Crisis, response and distributional impact: the case of Ireland

Ireland is one of the countries most severely affected by the Great Recession. National income fell by more than 10 per cent between 2007 and 2012, as a result of the bursting of a remarkable property bubble, an exceptionally severe banking crisis, and deep fiscal adjustment. The crisis culminated in a bail-out being required from the EU and IMF as the state’s borrowing costs on the financial markets became unsustainable. Ireland’s response to the crisis, as part of an Economic Adjustment Programme agreed at the time of the bail-out, is of particular interest, being widely seen as a "test case" for what is often described as the austerity approach. This paper focuses on the income distribution consequences of the crisis and of the state’s response, rather than on its merits or otherwise as a macroeconomic strategy: these consequences will be an important consideration in any overall assessment, and of relevance to other countries undergoing stagnation and fiscal ‘correction’. To analyse these income distribution effects, we make use of the latest available microdata, notably the 2011 round of the Survey on Income and Living Conditions (SILC) and the SWITCH tax-benefit model (Callan et al., 2012).

The macroeconomic and labour market context in which income inequality trends have occurred, and central features of the fiscal policy response, are summarised in Context and policy measures section. Key elements include a rise in unemployment from about 4 per cent to 14 per cent; sharp rises in taxation; reductions and restrictions on welfare payments; and progressively structured reductions in public sector pay. Income inequality, 2008 to 2011 section sets out the basic results on income distribution over the period 2008 to 2011, during which most of the fall in average incomes occurred. Our exploration of changes in income inequality over this period has three main elements examined in the succeeding sections.

In Methodological issues section, we consider some important issues concerning data and methods, which could have a bearing on the results. These include the treatment of lump-sum payments which became more common (especially in the public sector) due to early retirement and redundancy schemes. We also discuss the issue of representativeness of the survey data on income distribution, using the income distribution statistics as reported by the Revenue authorities as an external check.

Decomposing income inequality levels and trends section uses decompositions of income inequality by income source and by population sub-group to identify some factors behind the observed changes, and to guide further investigation. The role of taxes, transfers and public sector pay policies section explores how much of the total change in income inequality is due to changes in tax and transfer policy, and how much is due to changes in market incomes–including the loss of income for those becoming unemployed. Finally, Conclusions section sets out key conclusions and priorities for further analysis.

The focus of this paper is on the impact of the Great Recession on Ireland’s income distribution–both directly and via the way policy has responded¹. Annual micro-data on the Irish income distribution have been available since 1994–but during most of this period the economy was growing strongly. During the recessions of the 1970s and 1980s, microdata on the income distribution was not available on an annual basis making it difficult to track the impact of recession and associated policy responses on the evolution of income distribution. Nolan (1999) examined the distribution of household income in Ireland in 1987, when Ireland was experiencing a recession, and 1994, when Ireland was emerging from recession. Between these years, he found that the bottom decile slightly increased its share of equivalised household disposable income at the expense of the middle and top deciles. The Gini coefficient, however, was largely unaffected, remaining at 0.32 in both years. O’Neill and Sweetman (2001) came to a similar conclusion, irrespective of whether income or expenditure was used as a measure of resources.

A recent international study by Jenkins et al. (2013) conducted a large scale comparison of the effect of the early years of the Great Recession on 21 OECD countries. Although the recession was the deepest macroeconomic downturn since World War II, and there was substantial heterogeneity in the depth of the recession across countries, Jenkins et al. showed that for most of the countries studied, there was little change in household income distributions in the first two years of the downturn. In addition, the report presented some evidence from six detailed case study countries that elderly people were relatively well protected over the first two years of the Great Recession. Perri and Steinberg (2012) explored the impact of the Great Recession on economic inequality in the United States. They found that the lowest quintile fared worst in terms of earnings and wealth, losing about 30 per cent of earnings and 40 per cent of wealth. However, they suggest the tax and transfer system offset most of these losses for the bottom quintile, so that expenditure on nondurables for this lowest 20 per cent did not change significantly relative to other groups. Similarly, Larrimore et al. (2013) found an "unprecedented importance" of the direct effects of temporary tax and transfer policies for supporting median and bottom-quintile income during the Great Recession in the US.

