General Urban Decline and Suburban Growth

In the second half of the nineteenth century, urban infrastructure in the city of Dublin decayed, public health declined, and the quality of city life fell substantially. The social history literature investigates why Dublin suffered such a downturn throughout the nineteenth century and beyond, and in doing so, it logically asks why there was a large movement of wealthier classes away from the city. Dubliners have been known to simply blame The Act of Union 1800 for their city’s failure to thrive both at the time and still today (see O’Brien 1910, p. 246; Dickson Reference Dickson2001, p. 114).Footnote ³ However, academic studies on the urban decline in Dublin instead pinpoint the initial cause of Dublin’s decline to the city’s failure to industrialize its economy, leading to limited employment opportunities (Craft Reference Craft1970, p. 307). Despite the lack of jobs, the city was a magnet for unskilled rural workers, particularly in years of poor agricultural yield (Ó Gráda Reference Gráda1995, p. 250). As a result, Dublin was increasingly becoming a city populated by transient general laborers at times of downturns (Craft Reference Craft1970, p. 309), which was emphasized by the fact that Dublin was Ireland’s principal port for imports/exports related to the agriculture industry (Daly Reference Daly1984, p. 54).

At the same time, following a growing international trend, upper-class city dwellers relocated to the fresh and open environs outside of municipal boundaries: the “garden suburbs” (Harris and Larkham 1999, ch. 4). The innovation in improving modes of transportation made living in the suburbs and working in the city feasible, with horse-drawn buses and trams linking the suburbs to the city by the 1870s (Craft Reference Craft1970; Corcoran Reference Corcoran2000). Suburban growth and the related city decline peaked in the latter decades of the nineteenth century. To illustrate these contrasting trends, Figure 1(a) gives population figures for the city, two selected wards of the city (Inns Quay and Rotunda), and three surrounding townships (Clontarf, Drumcondra, and Rathmines) in the two census years 1891 and 1911. The areas displayed in the figure are those that later will become part of our empirical analysis. Inns Quay and Rotunda [CITY] will be the wards representing the city; Clontarf and Drumcondra [AT] the areas representing the annexed townships; and ultimately, Rathmines [IT] an area that was not annexed (independent). The figure shows that the whole city of Dublin city’s population grew by 24 percent between 1891 and 1911, mainly due to the annexation that occurred in 1901; without the annexation, the city remained relatively stagnant, Inns Quay and Rotunda grew mildly (about 10 percent), indicating that some wards in the city grew less. By contrast, the suburbs (we interchangeably use the terms suburb and township as they are both widely used in the related literature) grew substantially, by about 35 percent in Rathmines and 80 percent in Clontarf and Drumcondra.

Figure 1

AGGREGATE SUMMARY STATISTICS

Notes: *From 1901 Dublin City’s population is inclusive of the annexed townships Clontarf, Drumcondra, and Kilmainham. The population of this latter was 8,783 in 1901 and 9,630 in 1911. **Due to information limitation, the year 1891 is the actual year 1881 for The City and it is the year 1894 for the townships; the year 1911 is the actual year 1910. Note that in 1911 Clontarf and Drumcondra have the same tax rate as the city due to annexation. ***[CITY]=Relevant part of the city; [AT]=Annexed townships; [IT]=independent township.

Sources: Irish censuses reports for Panels (a), (c), and (d). Panel (b) from Parliamentary Papers c. 2827, p. 4, c. 7795, p. 56, and cd. 5091.

There was also a drastic change in the city’s occupational structure relative to the suburbs (see Table 1). According to the aggregate census data, in 1891, the suburban districts had 40.2 percent of their population in professional- and middle-class occupations, while Dublin only had 22.5 percent. Table 1 presents an index for the representation of different professional classes in the city as compared to the suburbs in 1881 and 1911 and shows that the suburbs were generally becoming wealthier, while the city was experiencing the opposite (Daly Reference Daly1984, p. 146).

Table 1

INDEX OF PROFESSIONAL OCCUPATIONS

Notes: †Suburbs include, Blackrock, Clonliffe, Clontarf, Drumcondra, Glasnevin, Glasthule, Kilmainham, Kingstown, Pembroke, Rathgar, and Rathmines. Clontarf and Drumcondra are part of the city in 1911.

Source: Daly (Reference Daly1984, p. 146).

