consumption, compared to traditional city.

So far, I’ve not tried tools that can automatically recognize planning types. When this report was being prepared, a Russian architect Roman Kuchukov published a territory recognition method in his blog. Using this tool should be the next step in this research.

# Method Overview

Calculations go in the following way:

1. Each house has source data: number of residents, completion year, time to travel from the city center.
2. For every house I calculate the number of objects within 500 metres: beer-on-tap shops, coffeeshops, grocery stores and pharmacies, number of residents or apartments.
3. We divide these values by population/apartment amount, obtaining specific values (in other words, clean the values from the factor of population density). These values are called diffusion of coffeeshops or beer shops.

In principle, one could try removing the economic activity factor, by simply dividing the diffusions by the diffusion of FMCG stores. Unfortunately, the resulting value will contain a product of their errors in statistical terms, and will be very unreliable. Also, these values oscillate a lot, and will require heavy smoothing. So I do calculate them, but don’t rely on them alone, preferring to have FMCG diffusions as a safe check.

1. We calculate aggregate statistics for blocks, years, time of commute to the center, etc. Average value per group is weighted by apartments number, e.g. coffeeshop diffusion for each house is multiplied by its apartments number, and then the total sum is divided by sum of apartments.
2. Time of travel (which I here call “commute time to the center” interchangeably) is rounded upwards by 5 minutes. Years of construction are grouped by epochs: imperial, early Soviet, etc.
3. Average values are calculated for every pair of values & , just like for blocks.

Steps 4 to 6 create what is called pivot table, or an OLAP cube) with multiple dimensions that have levels of hierarchy (e.g. years belong to epochs).

Having this cube, we check if there are any trends along some axes or their combinations. We also may look at one column in such a table and check values, e.g. how does beer shops diffusion change with the commute time. Or we compare different epochs for the houses withn the same commute time.

The best slice that I observed was in 2 dimensions: construction epoch and commute time to the center.

## Construction Epoch As a Factor

For an easier classification, I divided construction years into several epochs. Between each of them, either economic conditions changed dramatically, or construction approaches did. Here’s a summary of all of them. For the foreign readedrs, I expanded this section significantly.

### Construction in the Russian Empire

Only houses from the late 19th and early 20th centuries have survived until today, with very rare exceptions.

Those years, cities were built in a traditional European manne, with dense streets grid, facades standing right at the sidewalks, many displays and entrances (usually with no more than 2 steps) that added to street activity. With the revolution and property expropriation, all economic incentives for construction vanished, and new construction halted until late 1920s when the Soviet government started its own projects.

Despite the revolution, life did not stop completely, and in the periphery under the White Army rule, some buildings might have been completed, so for safety we put the end of this epoch at 1919.

### Early Soviet Period - 1920s, Constructivism

Le Corbusier still did not present his Radiant City, and architects working for the Soviet government kept the previous traditions. Even in early 1930s we can see architects making bulidigns stay in contact with the sidewalks and showing pedestrians walking by.

Under the new total central control, construction focused in the central streets and for the Communist Party elite.

This period lasted until mid-1930.

### Stalin Period

In 1930s, the private economy was almost eliminated, and with more funds, Stalin government tried tried to create a showcase of their success. Construction still focused on elite housing, but now bulidings became bigger and more iconic. 12-storey buildings in Moscow from that time are quite common, and the famous five Moscow skycrapers were built in 1950s.

Streets stopped being planned according to urban traditions: now they usually left a gap for a greed front yard.

In the other cities, construction was not as impressive, but still the principle was the same: elite housing and detachment from urban streets.

Meanwhile, for the working class, the state provided only basic houses, wooden or of slag blocks. Sometimes, they imitated stone buildings with plastering, but most of these houses did not last long, and by 1990s were in poor conditions. Right now, some of them, poorly maintained, start falling apart.

### Khruschov Period - 1958 to late 1960s

Eventually, the iconic construction for the few was seen as inappropriate, and this created an exaggeration in the other direction: to build as much simple but reliable housing as possible for the masses. It was that period when the USSR was urbanizing, and housing conditions after four decades of neglection, were very poor.