In order to understand the context in which the recessionary forces were operating in Ireland, it is necessary to keep in mind the broader backdrop. Over the years 1994 to 2007, economic growth in Ireland was among the highest in the OECD (see Figure 1 for the rise in GDP per capita). The period 1994 to 2000 saw an annual average growth rate in real GDP of over 7%. This growth was accompanied by sustained increases in the numbers in employment, rising from 1.2 million in 1994 to 2.1 million by 2007. Unemployment fell to just over 4% in 2000 and remained around this level until 2008 (Figure 1). Ireland had long been a country of net emigration, but this trend reversed as significant numbers of Irish emigrants returned and immigrants from other countries were attracted to Ireland.

Ireland’s economy entered recession in 2008, and by 2010 GDP per capita had fallen by more than 13 per cent, while unemployment soared to almost 14 per cent. This scale of economic deterioration was driven by three main factors:

Each of these factors contributed to a fiscal crisis, with tax revenues collapsing while increased unemployment led to greater demands on the welfare system. The banking crisis resulted in the government guaranteeing both investors and bondholders and led to unsustainable yields on Irish bonds as government debt grew. These unsustainable yields led to the Irish government seeking a financial ‘bailout’ from the EU, the ECB and the IMF in 2010.

The nature of the recession, and in particular the severity of the downturn for the construction industry, has contributed to a sharp differential in the evolution of the male and female unemployment rates. Figure 2 shows that the unemployment rates for men and women were similar, at about 4 to 5 per cent, for the years 2003 to 2007. By 2011, the male unemployment rate had risen by 13 percentage points, while the female unemployment rate had risen by about half that much.

What about developments in wages for those in employment? On average, there was a small rise in hourly earnings over the 2008 to 2012 period, but there is a great deal of diversity across sectors. There were falls of 4 to 5 per cent in public administration and defence, and in finance and insurance. At the other extreme, there were increases of 6 per cent for those in manufacturing, and 2 per cent for those in wholesale and retail. Wages in public sector organisations were reduced first by via a ‘Pension Related Deduction’ (PRD), introduced in 2009, whereby the first €15,000 of annual earnings were exempt, but a 5% levy was paid on the next €5,000 of earnings, 10% paid on earnings between €20,000 and €60,000 and 10.5% on earnings above €60,000. Later, in 2010, an explicit pay cut for public sector workers was implemented, with a reduction of 5% on the first €30,000 of salary, 7.5% on the next €40,000 and 10% reduction on the next €55,000. New entrants were also to be hired on salaries 10% lower than the level payable to current staff. The evolution of average wages in the public sector has also been affected by compositional shifts. For example, a policy of incentivized early retirement, made available to those aged over 50, may have removed from the payroll more of those with above average wages, thereby depressing average wages. A further complication is that the system of incremental pay scales means that the composition of the workforce as between those (typically older) at the top of their scale and those who would benefit from annual increments can affect the extent of observed increases in pay.

As well as the cuts to public sector pay, the deterioration in the economy and in the government’s fiscal position led to a variety of tax and benefit changes between 2008 and 2011. Looking first at the taxation side:

○ An income levy, payable on gross income (excluding social welfare payments) was introduced in 2009 at an initial rate of 1% on annual income up to €100,100 and 2% on income in excess of that. These income levy rates were subsequently doubled, as was the Health Levy. Both levies were then replaced in 2011 by a "Universal Social Charge" (USC)–a new form of income tax, with exemptions for annual income below €4,004 and a progressive structure above this level with rates of 2%, 4% and 7%.

○ The income ceiling above which no further social insurance contributions were payable was first raised substantially, and then abolished in 2011.

○ In 2011 the standard rate band of income tax was reduced (from an annual €36,400 to €32,800) as were the main tax credits.

○ A €200 per annum charge on non-principal private residences was introduced in 2009 as was a flat-rate ‘household charge’ or property tax of €100 in 2011, both payable by the owner of the property. This was the precursor to a full scale value-related property tax coming into force in mid-2013.

○ Tax relief on pension contributions was also reduced, with the annual earnings limit for determining maximum tax-relievable contributions down from €275,239 in 2008 to €115,000 by 2011, while employee pension contributions also became liable for PRSI and the USC.

