Coming to Terms with Termination of Public Organizations

Why do some public organizations survive and others do not? The bureaucratic retrenchment efforts since the 1980s showed public administration scholars that the structure of the state has its own demography. Yet studies on the survival of public organizations tested different hypotheses using different methods and definitions. This review article reflects on the consequences of different takes on termination, it compares and reports on the main findings thus far and argues that we should study organizational survival focusing on transitions to build on existing empirical work.


Introduction
Are bureaucracies really 'immortal' and-once created-do they become more powerful than the elected politicians that are supposed to control them? By the end of the 1970s, when welfare states had seemingly reached their limits in Western democracies, these questions gained prominence in studies of US and European public sectors. Political leaders such as President Ronald Reagan in the United States and Prime Minister Margaret Thatcher in the United Kingdom embarked upon a series of administrative reform trajectories to Bshrink the state^: privatization, deregulation, agencification, liberalization, decentralization, and New Public Management all aimed to downsize central government bureaucracies. However, those same politicians soon found out that changing the structure and organization of their central governments was a hard nut to crack. Not only did administrative reforms evoke harsh political opposition, they also increased-if successful at all-fragmentation and coordination costs (Pollitt 2003;Christensen and Laegreid 2007).
The politics of public sector reforms coincided with a growing body of literature on the demography of government. Starting with Herbert Kaufman's seminal Are Government Organizations Immortal? (1976), scholars on both sides of the Atlantic and beyond studied the transformation of government-including the survival of public organizations Hogwood 1988, 1991;Lewis 2002Lewis , 2004Carpenter and Lewis 2004;Adam et al. 2007;Boin et al. 2010Boin et al. , 2016Rolland and Roness 2011;Park 2013;MacCarthaigh 2014;James et al. 2015).
The conclusions of these studies are clear on one aspect. Public organizations are far from immortal. Across a diversity of public sector populations under study, one thing stands out: over time, most public organizations perish or lose their structural autonomy. The explanations vary. Public sector organizations brutally compete over limited government resources. They suffer from political turnovers and executive ambitions of each new incumbent. They risk liabilities of newness, adolescence and obsolescence. Their institutional design can protect them in the short run but over time turn into a liability or vice versa.
Also on other aspects of organizational mortality, existing studies disagree. The explanations for termination and survival of public organizations often ignore the findings of others both interdisciplinary and within their own discipline. More disturbingly, disagreement over definitions of organizational termination continues to feed the debate, and color the findings on survival. As if during an autopsy, different doctors arrive at different conclusions on the cause of death whereas a third one claims the patient has not died at all. We think much can be gained from overview of the debate on public organization survival, to increase the knowledge on the demography of government and increase our insight into political control over the organization of the public sector.
The aim of this article is twofold. First, we review existing articles that explain survival and termination of (public) organizations. We evaluate both hypotheses and findings from institutionalist theories that use a more 'biological' definition of termination (i.e. 'death' of the organization) and from theories of political control over the public sector that are more interested in a range of transitions of public organizations. We offer an overview of the findings. Results and findings from different analyses do not add up, given the use of varying definitions and incompatible methodology. But they do give us a set of factors that may matter. This relates to the second aim of this article, to evaluate the full range of existing survival explanations and build on previous findings to contribute to theory. Based on our review of termination definitions, we think different concepts have important implications for research findings. In the concluding section we propose to get rid of the 'termination' terminology and the biological life and death metaphors and replace it with a focus on 'transitions'.
The structure of this paper is as follows. In the following section we explain the design of the literature review. In section three we discuss the differences in conceptualizing termination in the reviewed studies and the implications terminological differences have for their findings. Section four introduces a model for survival of public sector organizations based on the reviewed literature and section five discusses the comparison of findings. We conclude in section six with reflections on the road ahead.

Design of this Study
To survey what factors influenced survival and termination of public sector organizations according to scholarly accounts of the past decades, we decided to cast the net widely and include all possible social science disciplines that could offer us potential explanations. We decided to focus on scholarly articles and used two systemic search methods to identify relevant works.