Apartment buildigs, nicknamed “Khruschovka”, initially had high ceilings and big windows, were built of bricks, but then they were reduced in dimensions and built of prefabricated concrete panels. Some experimental series were less reliable, but on average, espite laments from most Russians, “khruschovkas” are in good conditions if maintained, and will stand for 100 years.

Russians usually complain of small apartment area of khruschovkas, but in my opinion, the standard 32 m² of a one-room apartment are enough for an affordable housing class. For instance, in Hong-Kong, a huge metropolis that attracts a lot of immigrants, the most common housing for rent is 25 m² studio. Often times, Russians indicate as “good” the apartments that have exessesive service area, for instance, a hallway of 15 m² or big corridors. Many state that these have to be the new standard. While there is a place for this kind of housing, they are still too big for the mass affordable market.

Back to Khruschov period, whether such bulidings were small or not, they were a big improvement for Soviet citizens, who used to live up to 4 people in one room, and just 10-12 years before this, in 1947, there was the last wave of deadly repressions.

Urban planning changed a lot. Residential neighborhoods were built in large scale on then city fringes, and planned according to Le Corbusier’s Radiant City concept: free-standing multi-apartment buildigns, with much space between them for good air circulation and insolation. Perimeter buildings were becoming more and more rare.

In most cities, 1960s were the period of the most intensive residential construction.

### Late Soviet. From 1970s Until 1993

There’s no defined boundary between Khruschov period and late soviet times. “Khruschovkas“ of the most common series were built up until late 1980s. The beginning of Brezhnev period should be marked at 1975, when the average area of the apartments started growing with the new apartment series. (See “Back to Khruschovkas”, Lebedev & Kiselnikova).

Apartments became spacier, and city superblocks were growing bigger with free standing buildings and curvy streets.

### 1990s

By early 1980s, Soviet economy oriented itself on exporting oil, and importing more and more commodities, as there were little incentives to work hard for the employees, and the managers, not being owners had no material incentive to innovate. In 1986, oil price dropped dramatically. As a form to alleviate the inefficiency and burden, Soviet government tried to transition to a market economy, allowing a limited form of private enterprizes, called cooperatives. But there was much resistance in the system and too little time for a transformation. Soviet government power weakened, and was overtaken by national republics governments, that declared indepedency, and the USSR was dissolved in 1991. The president of Russia allowed free domestic and foreign trade, and reduced public spending dramatically, which led many state owned enterprizes in crisis, while private sector was still too small to compensate.

Construction dropped by the factor of 3 to 4, and did not recover until 2002. The only exception was Moscow city, for many reasons, like metropolis flexibility, and the federal government spendings that fuelled the economy.