On the social welfare side, income support rates were actually increased in 2009. The Budget for that year was brought forward from December to October 2008, and the full scale of the problems was not yet evident. However, the Budgets of 2010 and 2011 then reduced the rates of support provided by most social welfare schemes applicable to those of working age, and made deeper cuts in the universal Child Benefit payment. Payments to young unemployed people were reduced very substantially. Rates of payment for old age pensions, however, have remained unchanged to date.

Whether taking 2007 or 2008 as the end of the bubble/start of the recession, the Gini coefficient is very similar at that point and at the latest available date, 2011–a slight fall from the 2007 level, and a slight rise from the 2008 level. Indeed, over a longer period (1994 to 2009) which includes the strong growth of the Celtic Tiger era, Nolan et al. (2012) show that the Gini coefficient remains in the range 0.31 to 0.32 for almost all years. Against this backdrop, the fall in the Gini to 0.29 in 2009, the first year in which the full effects of the recession were felt, is quite striking: this is the lowest level the Gini has reached in Ireland, by some measure, over the years since 1980. Gini coefficients calculated from the Household Budget Survey remained constant between 2005 and the next time the HBS was undertaken in 2009/10, and are close to those found by SILC.

Against this broad stability over the full period, there were significant shifts on a year-by-year basis. The 2009 pattern confirms the reduction in income inequality suggested by the fall in the Gini: in fact, the 2009 distribution Lorenz-dominates the 2008 distribution³. By contrast, there is an increase in inequality (again meeting the Lorenz dominance criterion) between 2009 and 2010. The Lorenz curves for 2008 and 2011 cross, however, meaning that these distributions cannot be unambiguously ranked in terms of inequality using the Lorenz criterion.

In interpreting these results it is important to recall that comparisons of corresponding deciles in different years are not comparing the incomes of the same people, but are instead comparing what might be termed "income positions" e.g., the incomes of the poorest 10% in each year. Changes in composition (e.g., more of the bottom decile being unemployed, or self-employed with very low incomes in the recession) can also affect the observed patterns, and further research is needed to identify the contribution of such compositional factors⁶.

The picture of changes in the income distribution can be complemented by a brief summary of changes in measures of poverty (Figure 3 and Table 4). The percentage of individuals falling below 60% of median equivalised income (the Laeken indicator for "at risk of poverty") fell from 2005 to 2009, rising thereafter but with a net fall overall. The elderly (aged 65 plus) were the main exception to this pattern, with a substantial net fall in the risk of poverty over the full period 2005 to 2011.

We highlight here two methodological issues which could potentially have a significant impact on the results. The first relates to the treatment of retirement and redundancy lump sum payments; the second concerns the representativeness of the samples.

Analyses of changes in income inequality typically make use of a number of decomposition methods. In this section, we consider four broad approaches:

A different question–the extent to which changes in income inequality are due to changes in market income or in tax/transfer policy–is addressed in the following section, using an alternative approach.

Before considering formal decomposition results, it is useful to examine how the pattern of changes in decile shares seen above is broken down in terms of the major income components. Table 5 shows the changes in shares of total equivalised disposable income by decile, and the contributions of each of the main income sources to these changes. For most deciles, falls in income shares due to employee income and self-employment income are partly offset by increases in transfers. Higher taxes are paid by those in the upper half of the income distribution, especially those in the 9^th and 10^th deciles. The top decile is exceptional in that employee income contributes strongly to a sharp rise in the share of the decile in total income, offset by a reduction arising from lower self-employment income.

The pattern of the contribution of transfers warrants closer investigation. The last row of Table 5 shows that transfers made up almost 8 percentage points more of total income in 2011 than in 2008. More than half of this increase went to the four middle income deciles (4 to 7). The changes for the three lowest income deciles were more limited, with the transfer share rising by 1 percentage point. This may reflect the fact that the major change in transfers between 2008 and 2011 was the rise in number of people receiving unemployment compensation. Such transfers are often be received by individuals with some employment income for part of the year. Given that the analysis is based on SILC’s annual income construct, households containing these individuals are less likely to be in the lowest income deciles, compared with those depending on social security payments for the entire year. Furthermore, there may be greater numbers of self-employed in the bottom decile in 2011, and they would have limited entitlements to benefits: this is a matter for further investigation. The combination of these factors may explain why the increased contribution of transfers is larger in the middle deciles than the lowest deciles. The contribution of transfer payments to the rise in the income share of the top decile is unusual, and arises in part from the inclusion of retirement and redundancy lump sum payments, as these can boost income to levels which bring the recipients into the top decile.