First, we employed the advanced search of the Web of Science database provided by Thomson Reuters. Within English language articles of their Social Science index we used a Boolean approach to find key word combinations of 'organiz*' and 'termin*'or 'surviv*'or 'death'. We refined our initial 2393 results by excluding articles from the life sciences, biology, engineering, computer sciences, environmental and geosciences, linguistics, cultural and religious studies. The refined result (447 hits) was subject to further selection based on abstracts to identify 90 articles that focused on the demography of public sector organizations, excluding groups (regimes, coalitions, teams) and firms (profit-oriented organizations). We have conducted another search round on 'agenc*' and 'termin*'or 'surviv*'or 'death', but found only a handful new articles of interests because most overlapped.
Second, we searched the references of our selected results for additional relevant articles. This resulted in another set of 18 articles that we looked into. Our aim was to arrive at a comprehensive set of factors influencing the survival or termination of organizations in the public sector. We therefore further excluded articles that focused on non-state organizations, such as NGO's, voluntary associations, community-based organizations, non-profit agencies, interest groups, foundations, and other similar types of organizations. Also, for the sake of comparability we decided to focus on central government apparatuses and excluded studies on decentral (branches of) public organizations. Studies on central government cabinet departments were not included because such studies seem to address issues of portfolio allocation rather than administrative survival (Cf. Mortensen and Green-Pedersen 2015;Glor 2011). To check whether we applied our exclusion criteria well, we looked into the content of the articles to decide whether it should be part of the review set. The final set comprised of twelve articles published since 2000.
The review set was then studied in more detail to find contributions to our main research question: why do some public organizations die whereas others survive under similar circumstances? The results did not allow for a full systematic review or meta-analysis that are more common in health, education and criminology studies. Our approach could perhaps best be labeled a 'scoping study': a type of structured literature review that allows for different study designs and noncompatible results to be included (Arksey and O'Malley 2005). Scoping studies follow systematic steps in reviewing available literature, while maintaining tolerance for variation of the included approaches, because the ultimate aim is to map existing findings and gaps in the literature on a specified topic, in this case: organizational survival.

Termination Terminology
The Idea of Termination What constitutes survival and termination of public organizations, really? Existing studies clearly disagree. In the field of business administration there seems to be less discussion on this matter. BSurvival of a private sector agency is a relatively identifiable piece of data. The organization is either in business or it is not. Even in the private sector, however, mergers, takeovers and splits make survival more than a simply mechanical exercise^ (Peters and Hogwood 1991: 88-89).
When Kaufman initiated the debate on the mortality of government organizations forty years ago, he casted his discussion in biological terms, i.e. the 'mortality', 'life' and 'death' of public organizations. In business administration studies, this biological metaphor is commonly used. The 'population ecology' approach adopted the entire idea that organizations that compete with the same product or service (be it newspapers, shoe factories, restaurants) in a given area are a 'species' (Hannan and Freeman 1977). Population ecology studies on the private sector borrowed concepts from biology to explain firm mortality and organizational founding, as well as population growth and change. According to Gray and Lowery (2000: 58), population ecologists do not so much claim to equate organizations with organisms though, because they are inattentive to individual traits of organizations other than their status of being 'alive' or 'dead'. Population ecologists only look at how populations (large numbers) of organizations interact with each other and with their environment. This idea has since been applied increasingly on populations of public sector organizations as well.
Population ecology thus requires a strict, dichotomous definition of what constitutes survival. Yet defining termination for public organizations is an entirely different matter, surrounded with ambiguity. Being 'dead' or 'alive' in bureaucracy is not a dichotomy. Theoretically, the study of organizational termination used to be a subfield of the policy termination and policy cycle literature (DeLeon 1982). These literatures claim that the policies carried out by public organizations do not often come to a definite end (Adam et al. 2007). In fact, public organizations and the policy programs they implement seem to be subject to constant structural reform, which blurs the distinction between termination or continuation in a dressed up (or down) version. BWhen imagined as a binary choice, these details are lost^ (Bertelli and Sinclair 2016: 6). In fact, Breorganization is an ongoing pastime for politicians and managers in government^ (Peters and Hogwood 1991: 89).
This reorganizing capacity of political executives is what makes public administration and business administration such different fields. The difference between public and private organizations is not about lack of competition for allegedly 'monopolized' public organizations. Competition in the public sector is fierce, since the total budget is limited and a wide variety of species wants a piece of the same cake. Yet, public organizations do not only perish because rival organizations did a better job in securing essential resources. Public organizations can also be deliberately axed or denied funding by a political executive for ideological or electoral reasons. This difference matters for both termination definitions and their implications. In the study of the survival of public organizations, different perspectives on politics and administration have driven a wedge between existing studies leading to a considerable fragmentation of insights.