Worsening public health provisions further exacerbated these trends. With the lack of a main drainage system and the overcrowding of the poor in the tenement slums spreading throughout the city, disease was a serious issue. Table 2 illustrates this by showing high mortality in the city due to zymotic diseases that were thought to be particularly associated with poverty at that time (33.6 per mille, compared with lower proportions in all suburbs—as documented in The Housing Report of 1914, PP cd.7317).

Table 2

DIFFERENTIAL MORTALITY RATES

Notes: Deaths per 1,000 living in population. The table also compares mortality rates due to “zymotic diseases,” which were thought to be particularly associated with poverty at that time.

Source: PP cd.243.

Political Control and Taxation

Over the second half of the nineteenth century, culminating with The Local Government Act of 1898, there was an immense structural transformation occurring at the local government level in Ireland, shifting the balance of power and politics in Dublin. Changes were occurring on multiple levels. New legislation gave Dublin Corporation a wider scope of responsibilities and powers, particularly in taxation and spending (O’Brien Reference O’Brien1982, p. 74). Additionally, the right to vote and to hold office was extended to many citizens, previously prohibited, transforming party politics within the Council. By 1898, property qualifications were fully removed from the regulations for holding office and the municipal franchise was democratized—widening to 38,000 electors versus the mere 8,000 allowed to vote prior.Footnote ⁴ Accordingly, Dublin Corporation became increasingly Nationalist (used interchangeably with Catholic from here forward), and, parallel to this, control was being handed over to the “small ratepayers” (Pim Reference Pim1875, p. 114)—the so-called “petty business classes” (Daly Reference Daly1984, p. 207). By 1898, Unionist (used interchangeably with Protestant from here forward) representation had shrunk to a mere 11 representatives (out of more than 60).Footnote ⁵ It was a dramatic shift from the turn of the nineteenth century when Catholics (and those with marginal wealth) were not permitted to stand in an election, and even as late as 1840, when members of the city council continued to be exclusively Protestant in religion and in favor of the Union (O’Brien Reference O’Brien1982, p. 35).

Accordingly, the townships surrounding Dublin held a Unionist majority. Daly (Reference Daly1984) writes “Like Rathmines, Clontarf retained its strong Unionist complexion until the early twentieth century.” “The prosperous Protestant upper-class, rather than fight this issue in the cockpit of municipal politics withdrew to their comfortable suburbs where they pursued separate interests and separate development” (Daly Reference Daly1984, p. 17).

Wealthy and Protestant residents likely held contrasting values to the new Dublin Corporation on taxation and provision of local public goods (O’Brien Reference O’Brien1982). In 1874, Dublin’s property tax was 45 percent, whereas it was much lower in the suburbs (e.g., 23 percent and 20 percent in Clontarf and Rathmines, respectively, as shown by (Daly Reference Daly1984, p. 227)).Footnote ⁶

After 1900, city taxation was consistently above 50 percent. Figure 1(b) illustrates the total taxes in the city versus the suburbs in 1891 and 1911. We observe a sizable rise in the tax rates of Clontarf and Drumcondra after their annexation, and to a lesser extent in the independent Rathmines, between the two time periods. High property tax rates affected mostly owners of large properties (wealthy residents).

The way that taxes were spent is another important determinant of the decision of certain households to move. Possibly, this affected the decision of Protestants more as, potentially, they saw this as part of the lost political control (both O’Brien (Reference O’Brien1982) and Daly (Reference Daly1984) make this point). The annual expenditure on the relief of the poor in Dublin had become unsustainable, presenting an enormous burden on taxpayers (O’Brien Reference O’Brien1982, p. 74).Footnote ⁷ Along with the high cost of maintaining Dublin’s poor houses, the city also bore a disproportionate burden in terms of hospitals (including morgues and disinfecting chambers), alms-houses, policing, fire brigade, schooling, and improvement of tenements and thoroughfare roads (The Housing Report of 1914, PP c.2827). Taxes would only cover some of the excess costs faced by the Dublin Corporation, partly because not all users of public services contributed to taxation.Footnote ⁸ Borrowing covered the deficit, leading to the extreme indebtedness of Dublin Corporation, with a debt (£1,399,655) to revenue (£410,009) ratio of 3.41 in 1894 (PP 7795). The annexation reduced the debt (£1,876,987) to revenue (£655,395) ratio to 2.86 in 1901 (2.88 in 1904).Footnote ⁹