The disequalising impact of benefits in 2010 arises from the lump-sum payments (pensions and redundancy) referred to earlier. When these are excluded, benefits had a small equalising impact–but far less so than taxes. This result contrasts strongly with the findings of what Fuest et al. (2010) term the "standard approach": an examination of how inequality indices change in moving from market income to gross income (via the addition of benefits) and then to disposable income (with the deduction of taxes and social insurance contributions).

Paul (2004) points to the major reason why the Shorrocks (1982) factor source decomposition attributes a lower role for benefits in reducing inequality. One of the axioms underlying the decomposition is that "a given income source makes no contribution to aggregate inequality if every household receives equal income from that source". Paul argues that this requirement is at odds with the idea that "if each household receives a constant positive income from a source, then the aggregate inequality declines".

Fuest et al. (2010) expand on this, in the context of a strong contrast between the "standard" approach and the Shorrocks factor decomposition across all 27 EU countries. They note that within the standard approach "equally distributed transfers imply an inequality reduction". This is because the pre-existing distribution of market income is itself unequal, so that equal-valued cash additions tend to reduce income inequality. In the decomposition approach, however, equal-valued cash transfers make no contribution to the reduction of inequality because their correlation with disposable income is zero. This is axiomatic in the Shorrocks approach, but the axiom is implicitly rejected in the "standard" approach. Thus, differences in the results found by these two approaches derive fundamentally from the contrasting views they take on whether equal-valued transfers contribute to inequality reduction."

There is strong interest in many countries in assessing the distributional impact of austerity measures. Traditional decomposition methods focus on changes between observed outcomes in a base year, with its associated tax/transfer policy, and an end-year, with its associated policy. Such approaches may, for example, identify an increase in social assistance income, but cannot say if this arises from increased generosity of benefit payments or from an automatic increase in the incidence of transfers as unemployment rises. Bargain and Callan (2010) propose a decomposition which has particular advantages in addressing such questions. The decomposition partitions the total change into a part which reflects changes in policy, and all other sources of change. A counterfactual policy designed to be distributionally neutral plays a key role: this is simply the base year policy, indexed by the growth or decline in a broad measure of income⁸. The impact of policy change is then measured by estimating inequality measures under this counterfactual "distributionally neutral" policy and under actual policy, as simulated using a tax benefit model. Where possible this is done for both base year and end year data: the average of the two can be interpreted as a Shapley value decomposition.

Work along these lines is currently under way (Bargain et al., 2013). Figure 4 gives a broader picture of the impact of policy over the full 2008 to 2011 period. Here the analysis is based on 2010 data, and on a "distributionally neutral" policy which indexes 2008 policy in line with a fall (2.4 per cent) in average weekly earnings over the period. The analysis includes the main changes in income tax, social insurance contributions and the introduction of income levies as well as changes in benefit payment rates. In addition, the modelling includes the impact of three rounds of reductions in public sector pay, which were progressively structured⁹.

Over the full 2008 to 2011 period, the largest percentage losses of equivalised disposable income were in the top two deciles (9.5 and 11 per cent for deciles 9 and 10 respectively). The lowest losses were in the 1^st to 3^rd deciles (between 3 and 5 per cent). A key factor in the lower losses for deciles 2 and 3 is that payment rates for pensioners were held constant, while there were explicit cuts in payment rates for those of working age, and deeper cuts in Child Benefit. With the exception of the lowest decile, Figure 4 shows that, in aggregate, Budgets 2008 to 2011 were progressively structured. The bottom decile, however, have losses higher than the 2^nd decile, and comparable with those of the 3^rd decile.

Budgetary policy in the most recent years has involved somewhat greater losses for low income groups than those at the top. However, over the full 2008 to 2013 period it remains true that the highest losses are in the top quintile, with lower than average losses for other deciles.