Two Perspectives: Political Control and Institutional Legacy
Government organizations are supposed to submit to political control. The political control perspective therefore implies that structural changes to public organizations are first and foremost political acts. Each political act deserves similar attention, ranging from subtle name changes of organizations to fully fledged abolitions. The contrasting approach, we call it institutional legacy, implies that changes in the structure of government are best explained by the fact that public organizations have their own interests (probably continuation, according to Downs 1967). The public organization adapts to its environment and sees political interventions as a necessary evil that becomes part of the legacy to be carried into its future.
Studies on public organization survival have thus sprung from two different perspectives and two contrasting research approaches can be discerned. Either the observer -from a political science background presumablyhas an interest in studying political control over the structure of government, from the perspective of the principal. Or the organizations themselves take center stage and the studyfrom a public administration or sociological institutional backgroundfocuses on what organizations do to strike root, survive and prosper. The latter looks into explanations of the rise and demise of particular organizations from an institutional legacy approach.

The Political Control Approach
The political control approach tends to define most change events (i.e., political interventions) as termination and (re-) creation. While these changes reflect what really happens to these public organizations, their status, functions and form, above all they signal important political acts representing political control over the public sector. To indicate the subtle variety, one can discern between different 'termination types', such as termination by change of form, termination by merger or termination by acquisition of activity (James et al. 2015). It is important that no changes go by unnoticed because ignoring name changes and other events such as change of superior organization may mask important discontinuities.
Critics may argue that the least pervasive event (name change) does not really reflect termination. But from the above perspective it does reflect a significant act of political control over the organization. For instance, when the Dutch government decided to name the 'Luchtverkeersbeveiliging' (Air Traffic Security) from 1998 onwards the 'Lucht Verkeersleiding Nederland' (LVNL, Air Traffic Control Netherlands). This name change signals an outward and transboundary approach to air traffic control. ATC has long been a domestic issue, but increasingly became a joint European and global responsibility-an evolution reflected in its name change. Though the agency continues as an organization this event can be interpreted as a termination of its previous form, and a start of a 'new' phase in a new form. Such cases represent a discontinuation of 'key structural features'. From a political control perspective such a discontinuation or structural change justifies to be identified as an 'event', implying some form of termination (Hajnal 2012: 98;Cf. Greasley and Hanretty 2014).

The Institutional Legacy Approach
The institutional legacy approach would advocate a more dichotomous definition of termination instead of the subtle variety above. Organizations are only terminated if they are disbanded, when they cease to exist in any form-period. All other events in the life of a public organization are forms of adaptation, perhaps even on its own initiative, in its struggle to survive (see also Dommett and Skelcher 2014, for an account of agency responses when facing abolition and reform). In an impressive historical institutionalist account of the life and death of the Australian Wheat Board-later AWB Ltd.-, Botterill (2011: 637) shows the evolution from Ba venerable institution, anchored in collective agrarian values^, to an Binternational grain trader prepared to sell wheat whatever it took^-until the Oil for Food scandal unmasked its institutional drift and caused its final death blow. Institutionalists in Selznick's tradition (1957: 17) focus on how public organizations got 'infused with value' from the moment of its inception and how the legacy that the organization built over time became the seeds of its eventual demise. The political control approach above would instead dissect the life of the AWB in different durations for 'different' organizations, whereas students of institutional legacy emphasize the integrity of the institution, adapting over time.

Research Designs and Termination Definitions
The focus on institutional legacy lends itself better for comparative case studies than for large N studies on populations. Case studies allow for a definition of termination that depends largely on the judgment of the researcher. The researcher builds an argument and in turn provides the reader with ample substance on each case to validate the argumentation. Such an approach would also suit Kaufman's definition based on 'uninterruptedly maintained organizational boundaries ' (1976: 28) or Starbuck and Nystrom's (1981, xiv) identification of 'significantly different responsibilities', or 'substantial reorganization'. The larger the N, the more distance from the cases, the harder it becomes to make such substantiated judgments.
