Labour Mobility with Vocational Skill: Australian Demand and Pacific Supply

How many immigrants with less than university education, for a given immigration quota, maximise economic output? The answer is simple—zero—in the canonical model of the labour market, where the marginal product of a university-educated immigrant is always higher. We build an alternative model, following Jones (2005), in which national production occurs through a set of Leontief production functions that shift over time with technological change. This model, more consistent with historical data from Australia, implies a positive output-maximising supply of vocationallyskilled immigration, which we estimate for likely scenarios through the year 2050. Australian demand for vocationally-skilled workers will substantially exceed native supply, especially of fundamental workers performing tasks that do not require college education but are difficult to automate or offshore. Historical patterns of growth and technological change imply a demand-supply gap of two million vocationally-skilled workers by 2050. Australia can maximise future output with expanded vocationally skilled labour mobility (settler or temporary), especially by focusing on labour supply from the Pacific Islands. www.cgdev.org Satish Chand and Michael A. Clemens


Introduction
Suppose an ocial is charged with setting quotas on the number of visas for foreigners to work in Australia. She would face pressure to admit only workers that the economy "needs". Workers with university education are often considered to be needed. But how would the ocial place a number on the economic "need" for workers with less than tertiary education, such as those with vocational training?
A common stance in immigration policy is that the economic need for such workers is close to zero. The United States and United Kingdom, for example, have explicitly barred almost all employment-based settler immigration for workers without tertiary education. 1 Australia has largely lacked an ocial policy toward labour migration for vocationally-skilled work work (Wright and Clibborn 2017), despite a growing mismatch between focus on high educational qualications and the needs of the labour market (Wright et al. 2016).
In this paper, we build a simple model of the Australian economy in which the output-maximising quantity of vocationally-skilled immigration is positive. This is because the substitution of university-educated workers for vocationally-educated workers is 1) limited by the speed of technological change, as in the seminal model of Jones (2005), and 2) inherently limited for some tasks. We use the model to build rough estimates of the demand for less-educated workers in the Australian economy over the next three decades, and the likely supply of those workers among natives of Australia and of the Pacic region. Our contribution is to distinguish between falling relative demand for some tasks that do not require a university education-those that are being substantially automated or oshored-and other occupations with stable relative demand that 1 Under Prime Minister Theresa May, the United Kingdom restricted its General Work Visa (with minor exceptions) to migrants earning over £30,000 per year; the median worker even with the highest level of secondary education earns £18,450. A UK government review of the evidence concluded that "a policy on work migration that provided greater access for higher-skilled migration while restricting access for lower-skilled workers to enter the UK would be consistent with the available evidence. " (Manning et al. 2018, 111). In the United States, the inuential Jordan Commission created by Congress recommended "the elimination of the admission of unskilled workers"-that is, those with no college degree. Today the United States oers only 0.1% of its settler visas for newly-arrived immigrants on the basis of employment that does not require university education. It oers no temporary visa for nonseasonal work that does not require university education. UK data source in Appendix. Jordan Commission report is USCIR (1995,102). In scal year 2018, the United States admitted newly-arrived 528,727 permanent immigrants, of which 468 (0.086%) were for work requiring less than a college degree (EB-3 "needed unskilled workers", Yearbook of Immigration Statistics 2018, ' 1 we call fundamental workers. We nd a large gap between projected demand and native supply, reaching roughly two million by the year 2050 (stock, not ow).
Broadly, this gap can be reduced in three ways: accepting lower university-educated employment and economic growth, occupational downgrading by highly educated native workers, or immigration by vocationally educated foreign workers (temporary or permanent). We explore the case for this last option and in particular the case for recruiting vocationally educated workers from the Pacic region.
The paper begins by presenting the new model in section 2 and discussing its implications in section 3. Section 4 discusses how the model is more consistent with recent Australian data than the canonical model used to study immigration, and section 5 explores the implications for future scenarios of demand and supply of vocationally-skilled workers in Australia. The potential for managed migration focused on Pacic-region labour supply is considered in section 6, and section 7 concludes.