The Annexation of 1901

In response to the worsening of its finances and accumulation of debt, but also to correct the inequity of free-riding on public services by suburban residents (O’Brien Reference O’Brien1982, p. 82), in the latter half of the nineteenth century, Dublin Corporation initiated a movement to annex the surrounding townships into the city. The proposed legislation was expected to add 80,000, mainly middle-class, residents (around 30 percent of Dublin’s population) to the city and £350,000 to its valuation in taxable property—an increase of 50 percent. Not only did it mean an enlargement in the tax base for Dublin, but with expanded collateral, the city’s borrowing power would also be enhanced.

Township residents adamantly protested the possibility of annexation. They had a sense of independence and hostility toward the city and its administration.Footnote ¹⁰ A letter from township commissioners to the Lord Lieutenant defended their autonomy (Chief Secretary’s Office Registered Papers, 1885, no. 7695):

Rathmines and Drumcondra provide a useful case study on public opinion, as only the latter was eventually annexed. Circulars were sent out by township commissioners to gauge opinion on annexation. In Rathmines, 90 percent of residents were against amalgamation and the objection was even more marked in Drumcondra (PP c.2827, p. 19).

In 1900, legislation was laid down in The Dublin Corporation Act for the annexation of the highly professional and Protestant northern townships of Clontarf, Drumcondra, and Glasnevin, and the area west of the city known as Kilmainham. The Act was given Royal Assent in August of 1900 and took effect on 15 January 1901. At the last moment, the southern townships of Rathmines and Pembroke managed to be excluded from this act and were not annexed to Dublin until 1930.

The reasons why Rathmines and Pembroke were able to retain their independence differed between the two townships. Rathmines had low property tax rates, reduced public services, and a low-quality water scheme. The township commissioners saw the initial offer of accessing the River Vartry water by Dublin Corporation as too expensive and opted for polluted and low-pressure water from the canals. However, even when Dublin Corporation offered access to Vartry water at favorable rates, the commissioners rejected the offer for lower quality water from the river Dodder. Behind these decisions was the fear that accepting the deal with Dublin Corporation would have meant losing independence. The fact that the township maintained a majority of Protestants—many having escaped the city due to hostility toward Catholicism and Nationalism—played a pivotal role in this decision. Pembroke had a different structure. The Earl of Pembroke owned most of the township but did not reside there. The policies implemented were often favorable to his properties, but he would pay them back with generous financial benefits. The township faced a large number of sanitation problems. Upon pressure from its residents, it responded by agreeing to use Vartry’s water from the Dublin Corporation and increasing sanitation expenditure in the city, while remaining independent from the city. The township opted not to raise its property rates but, instead, to accumulate debt, eventually guaranteed by the Earl and wealthy residents. The debt accumulated made it less appealing to the city, and the presence of control by wealthy Unionists meant it preferred to retain its independence (for further details, see Daly (Reference Daly1984)).

The boundary extension occurred in the same year as the enumeration of the 1901 census. Its passing was quite unexpected, as numerous bills had been brought to the table over the previous 20 years and failed (Irish Times, 18 January 1900). It is possible that some residents, anticipating that their taxes would go up, would have moved before the taking of the census. However, equally, it is also likely that many did not foresee the annexation. In fact, in 1898, just two years before its passing, a similar annexation bill was taken before Parliament at great expense, with no success (Irish Times, 4 January 1900). In 1900, The Irish Times wrote that the passing of the Dublin Corporation Boundaries Bill at any time in the future was an “improbable event” (18 January 1900).

People living in the annexed townships had to pay more than double the tax rate that they had paid just before the annexation to a municipal government that they had not chosen to live under. As can be seen in Figure 1(b), over the years before annexation, those townships that were eventually annexed paid, on average, approximately the same amount in taxes as those townships that were to remain independent. It can also be seen in this figure that taxes went up in all areas around the turn of the twentieth century, which was largely the result of a need for funding for the elaborate infrastructural projects (mainly drainage) that were taking place at the time. However, rates went up much more dramatically for the annexed townships than for other areas, and these annexed areas were now paying taxes to fund the local public goods of a government that they did not elect to live under.