Yet even among the large N studies, one can discern approaches that take the organization and its institutional legacy as vantage point Hogwood 1988, 1991;Rolland and Roness 2011;MacCarthaigh 2014). They discern 'maintenance', 'succession' or organizational change categories as a go-between life and death. Succession could be linear, non-linear, partial, complex and/or it could involve mergers or splits (Peters and Hogwood 1988: 23-24). Rolland and Roness (2011) and MacCarthaigh (2014) introduce subcategories discerning the dominant partner in the merger or the dominant offspring that represents the continuation after the split. These studies use comparable definitions, but they do not explain events. Their distinction between terminating events and maintenance events does, for example, not allow regression analysis and event history modeling to generate statistically robust explanations.
The political control approach works best for large N statistical regressions. Studies within this realm seek statistical explanations of survival. Peering from the principal end of the discussion, authors look at organizational change as an act of political intervention. In order to find out which political factors influence survival, they employ event-history analysis to explain 'termination' events (cf. Box-Steffensmeier and Jones 2004). There is no room for categories between life and death here, so most changes in the organizational structure are considered a termination event. Bertelli and Sinclair (2016) offer a way out by discerning both the fully abolished cases, the reformed ones and the retained ones in different regression models. O'Leary (2015) also uses logistic regression to explain termination of public organizations but only considers fully abolished ones as terminated.
It is rather unfortunate that even studies that do statistically investigate the factors that affect survival, employ different definitions. Their findings have a different base and as such they do not add up in a meta-analysis. Park (2013) follows Carroll and Delacroix (1982: 170) in that she only considers after a merger Bthe resulting organization as new and the two merger partners as dead^. She adopts a termination definition that includes merging and splitting, plus shifting to the private sector, in line with Boin et al. (2010). The latter also treat organizational survival dichotomously (no in-betweens), but did not encounter mergers and splits (perhaps due to their specific population of New Deal organizations). To Lewis (2002) an organization is dead as the result of a combination of name change, change of proximity to the president and change of function of the organization, which leads to more terminations than the abolishment/merger/split definition. To Boin et al., the organizational structure and leadership would have to be changed, too, before the organization was considered terminated (2010: 6-7). These studies do seem to come close in their definition of events.
By contrast, Greasley and Hanretty (2014: 11-12) emphasize that Bto abolish, merge, rename and otherwise signals a discontinuity in an organization's life…^. They also see reorganizations and name changes as terminations. Of their sample, 58% gets terminated in a 23-year period. James et al. (2015) consider an organization not terminated when its name changes but when its Bgovernance structure (of defined roles for managers formally leading an organization and associated accountability mechanisms) has been ended^(p. 8). The James et al. (2015) definition deviates so much from other definitions used, that it is impossible to say what implications it has for the comparability of the termination rate. The number of terminations in their study is so high though (109 of 153 organizations in 23 years, which adds up to 71%) that it might indicate a bias in favor of discontinuation. As Park (2013: 416) concluded, Bthe distinction between termination and change remains a matter of pragmatic choice rather than strict theoretical underpinnings^(see also Hajnal 2012). Again, much is left to the judgment of the researcher. 1 On an interpretative continuum varying from biasing in favor of stability versus discontinuity of organization, the definitions of the studies discussed would rank as follows (Fig. 1): What we see is that most studies that focus on organizational characteristics as a theoretical explanation for survival, also tend to use termination definitions that imply more stability. Those studies that see political control as the main explanans for survival, focus on organizational discontinuity.
When we keep in mind the different definitions of termination, we are better equipped to appreciate the results of the subsequent analyses of different studies. Presumably, the institutional legacy definitions lead to less terminations and higher survival rates. In the next section, we distill from the literature the most important factors and findings, discussing them one by one and we present per independent variable the prevailing knowledge to construct a model of organizational survival.

A Model of Organizational Survival in the Public Sector
Though the definition of organizational termination varies per study, in this section we treat the definition as if it were 'constant' across all twelve studies. On the basis of an assessment of the findings in the existing literature we build a theoretical model of determinants of organizational survival as if termination is conceptualized in a similar way. This exercise provides us with an overview of common findings and disagreements in existing research. The section first introduces the hypotheses and ends with a summary of findings.
We have discerned: 1) demographic type explanations centering around the age of the organization; 2) design type explanations focusing on 'birth characteristics' of organizations that will have implications for future survival chances; and 3) contextual or functional explanations, affecting the population in total or organizations with a particular degree of specialization or niche, to elucidate why some survive changes in their environment as opposed to others.