Summary measures of income inequality have been broadly stable in Ireland over a long period, from the early 1990s through to the start of the current recession. There were, however, some significant shifts on a year-by-year basis in the years 2008 to 2011, during which average incomes fell sharply as Ireland experienced the full force of a major recession. The year by year pattern shows a fall in income inequality in 2009, reversed in the following years. Some of this is directly attributable to the timing of policy changes, with an increase in welfare payments in 2009, while taxes and levies were sharply increased. Later years saw more emphasis on expenditure cuts, and less on income-based taxes. Over the full period 2008 to 2011 the Gini coefficient has returned to the narrow range in which it has remained over since the 1990s. There have, however, been major losses in income share for both bottom and top deciles, with gains in income shares focused on the rest of the upper half of the distribution.

Decomposition analyses find that employment income acted to increase income inequality, while changes in self-employment income partially offset this. The large rise in unemployment was a major factor in increasing concentration of employment income, but there was also an increase in employment income in the top decile relative to others. This was partly offset by falls in self-employment income for the top decile.

What of the impact of policy changes in the areas of direct tax, welfare and public sector pay? The SWITCH tax-benefit model permits analysis of this issue, and found that policy changes over the 2009 to 2011 period tended to reduce incomes by more in the upper half than in the lower half of the distribution; with the percentage losses rising with income in the upper half of the distribution. The pattern of losses in the bottom half of the distribution reflected the fact that payment rates for benefits to those of working age were reduced over the period, whereas payment rates for pension benefits were increased in 2009 and then held constant. Thus, losses were lowest in the 2^nd and 3^rd deciles, which contained higher proportions of pensioners than other deciles.

We note two broader implications from these findings. First, the widespread perception that austerity policies in terms of tax, welfare and public sector pay are regressive is not supported by the findings in the Irish case. Thus, it cannot be argued that austerity policies must be regressive: the experience of Ireland over the 2008 to 2011 period provides a counterexample. The progressivity of the initial policy response in 2009 owed much to substantial increases in income-related taxes–exploiting some initial "headroom" in tax rates, relative to EU averages. As this gap has narrowed, income-related taxes have played a lesser role in later adjustment, so that the overall pattern of policy impact over the 2009 to 2014 period has become more complex¹⁰. Secondly, the Irish findings suggest that the squeeze, in income distribution terms, has been particularly sharp at the bottom and the top of the income distribution. Thus, while those on middle incomes have also experienced a squeeze, it is less sharp than that experienced by the top and bottom.

¹There were also substantial effects on the distribution of wealth, with the collapse in house prices being a major factor. This issue may be best studied in the context of data from the new Household Consumer Finance Survey, due for publication in 2014.

²The equivalence scale is the one used in Ireland’s official measures of poverty: 1 for the first adult, 0.66 for other adults, and 0.33 for children aged under 14. This approximates the scale used in social welfare payments.

³Lorenz dominance, in this context, means that the poorest x% have a higher share of income in 2009 than in 2008, for any value of x. The full Lorenz Curves are available from the authors on request.

⁴Nominal disposable income in 2011 is adjusted to 2008 levels using the Consumer Price Index, which fell by 2.9 per cent between 2008 and 2011.

⁵The policy analysis is conducted at micro-level, meaning that we do not allow for a macro feedback effect. In other words, we do not examine what would have happened to income if instead of the austerity policies actually implemented some other approach had been taken.

⁶Like most studies of income distribution (e.g., OECD 2010 and 2011) our analysis focuses on what has happened to the distribution itself–a distribution of "income positions" rather than of people. It is, of course, also of interest to track how individual earners and households have fared during the recession. McCarthy et al., (2012), in a study based on tax records, find substantial movement of individual tax units between quintiles of gross income over the 2006 to 2010 period. The panel aspect of SILC is not sufficient to support a similar analysis of the dynamics of household income distribution.

⁷This has been implemented in Stata by Jenkins (2009).

⁸When data for base year and end year are available, the change in gross income provides a natural indexing factor; where income growth must be based on forward looking estimates, changes in weekly earnings are often used.

⁹The first round of reductions was labelled a "Pension Related Deduction"; the other two rounds were explicit cuts to pay.

¹⁰Callan et al. 2013 present results which take into account a number of factors, over and above the direct tax, welfare and pay impacts assessed here. These include indirect tax increases, bearing disproportionately on low incomes, and restrictions on some tax reliefs affecting mainly high incomes. It remains the case that the greatest policy-induced losses in cash income are at the top of the income distribution–but the next greatest is now at the bottom, with broadly proportional losses elsewhere.