Table of Contents
1.1 The brief
1.2 Background information supplied
2. Differences in results between CIE2006 and CIE2014……………….4
2.1 The 2006 report
2.2 The 2014 report
2.3 Comparison - are the differences significant?
3. Explaining the differences……………………………………….……7
3.1 Model structures
3.2 Model parameters and data
3.3 The starting situations
3.4 The baselines
3.5 Reform scenarios
3.7 Norfolk Island Government finances and Commonwealth support
4. Concluding remarks…………………………………………………..11
References / Contacts
Michael Common (BA, BPhil University of Liverpool) is an English economist. Between 1968 and 1988 he held teaching appointments in economics departments at the Universities of Liverpool, Southampton and Stirling. From 1988 to 1998 he was a Senior Fellow at the Centre for Resource and Environmental Studies (now the Fenner School of Environment and Society) at the Australian National University, Canberra. In 1991 he served as an Associate Commissioner for the Industry Commission Inquiry into the costs and benefits for Australia of acting to reduce greenhouse gas emissions. From 1998 to 2003 he was Professor and Deputy Director at the Graduate School of Environmental Studies at the University of Strathclyde, Glasgow. He has held visiting appointments at a number of institutions including Rutgers Graduate Business School, the Open University, the University of York, and the Vienna University of Economics and Business. He is the author of many papers and a number of books, including being co-author of the globally successful textbook Natural Resource and Environmental Economics, now in its fourth edition. Currently retired and living on the Isle of Bute (Scotland), he remains academically active.
1.1 The brief
The Centre for International Economics, a Canberra and Sydney based consultancy, has provided the Australian Commonwealth Government with two reports, in 2006 and 2014, on the economic implications for Norfolk Island of it becoming, essentially, part of Australia. I have been asked to provide a short, plain language, and comprehensible, account of the important differences between the two reports.
I am a retired Economics Professor, now resident in the UK. I have never visited Norfolk Island, but I have knowledge of Australia, and of the model used in the two reports, based on being a Senior Fellow in the Centre for Resource and Environmental Studies at the Australian National University from 1988 to 1998.
1.2 Background information supplied
Dr Chris Nobbs provided me with copies of the two reports (1)(2), hereafter CIE2006 and CIE2014. Dr Nobbs has also provided me with background information, from which I take the following to be the most salient points for the purposes of this report.
CIE2006 was commissioned by the Department of Transport and Regional Services, which wanted an assessment of the economic impacts that would arise from the extension of Commonwealth legislation to Norfolk Island. The report was delivered in October 2006. In December 2006 the Minister for Local Government, Territories and Roads announced that the plan to restructure the governance of Norfolk Island would not proceed.
In November 2010 in a Memorandum of Understanding between the Commonwealth and Norfolk Island, the latter ‘agreed in broad terms to participate in the Australian taxation and social security systems on the basis that there will be a net benefit for Norfolk Island and its community and there is appropriate consideration of local circumstances’.
In November 2014 CIE2014 was delivered to the Minister for Infrastructure and Regional Development.
The Norfolk Island Legislation Amendment Bill passed the Australian Parliament in May 2015. As a result Norfolk Island will become a regional council in the state of New South Wales.
While there a number of other inquiries and reports over the period 2000-2014, it might be inferred from the timing that CIE2014, which could be read as forecasting substantial ‘net benefit for Norfolk Island and its community’, played an important role in the 2015 outcome.
CIE2006, as discussed below, estimated smaller Norfolk Island net benefit. Whereas CIE2014 was published at the time of its delivery, CIE2006 was not. In July 2011 Mr B. Sanderson applied to the Department of Infrastructure and Regional Development for a copy of CIE2006. The copy supplied was heavily redacted. Mr Sanderson appealed this decision and after a lengthy process the entirety of CIE2006 became publicly available in early 2015.
Reading CIE2006, it is unclear why, other than disappointment with the reported results perhaps, the Australian Government should have sought to keep it out of the public domain.
It is of interest that CIE2014 makes no reference at all to CIE2006.