First, and perhaps the most prominent and oldest factor of interest to students of organizational survival is the age of the organization as a factor in itself. Conventional wisdom suggests that organizations tend to survive infinitely after they reached maturity. The classic reference here is from Downs: Bthe older a bureau is, the less likely it is to die^(1967: 20). Lowi (1979: 309) explained: 'once an agency is established, its resources favor its own survival, and the longer agencies survive, the more likely they are to continue to survive'.
Organizational theorists demonstrated however, that an organization's survival chances do not increase with age in a linear fashion. Stinchcombe (1965) introduced the idea of a 'liability of newness', when new organizations have not been  Fig. 1 Organizational termination on a continuum able to nest themselves sufficiently in their environment, when routines are not well established and political volatility or growing pains may be fatal to the newborn organization. By contrast, some scholars (Baum 1989) have argued, organizations may also grow old and become obsolete (a 'liability of obsolescence') when their functionality cannot keep up with changes in their environment (think of a steamboat inspection service, or a firm such as Kodak) or when their mission has been accomplished (a Western organization fighting leprosy or polio). When both arguments hold, one would expect the survival chances of an organization to resemble an inverted U-curve. Another demographic liability involves the dangers of adolescence (Bruderl and Schussler 1990). Considering that legitimacy of the public organization abounds at the moment of creation, i.e. that its legislative support base has just been established and its proponents are still in office, the infant has a few years to develop unharmed. After more than a decade, the organization can be considered to have secured its resources and support base well enough to fend of competitors and opponents. However, in the period right after its 'honeymoon' and before it has taken firm root, the organization's hazards rise because the protective shield of its creators may have diminished and it may have become politically opportune for its opponents to attack the organization or its policy. In this same period of adolescence, legislative actors have had time to learn about the organization and its accomplishments or performance (Carpenter and Lewis 2004). This learning may result in the increased likelihood of fatal critique and termination.
Summing up, the text box below presents the most prominent age-hypotheses that we detracted from the prevailing literature. We chose to hypothesize each age effect separately in order to separate their estimated effects in order to give a better overview of those age effects confirmed in the literature (see Table 1).
H1 A public organization faces higher hazards in the first five years after creation (liability of newness). After five years the survival chances of public organizations increase and become stable. H2 A public organization faces higher hazards in its 'adolescent' years (>5 < 10). The organization is imbued with immunity against termination by its creator(s) in its first years and once it has overcome the hazards of adolescence its survival chances increase and become stable. H3 A public organization faces higher termination hazards when it has reached a mature age (>30).
Second, the survival of public organizations can be explained by looking at the 'birth characteristics' or organizational 'dna'. Some organizations are simply stronger in their set up or have better armor for survival. This theory prevails especially within the fields of political science, where organizations are seen as embodiments of political interests and the acts of organizational creation and termination are intrinsically linked to legislative-executive decision-making. Though public organizations do often have a clientele, and a support base, there is no market mechanism or invisible hand to take them out operation. Instead opponents are sometimes highly visible, and their 'attacks' are predictable in such a way that creators try to 'hardwire' their organizational offspring into survival.  Here, we look into the effects of such design. The work of Moe (1989) inspired Lewis (2002Lewis ( , 2004 and Boin et al. (2010) to study the effects of design on populations of federal US government organizations. From their studies we derived the following hypotheses: H4 Organizations established by formal law have higher chances of survival than organizations established by decree. H5 Organizations set up at arm's length of government have higher chances of survival than organizations that operate in the hierarchy of a ministerial department or executive office.
Though perhaps less a matter of intentional hard wiring, populations of firms can feel the effects of organizational structure. Moore and Kraatz (2011) compared the performance (with implications for survival) of 4000 Savings and Loans organizations in the US. In the long run, the mutualities (80% of the population) fared better and took less (unjustified) risk than the stock organizations (20%). The board structure and notfor profit legacy of the mutualities prevented them from the audacity that became detrimental over time to many of the stocks. In similar vein, Lewis (2002) and Boin et al. (2010) looked at characteristics such as board structure (vs single administrators) or political appointments of government organizations to see if those characteristics in fact increased termination hazards.