Demand for workers with less than university education
The eect of labour on national production can be modelled in various ways. The canonical model admits two types of workers: one with high education and high marginal product, another with low education and low marginal product. In this framework, national output is always strictly higher if a worker with vocational education is replaced by a worker with university education.
These assumptions yield a strong implication for immigration policy. In such an economy, admitting any given immigrant worker with vocational education imposes an economic cost relative to admitting a university-educated worker-a cost that could only be compensated by noneconomic policy objectives such as humanitarian protection or family reunication. It implies that for any given quota of immigrants, the output-maximising number of vocationally-educated immigrants is zero.
Here we set out a simple theoretical framework in which the output-maximising number of 2 immigrants with vocational skill is nonzero. Prior literature has focused on the impact of foreign workers in Australia on the distribution of income, generally nding little to no eect (Breunig et al. 2017;Crown et al. 2020). We focus instead on macroeconomic impact: estimating the demand for immigrants with vocational education that would maximise productivity and scal revenue.

Labour demand: The canonical model
Studies of the economic eects of immigration focus on the "canonical" model of the relative demand for labour by education level (Tinbergen 1974;Katz and Murphy 1992;Acemoğlu and Autor 2011;Dustmann et al. 2016). In the essence of that model, national output . is created by a long-run neoclassical production technology with two types of labour, constant returns to scale, and constant elasticity of substitution f: (1) ! ⌘ is workers who have completed higher education (diploma, associate degree, university degree), ! a is workers who have completed vocational education (secondary or Certicate I-IV), aggregate labour is ! ⌘ ! ⌘ + ! a , U < 1 captures their relative productivity, and constant is an overall labour productivity parameter capturing both capital and technology.
An ocial setting immigration quotas would want to know: When a growing economy employs more workers with college education (! ⌘ ), how many more workers with vocational education (! a ) does it employ? Assuming perfect competition in factor markets, output is maximised when the wage ratio for the two types of labour equals the marginal rate of technical substitution: Isocost curves are dened by a constant labour cost share Taking the total dierential of (1), we can use (2) and then (3) to arrive at a version of the Slutsky equation, in quantities employed: 2 The rst term in (5) is the pure scale eect of unity: In the short run-with a Leontief production technology, and thus xed factor proportions-a percentage rise in employment with college education (and thus output) would require the same percentage rise in employment with vocational education. That is, it reects the short-run elasticity of substitution f = 0. The second term captures the negative substitution eect in the long run: If the technology is not Leontief (f > 0) then rising employment of workers with college education can cause an even larger rise in output if a technology can be developed that is relatively more intensive in workers with college education, with an osetting negative eect on demand for workers with vocational education.
The sign of the net eect is ambiguous if the two types of workers are relative substitutes, that is, if the Hicks-Allen direct elasticity of substitution f exceeds unity (Stern 2011). Empirically, it does: In a group of advanced economies including Australia, f is estimated at 1.8-2.6 (Jerzmanowski and Tamura 2020). Thus it is possible that workers with college education, in this model, can displace workers with vocational education even as they cause the economy to grow.

Limitations of the canonical model
Two key assumptions above limit the policy relevance of the canonical model. First, the relative labour demand equation (5) applies to the ambiguously-timed "long run" (Hamermesh 1986), but immigration quotas must be set in the short run. Second, it assumes the same possibilities of substitution between workers with college education and all types of workers with vocational education, regardless of the tasks they perform. Intuitively, new ideas from college-educated workers to substitute for vocationally-educated workers might arrive at very dierent rates for, say, telephone operators than for childcare workers. Here we discuss relaxing those assumptions.
Rather than the ambiguous "long run", the immigration ocial might be interested in the demand for each type of labour at a moment in nite time. We can estimate that using Jones's (2005) interpretation of the national production function (1): It is a reduced form capturing not a single production technology, but the long-run substitution possibilities between dierent (Leontief) short-run production techniques. Combining the two types of labour in a new way, at a substantially dierent ratio, requires new ideas-and ideas arrive over time at a certain rate.
That feasible rate of change in the organisation of national production is captured by f. From (5) the instantaneous rate of change 3 ln(! a /! ⌘ ) If ideas capable of reorganising national production arrive very slowly (f ⇡ 0), the labour ratio remains constant during growth as the economy expands along Leontief isoclines. The shift in the labour ratio over a nite time period characterises the degree of substitution between production techniques made feasible by the arrival of new ideas during that period.
This points to a second important limitation of the canonical model. The ideas needed to reorganise production might not arrive at the same rate for all types of workers with vocational education: They might arrive relatively quickly for telephone operators and very slowly for child care workers. Thus in the medium run the employment of college educated workers might reduce the demand for telephone operators even as it raises demand for child care workers (Cortés and Tessada 2011;Cortés and Pan 2013). The labour literature has recently focused on this concern, as the canonical model has proven unable to adequately explain the heterogeneity of changing education premia across the education spectrum in many advanced economies over the last half century (Bárány and Siegel 2018), often called labour market polarisation (Autor and Dorn 2013; Goos et al. 2014). This new literature reconsiders the national production function as combining not workers directly but tasks that are in turn produced by workers. It explores how substitutability can vary across tasks performed by workers even within the same education level.
Such polarisation of labour demand by task has been evident in Australia as well (Coelli and Borland 2016) though it is less marked than in other countries, possibly due to wage regulations limiting employment growth in the lowest-wage jobs (Wilkins and Wooden 2014, 424). This new literature oers tools to consider the relationship between labour demand at dierent education levels in a way that accounts for Australian particularities. This focus on tasks has only begun to shape the study of immigration (e.g Foged and Peri 2016;Basso et al. 2017).