2. Differences in results as between CIE2006 and CIE2014
The basic structure of the two reports is the same. A model of the Norfolk Island economy is used to establish a ‘baseline’ which is the forecast future path for the economy in the absence of the changes of interest, given various assumptions about the external conditions affecting that economy. Then, the model is re-run after being modified to reflect the changes of interest, and with any necessary changes to the assumed external conditions. The ‘changes of interest’ are often referred to as the economic ‘shock’, and in the two CIE reports are referred to as ‘the reform scenario’, or ‘the extension of Commonwealth legislation’.
2.1 The 2006 report
The results relate to Commonwealth legislation being applied to Norfolk Island on 1 July 2007. Results are reported, in Chart 5.3, for: GDP, Household Consumption, Exports and Imports, Inflation, Capital Stock, Cost of Labour, and Employment. GDP refers to Gross Domestic Product, that is the total income earned by Norfolk Island residents. Results are shown annually out to 2016/17, as per cent differences from the ‘baseline’, which is what the model estimates would happen in the absence of the roll-out of Commonwealth legislation.
In 2007/08 GDP is some 4 per cent below baseline, by 2011/12 it is equal to baseline, and the graph continues upward until, in 2016/17, GDP is 9 per cent above its level in the baseline projection.
Consumption per household also drops initially, to some 4 per cent below baseline. It recovers to the baseline level in 2009/10, and in 2016/17 it is some 4 per cent above baseline.
Labour costs rise above baseline and by the end of the analysis period they are up by about 50 per cent over baseline. Because of the higher cost of labour, Employment drops below the baseline initially, but then recovers, and by 2016/17 it is 15 per cent above the baseline. For the other indicators shown in CIE2006 Chart 5.3, no information is given in CIE2014.
It is estimated that over 2007/08 to 2016/17 the Australian Government would need to provide financial assistance to Norfolk Island amounting to $286 million, $32 million per annum, in addition to the transfers required as a result of the extension of Commonwealth legislation to the island.
2.2 The 2014 report
The results relate to Commonwealth legislation being applied to Norfolk Island in 2016/17. They are reported in the Executive Summary, and in Chapter 5 in several tables. Whereas CIE2006 reports results for a single reform scenario, CIE2014 reports results for four scenarios:
S1: Core Reform – Income Tax, Welfare, Medicare
S2: Other Tax Related – Superannuation, Aus GST, Aus Tariffs, Aus Fuel Excise
S3: No Norfolk Island GST, Customs Duty, Fuel Excise
S4: Other Reforms – Minimum Wage, Government Business Enterprise Reform
Table 5.2 gives results for S1, S2 and S3, where those for S2 and S3 are ‘premised on the Core Reform set having been implemented’ (page 47). These results are ‘broadly additive’, i.e. the impact of all of these reforms applied together would be approximately the sum of the results reported for the separate scenarios. All these 2014 scenarios applied together, along with S4, would be a reform package very similar to that of CIE2006.
In CIE2014 what was referred to as GDP in CIE2006 is called GTP for Gross Territory Product. For S1+S2+S3 there is, adding the entries in Table 5.2, an immediate increase in GTP over baseline of 31.4 per cent. The size of the excess over baseline declines to 28.3 per cent in 2020/21 and remains at that level until the last year for which there is an estimate, 2023/24. The relative contributions of the three scenarios are almost constant over time. For 2023/24 the deviations from the baseline are:
The GTP impact of S4 is relatively minor - of the order of -0.3 per cent over the whole forecast period.
Estimates for Household Consumption are not given for all scenarios. For S1, the Core Reform Scenario, the immediate impact is up 39.8 per cent on baseline, declining to 37.9 per cent in 2019/20, and then remaining at that level until 2023/24 (Table 5.1).
Estimates for Labour Cost and Employment are not given for all scenarios. For S1, there is an immediate and sustained increase in ‘Nominal Wages’ to 21 per cent above baseline. There is an immediate increase in Employment of 8.2 per cent above baseline, which declines to 7.9 per cent in 2019/20, and then remains constant.
Table 5.3 in CIE2014 gives the estimates for the ‘Net financial impost of reforms on the Commonwealth’ for the years 2015/16 to 2023/24. The figures are for increases over the baseline, are net impacts reflecting the new balance of flows in both directions (to and from Norfolk Island), , and are estimates for the impact of S1, S2 and S3. Initially the increase is $6 million and this rises to $7.4 million in 2023/24. The total over the analysis period is approximately $58 million. These estimates do not include any infrastructure costs.