Russian economy started recovering in 1997, but the recovery was postponed by 2 more years by the Asian crisis. Then, in 2000, oil prices grew to $20..25. The new economic growth did not yet stimulate mass demand for housing, but the upper classes already could afford it, and private construction companies started new projects, mostly in city centers. In some rare cases, the reorganized construction factories, that made pre-fabricated concrete, started new projects on the fringe. Urban planning was not on national agenda. In that period, I recall myself observing construction sites and being pleased by new steel-and-glass high-risers. ### 2010s By 2008, oil prices grew from$10 in 1990s, to $120. The 2007 meltdown arrived in Russia about a year later, together with another drop in oil price. Construction companies, that formed in the rapid growth, were inefficient and/or risky, and in 2009 when many construction sites were frozen, the sector got state support. Regional governments started financing mortgages, usually for young families. Construction sector got a huge boost, and restarted, although with the same methods as before. The regional governments, dependent on the federal one, were trying to get as much floor area as possible, and boosted construction by selling cheap land in the gaps or on the fringe. In the biggest cities, construction companies produced built huge amounts of residential housing. In our work “Back to Khruschovkas”, we found out that construction focused moslty on small apartments and studios. In Novosibirsk city, the total area of new apartments was far bigger than in Khruschov time, but the average area of new apartments fell lower than in that period, to 43 m². Probably these records were broken in Moscow urban area, but this happened outside of Moscow administrative border, and should be investigated separately. ### Private Houses The goal of this classification is not exactly timing, but to categorize similar approaches of planning. That’s why we have a special “epoch”, or category, for detached houses. Since there’s no publicly available registry of these, I developed an algorithm to predict where and how many of these houses are there, described in a special section. ## Data Sources All together, there are 3 kinds of data in the research: 1. city blocks geometry 2. residential houses 3. retail objects Population data is taken and interpolated from Russian government database “Housing & Communal Services Reform”, that lists all multi-apartment houses, that usually have a communal service provider. I used the apartment number rather than data on residents, since the latter is a number of officially registered residents, and this figure is usually distorted in many places. For instance, in city centers, where many apartments are rented, there may be less registered residents than apartments. In the newly constructed houses, people may have already moved into, but either they delay registering, or the database is updated much later. Apartments number is a figure that is well known and gives a better estimate of residential density. In many places, students contribute to population and make a great portion of demand on daily commodities and alcohol. The dormitories where they live aren’t in the housing reform database. For some of them I found a standard report, that universities submit yearly, called “Monitoring of amount and structure of dormitory rent”, which contain numbers of residents and floor area. Dormitories compared to apartment buildings. Circle area = number of residents. Geometric objects were downloaded from OpenStreetMap with Overpass API. The retail data was taken from 2GIS business registry on October 5, 2016, from several sections (beer-on-tap shops, coffeeshops, grocery stores, pharmacies). Supermarkets or hypermarkets are usually not in these sections, and they were not included, although some of them may sell beer on tap. I did not estimate the completeness of the 2GIS registry, however in the cities where it works for years, it’s close to 100%. Reviewers have not found a single beer-on-tap shop or coffeeshop that was not in 2GIS. The cities studied are: Chelyabinsk, Chita, Ekaterinburg, Irkutsk, Kazan, Krasnodar, Moscow, Nizhniy Novgorod, Novosibirsk, Omsk, Perm, Rostov-on-Don, Saint Petersburg, Tver, Togliatti, Tomsk, Tyumen, Ufa, Volgograd, Voronezh. ## Detached Houses Interpolation These houses make up about 15% of cities population, but still there’s no publicly available database on them. Their share is relatively low, and density is low too, but in those parts where dense urbanized blocks are neighbored by these detached houses, the population of the latter contributes a good deal of customers. When I only calculated the apartment