A task-based alternative
Consider a national long-run technology that directly combines types of labour as tasks, not people (Autor et al. 2003;Autor and Dorn 2013). This conception of labour as tasks is distinguished with the script ↵. Some of these tasks are "complex" or abstract and require formal higher education (↵ ⌘ ), such as the services of a surgeon. Other, "basic" tasks can be performed by a person with formal education to the vocational level only (↵ a ). Some of these are replaceable (↵ a,A )that is, can be automated or oshored-such as the services of a medical transcriptionist. Other tasks that can be performed by low-education worker are not replaceable (↵ a,5 )-thus "xed" or "fundamental"-such as the services of a nurse's aide. 3 National output is produced by combining these tasks in the nested, constant elasticity of substitution production function Automatable and/or oshorable tasks ↵ a,A can be substituted by greater use of abstract, complex tasks ↵ ⌘ more easily than xed or fundamental tasks ↵ a,5 . This dierence is captured by the Hicks direct elasticities as f ⌧ X.
In this model, the creation of complex employment ↵ ⌘ again aects demand for the other tasks ↵ a,A , ↵ a,5 , through a short-run scale eect and a long-run substitution eect. In the short run, before new ideas to rearrange production arrive, production is Leontief and additional complex employment causes an equal percentage rise in demand for the other two employment types.
In the long run, the rise in demand for the other two employment types is mitigated by technological change to use complex employment more intensively. It might be tempting to characterise that substitution using the Hicks direct elasticities of substitution in (7) and use them to write a separate version of the Slutsky equation (5) for each labour type. But the response of demand for ↵ a,5 to a rise in employment of ↵ ⌘ depends on the substitutability of ↵ ⌘ and ↵ a,A . The labour demand relationship analogous to (5) for 'fundamental' tasks thus requires the Morishima (1967) elasticity of substitution between nests, which in this case is a weighted average of the Hicks direct elasticities (Anderson and Moroney 1993;Frieling and Madlener 2016). This gives the long-run gross elasticities where`is the (one-way) Morishima elasticity of demand for fundamental tasks to the marginal product of complex tasks. By denition, 0 <`< X. As above, these correspond to the nite-time That is, over any nite period, the change in the ratio of 'fundamental' employment to complex employment is slower than the change in the ratio of 'replaceable' employment to complex employment.
The last step is to introduce workers. Every task must be produced by a worker, so that ↵ ⌘ + ↵ a,A + ↵ a,5 = ! ⌘ + ! a . Some workers with only vocational education can manage to acquire the skills and experience to be as productive in complex jobs as those with higher education (upskilling: ! ⌘ a ), provided they do not face a licensing barrier. Others have higher education but nevertheless perform tasks that do not require it (downskilling: ! a ⌘ ). Still others work at the level for which they are educated: ! a a , ! ⌘ ⌘ . Thus, Either type of worker (⌘, a) can be native (=) or immigrant (8), and these are perfect substitutes 7 within education groups: This assumption is for simplicity in the baseline analysis, despite evidence that otherwise observably identical natives and immigrants are not perfect substitutes (Peri and Sparber 2009). We will discuss relaxing this assumption below. Finally, the whole population # includes non-working elderly or juvenile dependants ! 3 alongside workers, so that # = ! ⌘ + ! a + ! 3 .