2.3 Comparison – are the differences significant?
Commenting on the problems with data for its modelling, CIE2006 says (page 9) that:
……in the CIE’s view the data issues are not significant enough to change a result from being positive to negative, from being small to large or vice versa. (italics in the original).
At page 12, CIE2014 says that:
…..the modelling results presented in later chapters should only be used to infer the outcome of extending mainland taxation and welfare systems etc to Norfolk Island (positive or negative) and the magnitude of such impacts (small or large). It would be inappropriate to, for example, report modelling results to the 2nd decimal point and claim that as the unambiguous impact of any taxation/welfare reforms. That is, only broad messages and trends should be taken from the modelling results.
These statements reflect the general view of proponents of the kind of modelling used in CIE2006 and CIE2014.
It would be reasonable to apply the same standards to differences between the outputs from two applications of the same modelling technique to the same shock to an economy. Are the two sets of results qualitatively the same, in the same direction of response, positive or negative? Are the two sets of results quantitatively the same, both ‘small’ or both ‘large’?
Comparing CIE2006 and CIE2014 results for GDP/GTP, Household Consumption and Employment, the direction of change from baseline is, leaving aside the initial downward movement in CIE2006, the same in both. However, the numbers are quite different. Consider Household Consumption, which might reasonably be regarded as the best single economic indicator of the impact of the reforms on the welfare of Norfolk Island inhabitants. Whereas CIE2006 has it up by 4 per cent on baseline by the end of the projection period, CIE2014 has it up by 38 per cent for just the Core Reform scenario, and the figure would presumably be higher for the four scenarios together.
Given the inherent limitations of this kind of modelling, I think a practitioner would regard the CIE2006 impacts as generally ‘small’, but the CIE2014 impacts as generally ‘large’. Certainly, the differences between the CIE2006 and the CIE2014 impacts would have to be considered ‘large’.
The question which then arises is whether the differences can be explained in terms of changed circumstances as between 2006 and 2014? If they can, a reader could have some confidence in the usefulness of the modelling exercises. If no such explanation can be provided, the reader could reasonably suspect that one or both modelling exercises were of limited usefulness.
As noted, CIE2014 makes no reference at all to CIE2006. Consequently, the consultancy is offering no explanation for the differences, and, hence, offering no explicit basis for the reader to decide how much confidence to attach to either or both sets of estimated impacts, even in terms of positive/negative and small/large.
In the next section I examine what can be inferred from the two CIE reports about the origins of the differences in the results that they report.
3. Explaining of the differences.
There are several possible explanations for the differences, which are not mutually exclusive.
3.1 Model structures
The models used in CIE2006 and CIE2014 are particular versions of a generic model called ORANI-G. This is what is known as a ‘computable general equilibrium model’. Such models are:
Computable – they can be solved for numerical values of the things, such as GDP, that the model has determined by the functioning of the economy;
General – they include all markets;
Equilibrium – they assume that all markets clear, that is that in each market supply and demand are brought into equality by price movements.
It is not obvious that the market clearing assumption is appropriate for the Norfolk Island economy. Markets work well when there are many buyers and many sellers, whereas for Norfolk Island it is necessarily the case that in many markets there are few actors.
There is an important feature of the generic ORANI-G that is not mentioned in either CIE2006 or CIE2014. It has more variables, things like GDP, to be solved for, than there are equations. This means that the ‘things’ cannot be solved for. To get a particular solvable ORANI-G model it is necessary to include extra relationships – this is called ‘closure’. For proper understanding of the results of a particular version of ORANI-G the reader needs to be told what closure is involved. Different closures provide different, sometimes very different, results from the same model. In neither CIE2006 nor CIE2014 is the closure discussed.
ORANI-G models comprise an input-output table describing numerically the structure of the economy in terms of flows of goods and services between producing sectors, equations describing money flows between government and sectors and households, and equations describing how households respond to changes in their incomes and the prices facing them. For a particular application of ORANI-G it is necessary to define the sectors and put numbers in the input-output table, and to give numerical values to the parameters of the equations.
CIE2006 and CIE2014 both state the sectoral composition that they use. They are different. Whereas the model in CIE2006 has 18 sectors, that in CIE2014 has 25.