H6 Organizations governed by a board have higher chances of survival than organizations governed by a single administrator.
An increasing number of studies from the United Kingdom, focusing on the termination of public agencies, have included the agency 'type' as a factor of influence (O'Leary 2015; Bertelli and Sinclair 2016). Other studies mention the 'regulatory function' of an organization as particularly important (Lewis 2002(Lewis , 2004James et al. 2015). We have included both types here, in distinct hypotheses, because the assumed direction of influence on survival is different.
H7 Advisory organizations have lower chances of survival than other types of public organizations. H8 Regulatory organizations are more likely to survive than other types of public organizations. Lewis (2002Lewis ( , 2004 asserts that some organizations are set up temporarily from the start. Their creation involves a 'sunset clause' that determines when the organization will be dismantled (Cf. Boin et al. 2010).
H9 When organization are created with a sunset clause, they are unlikely to survive that clause and mature into 'old age' (>average).
Third, existing research on survival in the public sector does take into account how the odds of termination are affected by political and economic indicators (Lewis 2002(Lewis , 2004Boin et al. 2010;Park 2013;James et al. 2015). These studies hypothesize that a conservative political climate leads to slimming the public sector, and that turnover in general is likely to affect survival of organizations in a negative way because incoming elects will probably attack the creations of their predecessors. Economic indicators serve as an important control: when government revenues go down, survival rates go down as well (Carpenter and Lewis 2004). Others assert the opposite: recession makes dismantling government organizations unlikely because it will increase unemployment (Greasley and Hanretty 2014).
H10 Political turnover negatively affects survival of public organizations. H11 A right wing (majority in) government negatively affects survival of public organizations. H12 Incumbency of a political executive of another political color than the organization's creator (so-called 'unfriendly government') negatively affects survival of public organizations. H13 When total government revenues decrease, the likelihood of survival of public organizations decreases as well.
In the next section we will discuss the actual influence of these factors on the survival of public organizations, resulting from the studies reviewed here.

Discussion of Findings
So far, we have resisted the temptation to reveal how these factors fared in empirical analyses on the public sector. In the overview below we present the findings from the studies we reviewed in this paper and for each of the hypotheses above we give the outcomes per study. Positive relation to survival is indicated by a '+', a negative effect on survival by a '-'. So a '+' for newness does not mean that the liability of newness is confirmed, but that newness has a positive effect on survival. A tested but inconsequential factor is indicated by an 'x' and the 'U' stands for non-linear effects (for instance insulation from presidential interference is a liability in the first six years of an organization's lifespan but protects it in the long run, see Boin et al. 2010).
We see in this overview, that on average more than 50 % of the public organizations perish during the period studied. By no means does the myth of immortality hold up. Because the time periods of the studies differ, it is important to annualize the results. The comparably high percentages of 75% (Boin et al. 2010) and 72% (James et al. 2015) turn out to reflect quite different termination rates when divided by the number of years over which terminations took place. When significant, nine factors seem to have a unidirectional confirmed effect on survival. Old age makes organizations more likely to survive (H3), and so does a regulatory function (H8), and a firm legislative base (H4). Adolescence (H2) is a hazard in two studies, but for different reasons and under different conditions. Advisory organizations (H7) are less likely to survive and, not surprisingly, so are organizations set up with a sunset clause (H9). In general, political turnover has a negative effect on survival of public organizations, i.e. new incumbents start reorganizing the administration (H10). Unfriendly incumbencies (i.e. political executives of a different political ideology -H12) are likely to target the organizational creations of their political opponents previously in office. If the color of the incumbency matters, it is because rightwing incumbencies (H11) have a negative effect on survival.
In two studies the ideology of the incumbent government interacts with other factors. Götz et al. (2015) show that leftwing governments are less inclined to terminate public organizations than rightwing governments, unless budgetary pressure increases (factors 11 and 13). Similarly, Greasley and Hanretty (2014) conclude that public organizations face higher hazards operating under rightwing governments in 'normal' times (low to moderately high budget pressure). But when debt increases, organizations operating under a left-wing government face higher termination risks. Greasley and Hanretty add another factor though: Bthe effect of greater-than-average debt on agency termination is negative for very young agencies, and for older agencies, but positive for agencies in their third to seventh years^(p. 17). So leftwing executives under budgetary pressure are inclined to target adolescent public organizations for termination (in line with our hypothesis 2).