buildigns, the specific values got off-scale. For instance, if a buliding has 50 apartments and there are 2 beer shops nearby, the specific amount, shops per 10,000 apartments is 2 / 50 * 10000 = 400 Normal values are 5 to 10 per 10,000 apartments. But if we count 950 houses (or to be more exact apartment equivalents) in the 500 metres radius (which means each apartment has a lot of 1000 m2), the specific value becomes 2 / 1000 * 10000 = 20 20 is a high value but not off-scale. I developed an algorithm to estimate the distribution of people living in these buildings. The method is that I add points along roads and within areas occupied by such houses. Then I distribute the people evenly between these “houses”. This way I can reflect the difference between dense near-urban areas and those rural further outside cities. The algorithm makes this distribution more smooth, overestimating the rural areas, but it’s better than the previous approach: in 2015, I made a research on Novosibirsk parks where the population was distributed proportionally to the areas of the polygons, and inside them, the population was distributed evenly between points on a regular grid. This approach should have been much rougher. The way the new algorithm worked is this: with JOSM editor, me and my assistant drew polygons of the areas that these suburban/rural houses occupy. Then these polygons were imported into PostGIS, together with road lines. All lines were united in a multiline (a set of lines) and then a buffer of 30 metres was drawn around them. Buffers were cut by the rural areas polygons, and the cuts inside the polygons were taken. Then, points were put along these lines at a same step. Suburban area in Leninskiy district of Novosibirsk and generated houses Then, we calculate how much population remains not accounted in the apartment bulidings, and divide them equally between the generated houses. In Novosibirsk, this calculation produced ~4 people per house. The algorithm aims to distribute population proportionally to the real density of suburban areas. We can’t account floor area of these houses, but at least it reflects the different types of areas: the more dense suburbia closer to city center, suburban areas on the outskirts, and also villages that were included in the urban boundary. In all these areas in Russia, the houses are distributed at equal distance along the street, and the difference is only in the depth of the lot. You can see that distinction in the pictures of Novosibirsk above. Krasnoyarsk suburbia and generated houses. ## Formulae ### Source Data for Apartment Buildings $ y_i $ - construction year, $ h_i $ - residents number, $ a_i $ - apartments number, $ t_i $ - time to reach city center in public transit. ### Indicators of the Building Vicinity Apartments number: $$a_{ri}=\sum_r^j a_j \space if \space (g_j \leftrightarrow gi) \le 500 m \qquad (1)$$ Number of beer shops: $$b_{ri}=\sum_r^j 1 \space if \space (g_j \leftrightarrow gi) \le 500 m \qquad (2)$$ Number of coffeeshops: $$c_{ri}=\sum_r^j 1 \space if \space (g_j \leftrightarrow gi) \le 500 m \qquad (3)$$ Number of grocery stores and pharmacies: $$f_{ri}=\sum_r^j 1 \space if \space (g_j \leftrightarrow gi) \le 500 m \qquad (4)$$ ### Indicators of Retail Diffusion Since a simple number of stores correlates with population density, while the specific values belowe are free of it. Number of outlets per 10k apartments within 500 m: beer shops: $ b_i=\frac{b_{ri}}{a_{ri}} \qquad (5) $ coffeeshops: $ c_i=\frac{c_{ri}}{a_{ri}} \qquad (6) $ grocery and drug stores: $ f_i=\frac{f_{ri}}{a_{ri}} \qquad (7) $ ### Aggregating The Values Since we work with specific values, their sum or simple average make no sense. To obtain average specific values, we weigh them by the number of apartments. For a city block q, the average beer diffusion is: $$b_q=\frac{\sum_i b_i a_j}{\sum_i a_i} \space \forall i: g_i \subset g_q \qquad (8)$$ Same way, we obtain average values for each construction year, taking all the houses of the same year and weighing them by apartments number. We can also calculate the total amount of buildings from that epoch: $$a_m=\sum_i a_i \space \forall i: y_i \subset y_m \qquad (9)$$ $ y_m $ - year number m (e.g. 1992, 2016), $ y_i $ - a year a building was completed, $ a_m $ - apartments completed in year m. Same way, calculate the diffusion of beer shops $ b_i $ for the year $ y_m $: $$b_m=\sum_i \frac{b_i a_i}{a_m} \space \forall i: y_i \subset y_m \qquad (10)$$ Also, we calculate the same average spcific values for epochs and commute times from city centers. Epoch $ i: e_i=n \Leftrightarrow y_i \in E_n $ where $ E_n $ - is the epoch’s interval (see below). Some Russian data scientists divide buildings in periods by the country leaders (USSR General Secretary or Russian President). In this work, they are divided by the economic conditions or urban planning paradigm. • Empire: (-∞, 1919 ] • Early Soviet Union: [1920, 1931] • Stalin: [1933, 1955] • Khruschov: [1956, 1969] • Late Soviet: [1970, 1992] • 1990-s crisis: [1993, 1999] • 2000-s recovery: [2000, 2009] • 2010-s subsidized mortgage: [2010, ∞) # Miscellaneous Calculations and Failed Hypothesis ## Bars and Microfinance Bars can also correlate with excessive alcohol consumption, and I plotted them on the map. It turned out that bars are mostly located in the same pattern as coffeeshops, in the centers. This means consumers more likely make appointments there, far from their homes. Hence, numbers of bars do not reflect anything about local residents, but rather the transit traffic or centrality of the locations. Microfinance companies became very widespread in Russia in the recent years. Their sales offices spread widely, and reached most of residential districts. Mostly, these are minioffices, with dimensions of a trailer cabin, installed at important bus stops. A bus stop in Omsk seen in panoram on Google Maps. Yellow circles are microfinance kiosks, blue circles are the other retail shops. But looking at their distribution in the cities, they also are concentrated in the centers and local centers. Their consumers may get a loan in any place they want on their way in the city, which means they’re not attached to the place where people live. Hence they characterize transit traffic, but not the local residents. ## Heatmaps Heatmaps became very widespread in data visualizing. Their problem is that it’s not clear what values their pixels stand for. A heatmap crosses boundaries, like railroad or water, still indicating some values beyond them, when there’s nothing at all. Heatmaps also produce artefacts with specific values. These artefacts do not exist in reality, althogh they seem entities to unprepared viewers (basically, anyone who hasn’t prepared such a map is unprepared to this). The issue with specific values remains even if when a smooth kernel function is used. Conclusion is that heatmaps are good only for gross values. Heatmap of a quotient of two figures. Blue arcs are the pixels where beer shops are closer than houses. But in reality this is just a field of bushes. Heatmap with simple sum ## Does The Prohibition to Sell Near Schools Have Any Effect In the early stages of the research, there was a hypothesis that schools vicinity may have effect. Many regions in Russia impose restrictions on the proximity of schools, kindergartens or cultural entities. In Novosibirsk, this limit is one of the biggest: alcohol can be sold no closer than 100 metres from a school. I plotted circles with 150 metres radius around the schools, to be sure in case if local authorities measure the distance by a straight line from the school fence. It turned out that these circles only push the beershops to the outer side of superblocks, but there’s no big barrier that would prohibit beershops from opening in any district. There’s still plenty of places without beer shops between these rose circles. Some shops are inside the circles, but as mentioned above, the buffer is 50 metres wider than necessary, and in some cases the distance may have been measured along footways, which produces bigger distance than a straight line. In many regions these distance limits are smaller: in Altai Krai it is 40 metres, in Tomsk Oblast it’s only 20 metres from any public social entity. This means you can sell alcohol in the next or in the second next building to them. Overall, these restrictions have only minor local effect, and do not prohibit beer shops from opening in a walking distance from any residential building. ## Block-level Statistics Unlike raster heatmap, the map divided into urban blocks is much easier to read, and it’s the preferred way to visualize for architects and urban planners. Although I need to mention though that urban blocks have poor correlation between their characteristics and any of the indicators that I calculated. The resulting map looks like this: A map of Omsk, indicating beer shops per 10k apartments The block-level averages are calculated in 2 steps, first the specific values of a house: $ (5) \space b_i=\frac{b_{ri}}{a_{ri}} $, and then the average for the block: $ (8) \space b_q=\frac{\sum_i b_i a_j}{\sum_i a_i} \$`. Why not just divide the number of beer shops by the number of apartments, all within the block? The answer is that in many blocks the values will be nonsensial. A block may have houses without a beer shop, and the one next to it may have a beer shop, but no houses. So the former will have bq equal to 0, and the next one will have division by zero error.