In neither of the CIE reports are the equations used listed.
In CIE2006, in a footnote at page 6, a url address is given from which it is stated that the model it uses can be downloaded. Using that address produces a message that the page does not exist. CIE2104 only provides an address for the generic ORANI-G model.
1 A parameter is a constant in an equation. Thus in y = a +bx, a and b are the parameters, numerical values for which determine how the variable y changes with changes in variable x. For a = 0 and b =1, for example, y is always equal to x.
3.2 Model parameters and data
Given a model structure, in terms of sectors and equations, its use requires numbers for the entries in the input-output table, the values for the parameters of the equations, and the initial conditions. Clearly, for meaningful results all of these numbers should be those applicable to the economy for which results are sought. Ideally, the values for the parameters of the equations should be determined statistically from historical data for the economy being modelled.
Both CIE2006 and CIE2014 are explicit about the dearth of useful data for Norfolk Island, which has never had the data collection systems necessary to generate the kind of data that ORANI-G modelling requires. Both reports discuss examples where attempts to derive the necessary data from available information ran into contradictions, and how those contradictions were resolved. Interestingly, the set of contradiction problems described in CIE2006 is completely different from that described in CIE2014.
In neither CIE report is there any discussion of how the numerical values for the parameters of the equations used to describe household behaviour were obtained. It is a reasonable inference that both modelling applications followed what is standard practice when doing computable general equilibrium modelling of small economies with a dearth of actual local data from which to infer numerical parameter values – namely use numbers obtained for economies where the data does permit the necessary inference. Or, use a ‘plausible’ number that seems to produce ‘reasonable’ results.
Obviously this is problematic. In the cases of the CIE2006 and CIE2014 modelling it is clear, for example, that, given that the reforms involve the introduction of income tax and a minimum wage, the number used to describe the response of hours of work offered to changes in the going wage rate will be an important determinant of the model results. It is equally clear that it is questionable whether the number that describes that response in, say, Australia, is going to be the same as the number for Norfolk Island.
3.3 The starting situations
It is conceivable that the same ORANI-G particular model, in terms of input-output structure and numerical values for equation parameters, could produce different results for the same reform scenario when applied in 2006 and 2014. This could arise because of changes in the initial conditions in both Norfolk Island and externally, which affect the impact of the reform package on the Norfolk Island economy. Clearly, the model structures are not the same in CIE2006 and CIE2014, and there is no information on the numerical parameter values used in either. So, it cannot be concluded that the differences are solely due to changed initial conditions, and/or any differences in the reform scenarios.
Nonetheless, the question remains as to whether initial conditions were different in such manner as to provide something toward an explanation of the differences in model results. Given that CIE2014 does not acknowledge the existence of CIE2006, one cannot look there for an explicit answer to this question.
It does appear that the Norfolk Island economy may have shrunk between 2006 and 2014. CIE2014 estimates Norfolk Island GTP in 2013/14 as $68 million. CIE2006 does not report its estimate for GDP in 2005/06 (and does not produce any dollar figures for any variable for the baseline or for the impact of reform). CIE2014 reports, in Table 1.1 page 6, estimates by others as $80.3 million in 1995/96, $62.1 million in 2004/05, $89.5 and $82.0 million in 2009/10, $87.9 million in 2010/11. Using the last of these estimates, it states that its 2013/14 figure ‘suggests that economic activity has fallen by around 22 per cent since 2010-11’.
If, on this not very firm basis, it is accepted that the Norfolk Island economy shrank substantially as between 2006 and 2014, would such a contraction make the impact of the reform package substantially greater? All that can be said is that this is possible.
3.4 The baselines
It is impossible to compare the baselines established by the models in CIE2006 and CIE 2014. In CIE2006 the baseline is reported only in terms of per cent deviations from initial levels. In CIE2014 the baseline is not reported at all.
What is clear is that the two baselines are established from different assumptions about what would happen over the period of analysis in the absence of the extension of Commonwealth legislation to Norfolk Island. For example, there are big differences in what is assumed about infrastructure spending, and the financing of the Norfolk Island Government. Intuition suggests that these differences significantly impact on the two sets of baseline results.