Economic and fiscal implications of skill-selective immigration policy
The model has an important implication for the economically ecient regulation of immigration. Often, as discussed above, admitting a vocationally-educated immigrant worker is seen as necessarily reducing national productivity at the margin, relative to admitting a universityeducated worker. Indeed, countries with greater increases in the average education of the labour force experience greater economic growth (e.g. Cohen and Soto 2007).
But this reasoning implicitly assumes an innite elasticity of substitution between the tasks performed by high-and low-education workers. The optimal labour demand equations (9) are derived from the output-maximisation condition (2). So when there is a rise in employment that requires high education, relative employment requiring low education must change to the degree shown, or output would be lower than if it had. If the right sides of either of the equations (9) is positive, employment requiring low education must rise as well.
For example, if an immigrant engineer can only produce engineering work (↵ ⌘ ) by consuming the services of a child care worker (low substitutability`), then maximizing output requires raising employment in child care (↵ a,5 ). If the supply of natives with low education is rising, that demand can be lled domestically. If not, then maximising output requires admitting a low-education immigrant (! 8 a ) or native downskilling (! ⌘ produces ↵ a,5 ).

8
The model also has an important implication for the scal eect of low-education immigration.
This eect is typically analysed with individual-level accounting exercises: summing the lifetime taxes paid directly by such immigrants, net of benets they receive. This too would suggest a rule always preferring high-education immigrants, who on average pay higher taxes.
Again this implicitly assumes perfect substitution between the work performed by immigrants and natives. But if they are not perfect substitutes, and if low-education immigrants raise the productivity of other workers, they generate tax revenue indirectly (Colas and Sachs 2020). This shortcoming is recognised by inuential studies of the scal impact of immigration, which nevertheless proceed to carry out individual-level scal accounting (Blau and Mackie, eds e.g. 2017, 324;Manning et al. 2018, 71). because the indirect eects are dicult to measure.
But suppose that net tax revenue, and thus the long-run size of government, is a constant share of national product. 4 This implies that net tax revenue is maximised by maximising output. So again, the addition of new high-education employment requires the changes in low-education employment described in the labour demand relations (9), or net scal revenue is lower than it could have been. In the example above, the tax revenue from the engineer's employment is foregone, indirectly, without the employment of the child care worker. It is the overall eect of the addition of both workers to the labour force that determines the scal eect of low-education immigration, not a simple account of the taxes paid by the child care worker.

Trends in the tasks and education of the Australian workforce
Immigration to Australia raises GDP and reduces unemployment (AboElsoud et al. 2020), today as it has for the prior century (Pope and Withers 1993). 5 The Australian economy has been shifting to use complex tasks relatively more than basic tasks over time. But sizeable employment in some basic tasks has remained essential for employment in complex tasks. Consistent with the task-based production function (7), the demand for fundamental tasks relative to complex tasks has remained remarkably constant over time, while the relative demand for replaceable tasks has fallen. This suggests a hard limit on the substitutability of complex workers for basic workers, one that is not present in standard models of the labour market with two types of worker skill.
Seeing this in the data requires an empirical denition of the types of tasks in the model. For now, we dene 'complex' occupations as requiring a university degree or equivalent experience.
'Fundamental' occupations are those that are neither oshorable nor automatable. We dene an 'oshorable' occupation as one that is not rated as strictly 'not oshorable' by Blinder and Krueger (2013). We dene an 'automatable' occupation as one placed in the top third Dorn 2013, 1571) of the Routine Task Index proposed by Acemoğlu and Autor (2011). Table 1   The trends in relative demand for these task types is shown in Figure 1. The demand for fundamental workers relative to complex workers has changed little in the past two decades, even while relative demand for replaceable workers has collapsed. Since the late 1990s the number of replaceable workers employed for each complex worker has fallen by more than half. But the number of fundamental workers employed per complex worker-roughly one-did not change over the same period. By equation (9), this implies an elasticity of substitution between complex employment and replaceable employment (X) that is large in absolute value, but an elasticity of substitution between fundamental employment and complex employment (`) that is very small. This in turn suggests that demand for many tasks not requiring university education-those classied as fundamental-will remain strong decades into the future as a necessary complement to projected advances in the demand for complex tasks.  This trend is highlighted in Figure 3. It presents the ratio between the 'Education, native' line and the 'Tasks' line for each task level from Figure 2. While natives with task-appropriate education (or less) have made up a historically rising and recently steady share of employment at the tertiary level, natives are a declining share of those with task-appropriate education at the vocational level. The trends imply that within a few years, more than half of all vocational-level work in Australia will be lled either by immigrant workers or by native workers with more education than is needed to perform their jobs. Year Two important trends are clear in these graphs. First, demand for some basic occupations will remain strong into the future and will go hand-in-hand with rising demand for more complex occupations. Second, that demand is being increasingly met by immigrants or by native downskilling (working at tasks below one's education level). This suggests that, if native downskilling is considered undesirable, rising demand for complex workers might require a certain degree of immigration for work in vocational-level tasks. In the next section we estimate this relationship between complex employment and the degree of immigration for basic tasks, given trends in native labour supply to those tasks.