The radius-based approach resolvel this problem and smoothens the values.

I hypothesized other approaches, like dividing consumers between beer shops, or other ways, but these calculations turned out to depend not only on how many consumers there are, but on how they are divided between buildings (1000 apartments in 1 house make not the same values as 1000 split in 10 houses).

## Coffeeshops Inside and Outside Malls

I was worried that many coffeeshops are concentrated inside shopping malls. European new urbanists criticize malls for killing the street life in the cities. Judging by the number of coffeeshops, a mall looks like a city center on a map, although it isn’t.

I tried making a map of only those coffeeshops that are in the streets, and not in malls. To remove them, I tried three approaches, and here’s the first one: keep only those coffeeshops that have no others in the same coordinates, but do have other neighbors in about 200-250 metres.

Can we trust this approach to find the real city center, or local centers? This filter actually kept two coffeeshops at railway stations, in 200 metres one from another, and between them there’s a rather repelling tunnel passage under railways. This is definitely not a city center.

Other approaches to filter the real street coffeeshops did not produce any satisfactory result:

• Approach 1, to remove coffeeshops that have no neighbors or only neighbors in the same building (which usually is a mall or a department store), is not working everywhere. In Novosibirsk city, it seemed to work, but failed in many other places. In Kuznetskiy Most street in Moscow, there are perimetral buildings that have 5 or more coffeeshops, and the algorithm marks them as being in a mall. To remove these false-positives, we could try increasing the threshold to more than 5. But according to statistics, in all the Russia, there are just several malls with 6 or more coffeeshops, the rest of malls would become false-negatives then.

• Approach 2, we check if a coffeeshop has other coffeeshops farther than 10-20 metres, but closer than 250 metres. If there are none, it’s an outlier or it’s inside a mall. Unfortunately, this method is even less stable to random fluctuations. In Togliatti, there’s shopping mall with some coffeeshops. They’d be classified correctly, if there weren’t a takeaway coffeeshop near a mall. This way, every coffeeshop inside the mall has a neighbor in the correct range, and is classified as a street one. I can’t even guess how many false positives or false negatives it produces.

Why in the first place did I want to distinguish these places? If we try separating malls from non-malls because the former is an artificial place, then what are modern pedestrian streets? What are modernistic streets with houses standing randomly? Both streets are to some extent artifical. Hence, if we measure characteristics, and how many coffeeshop they host, we should measure all of them.

## Variety of Architecture Types

Daria suggested that in active streets, new buildings were always added over time, and their age dispersion is bigger than that of modernist megablocks. This means that diffusion of beer shops and coffeeshops should differ in these two kinds of blocks.

No trend can be seen in this chart, so this hypothesis was left.

## Heatmaps of Coffeeshops

Heatmaps were the initial plotting tool, and but as the research progressed, lost their importance.

What is curious, is that the city center can indeed be seen, though we have no statistics to check for any correlation.

• In Novosibirsk, there are 5 centers and 3 malls in the central part.
• In Chelyabinsk, the center is smaller, and outside of it, all coffeeshops are located in malls.
• In Yekaterinburg and Perm, big malls exist, but the central part of the city, where coffeeshops are distributed evenly, is bigger than in Novosibirsk or Chelyabinsk.
• In Rostov-on-Don, despite a regular blocks grid, the “coffee center” is only in a half of the central part, and there’s little activity beyond the railroad that surrounds the center.
• In Volgograd, essentially a conglomerate of factory settlements put along Volga river, there’s one main center, and a local one, and the rest of the territory has only beer-on-tap shops.
• Moscow and Saint Petersburg, being 3-8 times biger than the third-largest Novosibirsk city, have a very big center with lots of coffeeshops, also fuelled by foreign tourists. But outside of the central historic parts, both have very few coffeeshops, mostly right at the subway stations. In Moscow, there are more coffeeshops in the South-West, where university and lots of students live, and more beer-on-tap shops in the South-East, that consists of big apartment buildings and hosts low-income immigrants.

The main conclusion of these maps is this: the wider the regular grid of streets is, the more center there’s in the city. Center in Yekaterinburg (1.4M people) and Perm (1.0M people) spreads more than in the respectively equal-sized Novosibirsk (1.5M) and Chelyabinsk (1.1M). This is still a hypothesis that needs a strict mathematical proff.

Building the cities around subway and highways as in Moscow and Saint Petersburg, led to creation of huge dull areas, deficient micro-cities growig around life-supporting subway station.