Many of the larger property owners had threatened to leave as the result of a successful annexation (Irish Times, 11 July 1899). Aggregate statistics on the townships that became part of the city indicate that certain groups of residents may have been “voting with their feet” against living under the imposed municipal jurisdiction of Dublin. Figure 1(c) shows that there was a larger percentage drop in the number of Protestants in the annexed townships between 1891 and 1911 than in those townships that remained independent. Finally, Figure 1(d) offers an indication of the relocation of wealth (proxied by the percentage of workers in top jobs), showing that the suburbs of Clontarf and Drumcondra became poorer, while minimal variation took place in Rathmines (and the city).Footnote ¹¹

Sources and Sample

The effect of the Dublin annexation took hold between the only two years, 1901 and 1911, for which the records of the Irish Census are available to the public in their entirety.Footnote ¹³ We employ the returns of these Irish censuses to construct a sample of household-level data for houses located in three distinct areas of Dublin, corresponding to those defined earlier and described later.Footnote ¹⁴ We select sample locations based on their distance from the original border of the city of Dublin. We focus on households near the boundaries of our areas of interest to minimize confounding factors. The sample streets and houses meet the criterion of being in existence (fully built-up) in both the 1901 and 1911 censuses. For each year, we collected data on households residing in houses on the edge of the Inns Quay and Rotunda wards of Dublin, which had long been a part of the municipality but which are near the city/suburbs border [CITY]; houses in a location bordering the city in Rathmines, a township just south of the border of Dublin that was not annexed in 1901 ([IT]), and houses in another location bordering the city in Drumcondra and Clontarf, two townships north of the border of Dublin that were annexed in 1901 ([AT]). See Online Appendix B for a more detailed discussion. So, for the selected wards, only streets either bordering or near the border with the city have been chosen to be part of the dataset. These streets represent 16–23 percent of the population of the wards in which they are in and, for these streets, we have information on all households.

The Dublin metropolitan map in Figure 2 shows where the areas described are physically located. Ideally, we would have liked to compare annexed versus non-annexed townships that were physically next to one another, but due to the geography of Dublin, and the fact that only those townships to the north (Clontarf, Drumcondra, Glasnevin) and west of the city (Kilmainham) were annexed, this was not possible.Footnote ¹⁵ As mentioned, all samples examined are directly next to the jurisdictional boundary of Dublin. Homes on the fringe of the Dublin metropolitan border were generally built in a single era (the mid-to late-nineteenth century) and, as such, would have experienced an equal rate of natural decay. Moreover, all areas had similar modes of transportation in place for the period studied: by the 1870s Rathmines, Drumcondra, and Clontarf had horse-drawn tramlines along their main streets leading towards the city center of Dublin (Corcoran Reference Corcoran2000). The regulated charge for a single trip from the city center to these townships was equally one shilling (Craft Reference Craft1970).

For each household in the dataset, we determine whether it is a “stayer,” a “mover-out,” or a “mover-in” by manually finding the corresponding houses and the residing individuals across the 1901 and 1911 censuses. A household is a stayer if at least one family member lives on the same street in both census years (first, middle, and last name, and age match). It is a mover-out if all family members are only on that street in 1901 and a mover-in if all family members are only on that street in 1911 (see Online Appendix C for a more detailed description of this process, including how we addressed the difficulties such as handwritten returns and varying house numbers).Footnote ¹⁶

The individual returns of the census contain detailed information on each member of the household. For this, we can extract the following variables: religion, occupation, birthplace, age, number of children under 14, number of children 14 and over, and number of non-family members living in the household. These can be thought of as potential control variables for residential choice. The birthplace variable is aggregated into six categories based upon whether the individual is originally from Dublin City, County Dublin, another area of Leinster (the province in which Dublin is located), wider Ireland, the United Kingdom, or outside the British Isles. For additional members of the household, we divide the number of children into two age groups (0–14 and >14) based on the view that families with school-age children are bound to behave differently than those with working-age children when making a residential choice.Footnote ¹⁷ We have also included count indicators for the number of non-family members living in the household, further broken down by whether they are boarders or servants.