Economic growth and the demand for basic tasks
Under the model's assumptions, we can create scenarios for the evolution of demand for vocational tasks as the economy grows and demands more complex tasks. We begin by estimating a relationship between growth and complex tasks, under dierent assumptions about techno-14 logical change. The model above then implies a given change in demand for basic tasks. That change in demand will be compared, in the following section, to the evolving supply of native workers and Pacic immigrant workers.

From growth to demand for tasks, with demographic change
First, the model implies that a given rise in income per capita (. /# ) requires a certain change in complex tasks. Start with a simplied version of the task-based production function (7) that lumps together the basic tasks: . Dividing by # , taking the total derivative with respect to ↵ ⌘ , and rearranging gives The rst term on the right side (hereafter W) includes the co-movement of complex employment and overall productivity as well as change in the demographic structure # ↵ ⌘ . The second term (hereafter \ ) accounts for technological change that uses complex tasks relatively more intensively than basic tasks.
The equation (14) can be easily estimated from the data for 1986-2016. In the estimation, . is national output in each year at constant prices, # is the entire resident population of Australia.
↵ ⌘ is total employment in occupations requiring tertiary education or commensurate knowledge and experience, as rated by the Australian Bureau of Statistics, 6 and ↵ a is all other employment.
The best t to the data givesŴ = 0.366 and\ = 1.10.
This allows us to create scenarios of the future demand for complex tasks given any scenario for the growth of GDP per capita given past rates of demographic change as shown in the regression equation (14). If future years bring a faster percentage change in the overall population (including elderly dependants) relative to the number of workers in complex tasks, additional workers are required to achieve the same GDP per capita growth for all residents including those not working.
The demand for complex tasks must rise by that additional amount (the increase in growth of 3 ln # 3 ln ↵ ⌘ in future years relative to past years). 6 The method for classifying occupations in the census data is described in the Appendix. We begin this estimation in the rst row of Table 2. The rst row shows the assumed annual rate of growth in real GDP per capita up to each year shown, beginning in the last census year of 2016. The 1.8% annual growth rate represents a continuation of the trend from 1986 to 2016.
Note that this growth rate assumes past rates of demographic change, as in equation (14). But the demographic change term 3 ln # 3 ln ↵ ⌘ is not forecast to be the same in coming years. Figure 4 shows estimates of the total Australian population relative to the stock of tertiary-educated workers since the mid-1980s (green) and projected over the next three decades (red). The ageing of the Australian population means that each increment in complex workers will be met with a larger rise than before in the total population: the dierence in the two slopes is 0.003. In other words, the size of the complex labour force will need to rise 0.3% faster each year to achieve the same historical rise in GDP per person due to this acceleration of demographic change. Thus the assumed growth rate of GDP per capita in the fourth through sixth columns of Table 2 is 2.1% per year.
We choose the an age range for "working age" that stops at 65 because Australian labour force The Demand section of Table 2 uses this growth rate to estimate the necessary rise in complex employment. Because the estimation ofŴ above already accounted for the shifting composition of employment across tasks, the rise in complex employment associated with the assumed economic growth is simply the cumulative economic growth from the second row times the estimateŴ from equation (14). Starting from the size of complex employment in 2016, this cumulative growth rate in complex employment gives the new, additional complex employment needed to achieve the assumed growth rate (row 4).
Given the projected ratio of tertiary-level task employment to vocational-level task employment in Figure 2c, this implies the associated amount of new, additional vocational-level employment (row 5). These estimates do not assume current technology, but rather a continuation of the past rate of technological change shifting relative demand for the two types of workers, as in equations (6) and (9). Year Source: Table 2.