On the other factors, contradictory evidence exists. Carpenter and Lewis (2004) contradict conventional wisdom that budgetary pressure negatively affects survival (H13). In their study, the effect is the opposite (and robust in three different models): budgetary surpluses increase the hazards, and deficits make terminations less likely. Due to the short term cost of organizational termination, such decisions would be unfavorable in times of fiscal constraints and budgetary austerity (Carpenter and Lewis 2004: 222). Park (2013) tested this claim but found no significant effects.
In Lewis's 2002 and2004 studies, insulation from executive interference (arm's length status -H5) makes organizations fare better than those created in proximity to the executive but this is contradicted by other research (Boin et al. 2010;Hajnal 2012). Newness of the organization is an asset according to some and a liability according to others (H1).
An updated model would predict that regulatory public organizations created by law and existing for at least 12 years are the most termination-proof, but by no means hazard-free. Newly created or adolescent advisory organizations created by executive order are much less likely to survive. In general, hazards increase when political winds change. New incumbents tend to weed the administration upon entry into the executive office targeting the showcases of their political opponents but also reforming the administration set up by predecessors of the same ideology. Rightwing executives are particularly likely to terminate public organizations and leftwing governments become likely to do so under budgetary pressure.
The recent studies on survival and termination of public organizations have done much to debunk old myths. Downs (1967), Kaufman (1976) and Lowi (1979) were proven wrong on their assumptions of near-immortality of bureaucracy. Stinchcombe's liability of newness (1965) does not always hold up either. Time to move on to new hypotheses in the future. The next section will discuss some promising avenues for further research.

Reflections on the Way ahead
This paper has provided an overview of factors that could explain survival of public organizations and evaluated the main hypotheses from the literature by comparing the results. To improve the future debate on organizational survival in the public sector and build on existing research, we propose to replace the binary definition of organizational termination with a focus on 'transitions'. Termination implies a definite 'end' that may apply to a period in which an organization took on a particular form, while the organization itself does not end. We prefer to use the more neutral term 'transitions' to refer to the events through which organizations move from one status to another. Transitions could be studied as dependent variables in their own right. Together, many transitions of individual governmental units could result in a transformation of government.
Transitions include all the structural changes to the organization that can be measured consistently and reliably over time, such as name changes, change of superior organization, change of hierarchical level, change from staff to line unit, changes in structural autonomy by law, as well as mergers, splits and abolishments. This more fine-grained approach allows building both on existing datasets and on studies that have a dichotomous or biological take on survival. When detailed transitions are coded as events, data can still be made compatible to studies focusing on institutional legacy-by systematically ignoring the subtle transitions such as name changes or changes of level in the bureaucratic hierarchy and focusing on the more definite passings only such as full abolishments or mergers. Meanwhile coding all transitions as events affords employing event history modeling for multivariate regression analyses and explain what affects survival in the public sector. At the same time it affords studying the transformations of the state in a variety of ways.
Coding transition events allows to identify transformations of individual organizations. Sequential transitions that we include in a single analysis as forms of 'termination' could turn out to be a clever organizational adaptation to changing resource levels that ultimately save the organization from abolishment. Alternatively, Levine et al. (1981) teach us that more subtle transitions could also represent staged responses to fiscal stress (ignoring changing resource levels by issuing window-dressing reforms or delaying tactics) that are ultimately the prelude of a more profound administrative restructuring of individual organizations (Levine et al. 1981). Future studies on organizational termination should probe into this.
The current studies on survival of public organizations mainly focus on independent agencies, in majoritarian political systems. Rich harvest is waiting in the study of reform of both public agencies and national ministries in continental Europe. Such populations would allow for systemic comparisons between majoritarian and consensual systems but also afford studying the effect of policy preferences of different parties in coalition governments, on administrative reform and organizational termination. It could build forward on findings of Götz et al. (2015) and Greasley and Hanretty (2014) that political ideology has different effects under different circumstances or for different organizations. In addition, such future research could look into the link between policy preferences (the Manifestos data on electoral programs) or policy agendas (the Comparative Agendas project) and organizational reforms within specific policy domains.
We know now that public organizations perish. We also know that organizational design can help to protect organizations against termination hazards. Time has come to probe deeper into how the structure of government relates to specific political power plays and particular policy preferences.