In CIE2006 it is assumed that an Asset Management Plan, AMP, is introduced in 2006/07, with spending initially at $9.9 million per year, rising to $17.4 million in the final year. The Norfolk Sustainability Levy, NSL, was introduced in July 2006, in order to broaden the tax base for the Norfolk Island Government. It is a consumption tax, initially set at the rate of 1 per cent. In constructing its baseline, CIE2006 sets up its particular ORANI-G model so that it is used:
to determine the rate at which the NSL needs to be set so as to balance government expenditure…with government income. That is, revenue raised via the NSL meets any revenue shortfall. (page 15)
The model solution for the baseline has the NSL rate necessary to balance the government books rising from 7 per cent in 2006/07 to 12.3 per cent in 2016/17.
It is unclear what assumptions about infrastructure spending are used to establish the baseline in CIE2014. It is clear that it is not assumed that the NSL raises the money to cover any cash flow shortfall for the Norfolk Island Government:
…for the purpose of the baseline, it has been assumed that Australian funding moves to meet Norfolk Island budget deficits. (page 17)
There is some evidence supporting the intuition about the importance of infrastructure spending and financing in CIE2006. It reports the results of sensitivity analysis of the infrastructure assumptions. The baseline and the reform package impact are re-modelled for infrastructure spending at 75 per cent, 50 per cent and 25 per cent of the levels assumed in the original modelling exercise. The results for GDP and the NSL are shown in Chart 5.2. The NSL rate required to balance the Norfolk Island Government cash budget in the baseline drops with the level of spending. GDP also falls with the level of infrastructure spending – for 75 per cent the effect on the deviation from initial conditions is negligible, for 25 per cent the deviation is reduced from about 6 per cent to about 1 per cent. The results for the reform
package are more striking – for 75 per cent the impact drops from about 8 per cent over baseline to about 1 per cent over baseline, and for 25 per cent to 9 per cent below baseline. The comment in CIE2006 is:
This reflects the fact that under the alternative levels of AMP expenditure, the required rate of NSL is lower, and hence the gains from replacing the Norfolk Island tax regime with Australian taxes (and transfers etc) is not as great. Hence GDP is lower as the Australian taxes (net of transfers, subsidies etc) impose a net cost on the local economy.
3.5 Reform scenarios
In broad terms, the single reform package modelled in CIE2006 is qualitatively the same as the four scenarios together in CIE2014. There may be some numerical differences in, for example, the tax rates or the minimum wage level. It seems unlikely that these play a significant role in explaining differences between the two sets of results.
Both CIE2006 and CIE2014 are clear about the major role of tourism in the Norfolk Island economy.
The importance of tourism is illustrated in Chapter 6 of CIE2006 where the sensitivity of the modelling results to changes in the assumption about tourist numbers is considered. In the standard model run tourist numbers are held constant at 30,000 over the period covered. Chart 6.1 reports results for constant 35,000 and constant 28,000. For the baseline at the end of the period GDP is up by 44 per cent over the initial level for 35,000, and down by 20 per cent for 28,000. For the extension of Commonwealth legislation, the deviation from the baseline for GDP is plus 40 per cent for 35,000, and minus 8 per cent for 28,000.
It appears that there is a change between the model structure used to generate the estimates for the baseline and that used to estimate the impact of the reform package in CIE2006, which is not mentioned in the report. In discussing the baseline estimates, it is stated (page 11) that tourism numbers are held constant at 30,000, and that (page 12) tourist spend per day on the island is held constant at $215. There is no discussion of what is assumed about length of stay, or its determinants. However, when considering the impact of the reform package, there is (page 55) a discussion of the response of ‘tourist exports’ to the changes in Norfolk Island prices due to the opposing effects of replacing the NSL with Australian GST and the higher labour costs consequent on the reform package. Whereas it appears that tourism exports are a constant in generating the baseline, they are a variable for the calculation of the impact of the reforms.
CIE2014 does not report results for the impact on tourism of the extension of Commonwealth legislation to Norfolk Island.