For the occupation variable, we first combine similar occupations (some individuals gave elaborate job descriptions on their Census return form) into job titles that could be coded under the “HISCO” (Historical International Standard Coding of Occupations) classification scheme (Van Leeuwen, Maas, and Miles Reference Van Leeuwen, Maas and Miles2002). We convert this classification into a proxy for income using the HIS-CAM scale, producing values that range from 30 to 100 with a standard deviation of 14.71 (in shorthand, we refer to this variable as income). See Online Appendix D for details.

Finally, housing prices have been collected from Thom’s Irish Almanac & Official Directory.Footnote ¹⁸ We have consulted the 1892 and 1912 yearbooks due to their proximity to the periods of interest and their online availability.Footnote ¹⁹ We utilized the Street Directory portion of the publication, with information on street and house numbers, and most importantly, the “rateable valuation” of the property, which represents a proxy of the market value of each property, net of property tax.

Descriptive Statistics

Our sample consists of 3,257 households over two time periods, 1901 and 1911, with a total of 15,896 individuals. Table 3 indicates that for 1901, there are 1,516 households in the entire sample, and for 1911, this figure rises to 1,741. As mentioned, in the census data, all houses were observed each year. The increase in households across census years is put down to the fact that more houses were empty in 1901 and that houses were more likely to contain more than one household in 1911.

Table 3

SAMPLE SIZE OF STAYERS AND MOVERS BY LOCATION

Notes: Movers are implicitly outward movers in 1901 and inward movers in 1911. The lack of correspondence between stayers in two periods is often due to a missing head of household in 1901 and a young member of the family becoming head of household in 1911. This table counts the households based on the number of head of households.

Source: Authors’ calculations.

We have been able to collect aggregate information from census returns and analyses by social historians to test how representative our sample is of the population. Figure 1 (discussed earlier) presents aggregate statistics in four charts showing differences between the three areas of interest in terms of population, taxation, religion, and an indicator of wealth—before and after the annexation. For the religion variable, we test the mean of our sample against the mean of the census return (using a t-test of difference in means) and find that they are representative of the population, except for Inns Quay in period one and Drumcondra in both periods. In these three exceptions, Protestants are over-sampled.

The top panel of Table 4 provides a comparison of statistics showing how the areas differed before the annexation. The t-tests for the difference in means are reported in the notes of the table.Footnote ²⁰ We observe that [AT] and [IT] are more similar to one another than to [CITY] before the annexation. This is true in terms of religion, rental prices, and income. [AT] and [IT] are more similar in terms of religion before the annexation, with comparable proportions of their populations reporting as Protestant at 0.41 and 0.51, respectively. Though this difference is statistically different from zero (t-stat 2.95); its significance is well below that of the difference between the two townships and [CITY], where the proportion of Protestants is 0.17, with t-stats exceeding nine. The message from comparing the rental prices (rateable valuation) is much the same, the only difference being that the t-stat of the difference in means between [AT] and [IT] is not significant. Finally, when it comes to income, the difference in average HISCAM, 68.23 in [AT], versus 71.03 in [IT], is well below the difference in average HISCAM, when comparing with 63.37 in [CITY]. For income too, there is a smaller statistical difference between the townships (t-stat 2.53) than there is between the townships and [CITY], where the t-stats are above 6.5. Overall, we find evidence supporting Conjecture 1: The annexed townships [AT] and the independent township [IT] are more similar to one another than to the city [CITY].

Table 4

SAMPLE SUMMARY STATISTICS BY AREA

Notes: Bottom panel (Year 1911) includes whether there is significance difference between means in year 1901 and year 1911.

^a p<0.01, ^b p<0.05, ^c p<0.1. †Accessing 1892 rather than 1901 Thom’s Directory has the disadvantage of creating missing values in 1901. This was particularly the case for Rathmines since four of the streets were built between 1892 and 1901, causing about 200 missing values, which may explain the significant difference in rateable valuation between the two time periods for Rathmines. T-stats of difference in means for the year 1901 (the order is the same of the coefficients in the table, the first is rateable valuation, the second is income, …): ([AT]-[CITY]: 18.14, 6.54, 9.33, 4.21, –3.04, 1.94, 8.67, –2.27, 1.19, –4.56, 3.93, 3.16, 0.33); ([IT]-[CITY]: 8.76, 7.51, 11.02, 4.22, –2.60, 0.10, 12.38, –0.69, 2.69, –2.53, 3.04, 0.26, 1.72); ([IT]-[AT]: 1.51, 2.53, 2.95, 0.41, 0.14, –1.72, 7.08, 1.66, 1.76, 1.69, –0.43, –2.76, 1.50).