# Positive Results

As it was mentioned in section Problem Statement, the key hypothesis was that the plan of neighborhood affects life style, hence the demand for beer, and the number of beer-on-tap shops. There’s no classification of plan types yet, so I had to resort to their indirect properties:

• Construction year
• Construction epoch
• Block area
• Transit travel time from the center

To confirm the hypothesis, I had to find a correlation between some of these and beer shops diffusion (it should be positive correlation in all cases: newer buildings, bigger area, farther from the center). If no correlation is observed, then I’d have to admit a mistake.

First, I describe a refuted correlation:

## Block Size and Time from The Center

The central areas of Russian cities were inherited from the imperial period, when cities were built in the same traditions as in Europe, with active facades and dense grid of streets. In Yekaterinburg and Novosibirsk, blocks were relatively big, 125 by 250 metres (3.125 ha). In 1920-s and 30-s, planners of USSR decided to increase blocks size, uniting those that existed and closing some streets. These blocks have area of 7.25 ha. During Khruschov period, planners started building blocks like in Le Corbusier’s “luminous city” concept, with megablocks of 20-30 ha, and later they grew even bigger.

As new urbanists mention, these modernist blocks lack any street life. My theory is that these blocks should have had more beer shops. On the opposite, the small central areas have much more life in them, hence more coffeeshops.

I plotted houses, grouped by their block size, on a chart, and there’s no noteable trend. The only exception is that the dominant blocks of the center produce spikes in coffeeshops line.

In Krasnoyarsk, beer shops diffusion grows with block size, but grocery stores have the same trend too, which dose not prove the main hypothesis.

Same way, values on 2 axes: area and commute time from the center, produce some sort of trend only in some cities (Irkutsk, Krasnoyarsk, Nizhniy Novgorod, Omsk, Rostov-on-Don, Togliatti, Volgograd).

In Novosibirsk, blocks of 3 and 7 ha have a high amount of coffeeshops, simply because blocks of this size are mostly in the center.

No other correlation with block size has been found.

## Construction Year

Depending on the year the house was constructed, the diffusion of retail objects differs, with different trends over years. Apartment buildings of 2010-s have noteatbly more beer-on-tap shops per apartment than those built in other epochs. Older houses don’t have clear trends regarding beer shops.

Coffeeshops are much more popular in imperial and early soviet epochs. In houses of recent years, they are almost absent, except for 2000-s, when (except Moscow) new bulidings were erected near city centers.

## Commute Time from The Center

One of the hypothesis was that living far from most of the city destinations, stuck in the microdistrict (the Soviet term for modernist superblocks) on the fringe, citizen travels rarely. As many new urbanists point out, the Moscow periphery dwellers get so tired of commuting that on saturday and sunday they just stay idling at home. Probably, some of them have beer while doing so.

That’s why I introduced the commute time. For every house, I measure time to commute to the center by public transit (in Moscow, in 2017 the share of public transit was 70%. In other cities it may be lower, but the research of Gorod.PRO showed roughly the same amount in Ekaterinburg.)

As mentioned above, the distance of commute to the center is calculated for each apartment building. The center of each city is chosen arbitrarily using the experience from the previous works. (The coordinates of the centers can be seen in the maps as the star icons.)

In Moscow, the new periphery has clearly more beer shops per person. The streetscape there clearly has low quality. Any trip to any place takes significant amount of time. For instance, in Maryino, cultural centers are located at least in 1 kilometer from subway stations, which isolates them from being a center for the neighborhood.

Near the center, in particular between the Garden ring and the Third Transport ring, streetscape is also poor, but surprisingly the beer shops diffusion value is low. Starting the research I was trying to find a more straightforward dependency. The reason is probably that it’s quite expensive place, and the residents belong to upper classes or the new urban creative class, hence they either can afford or purposefully try to find themselves entertainment.

Another important observation: tn the other cities, the peak of beer shops is roughly where the majority of apartments are located.

Half of the cities show the common trend: more beershops in the periphery built in 2010s. Some other cities, though, show chaotic distribution, or even the opposite: more BD is near the center.

## Construction Epoch and Center Commute Time

This approach gave the most expressive results.