Tourism is an area where the data problems attending this kind of modelling for an economy such as that of Norfolk Island are well illustrated. CIE2006 has (Table 2.4, page 11) spend per tourist at $1638, while CIE2014 has an estimated spend of $980 for each tourist arriving by air (air arrivals are in excess of 97 per cent of all tourist arrivals). Inflation over 2006 to 2014 would, other things equal, see the dollar number increase rather than decrease. In nominal terms the CIE2014 spend per tourist is 60 per cent of the CIE2006 spend. Such a large decrease is implausible. It is also noteworthy that CIE2006 has the average tourist spending $532 on accommodation and sustenance (‘Cafes etc’) and $823 on ‘Retail’.
3.7 Norfolk Island Government finances and Commonwealth support
In CIE2006 the impact of the reform package on the amount of Commonwealth support needed by Norfolk Island is approximately $32 million per year, while in CIE2014 it is approximately $7 million per year. It must be recalled that for both the figures given for necessary Commonwealth support are deviations from the baseline. It appears reasonably clear what is mainly driving the difference here between CIE2006 and CIE2014 in forecast levels of necessary Commonwealth support – assumptions about infrastructure spending and financing.
In CIE2006 the baseline has the NSL meeting the excess of Norfolk Island Government expenditure over receipts, so the baseline figure to which the $32 million relates is zero. With the extension of Commonwealth legislation to Norfolk Island the NSL drops out of the modelling, but the AMP remains. It is the AMP that drives the infrastructure costs which are, in the modelling, the main component of Norfolk Island Government expenditure.
In CIE2014 there is no NSL in the baseline, and the Australian Government is assumed to meet any Norfolk Island Government deficit. It is unclear exactly what is assumed about infrastructure spending in the modelling of the baseline and of the impact of the reform scenarios. The estimates of the required transfers from the Commonwealth to the Norfolk Island Government do not include an infrastructure component.
4. Concluding remarks
It is standard practice in ORANI-G applications to report the results for the economic shock being considered in terms of per cent deviations from the baseline. In the case of CIE2006 the baseline itself was reported in terms of per cent deviations from an initial position which is not clearly set out. In the case of CIE2014 there is no information about the baseline. This makes it more difficult to compare, and attempt to explain differences between, the results from the two modelling exercises.
The particular models used in CIE2006 and CIE2014 are different in structure, and numerically. In both cases the quality of the data used for the initial conditions and putting numbers into the structure is poor.
It is probably safe to say that all of the possible sources considered in section 3 above have some role in explaining the differences in the results presented in CIE2006 and CIE2014. It is impossible say much about their relative importance in regard to the forecasts for the main economic indicators, GDP/GTP and Household Consumption. It appears unlikely that differences in the reform scenarios were a major driver of the differences in these variables. A proper investigation of the roles of the possible explanations would be a major undertaking, requiring access to the data bases and the details and numbers for the particular models used, each of which would need to be run under a range of different assumptions. In regard to this, it is noteworthy that although the same consultancy wrote both reports using the same generic model, CIE2014 does not mention CIE2006 and, consequently, offers no analysis of the differences.
Given the poor quality of the Norfolk Island data available in both cases, and the questionable relevance of computable general equilibrium modelling for short and medium term forecasting of reactions to a major shock to a very small economy, my own view is that neither CIE2006 nor CIE2014 was a useful exercise. The substantially different results may be taken as support for that view. It is probably reasonable to conclude that, in terms of the economic indicators considered in these reports, extending Commonwealth legislation would be broadly, and on balance, beneficial to the Norfolk Island economy, to an extent impossible to quantify with any precision. However, it appears to me that one could have arrived at that conclusion by much simpler and much more transparent means. It should also be noted that there are possible economic impacts, such as on income and wealth inequality, that are not considered in these reports, as well as un-considered social impacts.
(1) Centre for International Economics, Economic Impact Assessment of Extending Commonwealth Legislation to Norfolk Island. Report prepared for the Department of Transport and Regional Services, CIE: Canberra & Sydney (October 2006). Document made publicly available in August 2015 and currently available for download at: http://www.norfolkonlinenews.com/).
(2) Centre for International Economics, Final Report: Economic Impact of Norfolk Island Reform Scenarios. Report prepared for the Department of Infrastructure and Regional Development, CIE: Canberra & Sydney (November 2014).
Professor Common will be pleased to respond to questions and comments on his report (email@example.com). More general issues may be addressed to Dr Chris Nobbs (firstname.lastname@example.org).
Please 'contact us' for more information.