Source: Authors’ calculations.

Comparing the top and bottom panels of Table 4 show how the areas changed over time. We observe that heads of households are becoming older, and there are fewer young children, a lower number of servants, and—for most of the locations—fewer boarders. For the religion variable, the proportion reported as Protestant in the annexed townships drops from 0.41 to 0.34 after the annexation, and is highly significant. There is also a decline in the Protestant group between the two periods in the independent township, but it is not statistically significant. As for the other variable of interest, income, we observe that in the annexed townships (and the city), income drops significantly, while it rises (though not significantly) in the independent townships.

Here the statistics are supportive of Conjecture 2: The annexed townships became less wealthy and less Protestant after the annexation, while the independent township remained relatively stable.

Drumcondra versus Clontarf

So far, we have included houses from two of the annexed townships, Drumcondra and Clontarf, grouped to comprise the annexed townships. Each of these two areas was first established as townships in the same ten-year period (1869–1879); Dublin was immediately accessible to them both, as each bordered the city to the north, and each paid a virtually equivalent rate of taxation. Still, since the two areas are geographically and jurisdictionally distinct, we also ran an alternative analysis with Drumcondra and Clontarf as separate entities, to assess whether residential behavior varies across the annexed area (see Online Appendix B for a more detailed discussion). We decompose the annexed townships into [AT1] (Drumcondra) and [AT2] (Clontarf).

The results, displayed in Online Appendix Table B.1, show that said effects are much stronger for the residents of Drumcondra, and on its own, there is no effect of income and religion for Clontarf.

In researching potential explanations for these findings, we find that the two incorporated areas are quite distinct in a major respect. Many of the residents of Clontarf had petitioned for annexation, while every other township was emphatically against it. The Report of the Royal Commission on Municipal Boundaries, published in 1881, provides a telling story of Clontarf’s plight for inclusion in the proposed boundary extension. The report details the investigation by the Commission, in which the Dublin Corporation made it clear that they did not wish to invite Clontarf to join the city. The Clontarf Township Commissioners were themselves against annexation; it was the ratepayers who supported joining the city of Dublin. During the debate, the Commission received a petition from a large number of Clontarf ratepayers. Over 400 people, representing one-fifth of the rateable valuation of the township, stated that they wished to be brought in.Footnote ²⁴ The reason for this support for the annexation simply came down to necessity: despite the prognosis of higher taxation, the public services gained from amalgamating Clontarf with Dublin outweighed its cost. Moreover, historically, Clontarf was associated with poor quality of air/water, prostitution, and polluting industries. This exercise has proven meaningful as a check on whether different attitudes toward annexation between the annexed townships played a role in determining the effects of annexation. The fact that residents in Drumcondra were averse towards the annexation and not many of those in Clontarf has proven a determinant in the decision of certain types of residents to leave the area. A more elaborate discussion of this result is presented in Online Appendix B.

Movers Relative to Stayers

An alternative way to show evidence of Conjecture 2 is by comparing the households that move-in (move-out) with those that remained in [AT] and [IT]. In this way, movers are compared to stayers, with different types of moves being separated into different regressions. Online Appendix Table E.2 shows the estimates of the separate regressions. The richness of this comparison is that, relatively to the stayers, households that move into [AT] are, all else equal, less likely to be wealthy and Protestant, while this effect does not hold for the area that remained independent. Instead, when we shift the analysis to households that moved out, there is not such a difference in terms of religion and income. This is suggestive that the response to annexation is more apparent in terms of who moves into the area than who moves out of the area. We then add the interaction effect between religion and income to study what happens to wealthy Protestants. The results suggest that wealthy Protestants are more likely to relocate to an area that remained independent.

Matching [AT] and [IT] Households

As an additional robustness check, we study what would have happened if the suburbs, instead of just being more similar to one another than to the city, actually had the same demographics. In other words, we study what would have happened if we had a treatment and control group. Based on the same covariates used in the various regressions, for the initial period we calculate the propensity score where the dependent variable takes value one if the observation is from [AT] and zero if it is from [IT], and retain the observations that share common support based on a 5 and 95 percent interval.Footnote ²⁵ For this subsample, for [AT] and [IT], we rerun the estimates documented in Table 6 and find that the effects of income and religion do not change much (see Online Appendix Table E.3), confirming the robustness of our original results.