As mentioned above, construction years were grouped in epochs, in which typology was distinct from those of the neighbor periods. (See “Formulae”). The epochs are these: imperial (up to 1919), early Soviet (1920-1932), Stalin rule (1933-1955), “Khruschovkas” (1956-1969), late Soviet (1970-1992), 1990-s (1993-1999), 2000-s (2000-2009), 2010-s (from 2010).

The other axis is time to commute to the center. It was rounded up to multiple of 5 minutes.

Diffusion values were averaged as usual, weighed by apartments.

The results are in the following charts.

As you can see, beershops diffusion in higher in the upper-right corners of these charts. Beer shops are abundant in the neighborhoods of the last epoch. The trend is observed in 12 of the 20 cities researched: Ekaterinburg, Irkutsk, Krasnoyarsk, Moscow, Nizhniy Novgorod, Omsk, Rostov-on-Don, Saint Petersburg, Tver, Tomsk, Tyumen and Voronezh.

There are some dotted neighborhoods closer to the centers and of older epochs that have higher BD, those are usually squalid residential complexes that are undermaintained, and where residents are now only poor, other residents fled. Apart from these, other neighborhoods follow the trend line.

The trend is not observed in Chelyabinsk and Chita, Kazan and Ufa (these probably have different consumption habits due to turkic traditions, Muslim community and an influx of new residents from rural areas), Krasnodar, Novosibirsk (probably due to the presence of the manufacturer and franchiser of beer taps equipment), Perm, Togliatti (almost uniformly problematic city), Volgograd (a city with a very unusual linear shape and a very small center).

Houses of the imperial and the Stalin epochs look saner than those of other epochs equally far (by commute time) from the center.

## Traditional Urban Streets and Coffeeshops

One of the two big goals of this research was to prove that traditional urban streets are better than modernistic ones.

The problem with the old-style streets in Russia is that they’re rare. Only few cities do have such old historic center, but it makes no sense to compare the activity in the center with activity elsewhere. No way we can substract the effect of the center.

To make a point, I had to compare buildings equally distant from the center (again, by commute time). This was doable in 2 cities: Nizhniy Novgorod and Saint Petersburg. See the following charts and notice that indeed in any column (equal commute time), older epochs have more coffeeshops than the newer ones.

# Consclusions

1. Arkadiy Gerschmann was right in his observation. Residential buildings of 2010-s have much more beer shops per inhabitant and per apartment, than those in the older parts of the cities. A counterargument that commerce is easier in those houses, is not true, because there are less grocery stores and pharmacies. What is the real mechanism, is an open question to sociologists, but there’s a reason to worry.
2. Separately, the commute time and the epoch of the building aren’t strong factors. Equally far houses may have different diffusion of retail stores.
3. Modernist urbanism perporms worse as city center: the imperial and early soviet epochs gave houses that have more coffeeshops, than those of later epochs (I compare houses equally distant from the center).
4. Cities that have bigger uninterrupted grid of streets have more coffeeshops spread around the center. Examples are Perm, Chita and Nizhni Novgorod. (This observation still requires a thorough check.)
5. It was expected that the bigger modernist blocks have more beershops, but this turned out not true.

Overall, the research reached its goal and proved the hypothesis, that urban planning affects lifestyle and habits, even though using indirect indicators. Direct indicators should be used further, for example, houses should be classified by their planning type: perimetral, semi-perimetral or free-standing modernist. CityClass project already implemented this to classify territories.

The traditional urbanism is still being re-evaluated in Russia, here’s an example of Alexander Lozhkin’s Essay on cityscape #7: tradition framework:

Of all human urbanism models, only one turned out to be self-sustaining and comfortable, and that is the historic city model, because it wasn’t invented but gained through a lot of pain. The search of newer models started only when the traditional model failed to cope with hyperurbanisation. But this search gained nothing.

Read my and Gerschmann’s comments on the research. Also check out charts and maps.

Dmitri Lebedev, september 2016-may 2017.

A big thanks to Daria Kiselnikova, Alexander Lozhkin and Dmitriy Oschepkov.