The Role of Rental Prices

The theoretical model presented in Online Appendix A shows that rental prices are an important determinant of where to locate. We have so far omitted prices from the analysis because in collecting this data we came across non-random missing values. This is particularly the case for the year 1901 and the area of Rathmines since we proxy for 1901 by using the 1892 Thom’s Directory when some of the properties were still under construction. In this section, we add the observed prices and repeat the regressions reported in Tables 5 and 6. The estimated effects of these new regressions are given in Tables E.4 and E.5 in the Online Appendix. The results are much the same and Conjecture 1 holds even more tightly.

We use Thom’s Directory year 1892 as a proxy for the rateable valuation (net of tax prices) in the year 1901; we use Thom’s Directory year 1912 as a proxy for the year 1911, as these are the closest directories available online. For both periods, we have filled in the values using the Directories if they were listed. For houses that have been divided into sections, we divided the price by the number of sections. For example, if house 3 was valued at £10 and was recorded as divided into houses 3.1 and 3.2 in the 1911 census, we would have divided the registered value by two and valued them as £5 each. If the value for a house is not listed in one Directory but is in the other, we assumed that the value had not changed and replaced the missing information with the available value.

A comparison of the average house price in the three areas in 1892 suggests that prices in the city were far lower than in both the townships to be annexed and the township that remained independent (£9.41 [CITY], £17.25 [AT], and £18.92 [IT]). Also, we observe that the average price dropped in 1912 relative to 1892 in [CITY] (from £9.41 to £7.9) and in [AT] (from £17.25 to £16.42) but rose in [IT] (from £18.92 to £23.33). The positive change in prices in [IT] is what we expected in line with our analysis. However, it must also be considered that new houses were built in the independent township and those may have contributed to the large rise in the rateable valuation. As the use of the 1892 Directory to proxy for 1901 prices meant that some houses in our data set had not yet been developed (particularly in Rathmines), we are unable to directly compare the rateable valuation by street and house number in Rathmines. The only possible comparison is for houses that are inhabited by stayers (as we can assume that they were in the same houses even if the number had changed); for this group of houses, prices have increased from an average of £21.91 in 1892 to £29.43 in 1912, which is quite dramatic. Finally, Figure 4 shows how the empirical distribution of rateable valuation has changed during the two time periods and three areas. [CITY] and [AT] share a similar pattern, while this is not the case for [IT].

Figure 4

RATEABLE VALUATION (NET PRICE) DISTRIBUTION IN 1901 AND 1911 BY AREA

Notes: Kernel density of rateable valuation for houses built in 1892 (proxy for 1901) and 1912 (proxy for 1911) in [CITY], [AT], and [IT].

Source: Authors’ calculation from Thom’s Directory original data.

Disregarding Middle Names in Defining Stayers and Movers

Census data do not have identifiers to match households over time. Additionally, several houses changed numbers between the 1901 and 1911 censuses, making the matching of houses non-reliable. In this situation, we opted to match households using information on household members, the street where they reside, their full name, and their age between the two periods.

As a robustness check on our method of finding corresponding names across the two time periods to identify stayers, we inspect how reliable this technique is by using a shorter version of the name without a middle name (24 percent of the sample has a middle name). Thus, omitting the middle name can, potentially, increase the number of households that we record as stayers. For example, if there was a Mary Frances Delaney age 30 in 1901 on a street, and a Mary Theresa Delaney, age 40 in 1911, on the same street, we would mistakenly record Mary Frances Delaney’s family as stayers. Alternatively, if Mary Frances Delaney reports her full name in one census but only reports Mary Delaney in the other census, then we will have mistakenly recorded her as a mover in/out when she is actually a stayer.

This modified pairing method with no middle names identifies the same stayers apart from one observation. Though this robustness check confirms the number of stayers (and leavers) in our samples, we cannot fully rule out that we may identify as stayers, households with members having the same name and age and living in the same street between the two time periods but being different households. There are very few instances of stayers that have the same name, surname, and age but live in different house numbers between the two periods. Since this feature happens only on certain streets, and because house numbers have changed on some streets between the two periods, our pairing seems to be superior to one where we rely on house numbers being the same between the two time periods.