Native supply
The next rows of Table 2  Because the number of working-age natives will not substantially rise over the forecast period, and because natives are a falling share of vocational-task work, the number of native workers in vocational-level tasks is projected to fall over this period. The native decline in this estimation reaches almost 1 million by the year 2050.
The nal row of Table 2 simply takes the dierence between the additional demand for vocational tasks (row 5) and the additional native supply of vocational tasks (row 7). With accelerated demographic change, the gap between new demand and native supply reaches about 870,000 by 2030, 1.5 million by 2040, and over 2 million by 2050. Figure 5 shows the same data graphically.

The potential for managed migration
The quantitative exercise reported in Table 2  An important role for managed migration for targeted sectors is the subject of some degree of political consensus. The last Intergenerational Report by the Australian Commonwealth Treasury notes, "Migrants are expected to continue to be the largest source of population growth.
Migration contributes to economic growth and can help oset population ageing. However, migration needs to be managed well to ensure it supports higher living standards" (Commonwealth of Australia 2021, vii). The report places priority on "improving the skills of workers" (p. xiii), noting that permanent migrants arrive younger and better skilled thus make "a higher economic and scal contribution during their time in Australia" (p. 22).
We use three examples from the care industry to explain how matching the supply of TVETtrained migrants for existing and emerging employment in Australia can make a signicant contribution to the economy and welfare of natives.
Aged care and the National Disability Insurance Schemes (NDIS) are two of the largest service industries, and those projected to create the most employment until the middle of this century.
Both these sectors require skilled workers, the majority of whom provide personalised and es-sential care to a vulnerable segment of society that is neither routine nor oshoreable. Aged care as of 2016 employed 366,000 workers, 68,000 volunteers, and another 28,000 unpaid helpers (Pagone et al. 2021, 63). The sector is estimated to account for 3 percent of the Australian workforce, with the challenge of the future being able to "attract and retain a skilled workforce from a relatively smaller working population" (p. 24). The projections are that aged care will need an additional 80,000 workers by 2030 and 180,000 by 2050 (p. 42). Similarly, the NDIS was anticipated to create 162,000 full-time equivalent positions by 2019-2020 when the scheme was to come into full operation (NDIS 2015, 19). While the implementation of the NDIS has been delayed, the growth in workforce requirement is projected to accelerate until 2030 before stabilising at the rate of population growth (Pagone et al. 2021, 107).
Further, the participation of women in the workforce rests on access to aordable childcare, another non-oshoreable and non-routine service in which appropriately skilled migrants can make a signicant contribution. The demand for vocationally-skilled workers within the care sector as a whole is likely to increase in line with the ageing of the population, the expansion of access to the NDIS, and increases in participation of women in the workforce. The costs within the care sector are largely driven by wage costs, which in turn are dependent on supply of appropriately skilled workers.
It is unlikely that the supply of a sustainable skilled workforce for aged care and the NDIS will be satised locally. Permanent migrants have traditionally lled long-term needs of the economy (Ongley and Pearson 1995) while temporary migrants have met transient need for skills such as those for seasonal agricultural workers (Fels et al. 2019).

Supply of vocationally skilled labour in the Pacific
Managed migration to ll these gaps could, in the abstract, come from anywhere. But perhaps the largest opportunity for mutually benecial managed migration in the decades to come exists in the Pacic Islands. This region has provided agricultural workers to meet seasonal demand, and the same could apply for vocationally-skilled workers more broadly.
To date, however, employment immigration from the Pacic to Australia has been extremely low. Figure 6a shows the fraction of the Australian workforce by task composed of immigrants from all countries, comparing it with the fraction from Pacic Island countries. Figure 6b shows the same breakdown by level of education. Broadly speaking, almost none of the work-based immigration to Australia has come from the Pacic, at any level of work. And even the scant work-based immigration from the region has been heavily dominated by workers from Polynesia rather than Melanesia, where the great supply will exist.
A similarly sharp skew against the Pacic region exists in temporary work visas. The microstate visa and Pacic Labour Scheme remain small, and the Seasonal Worker Program is not focused on building vocational skills (Brickenstein 2015). As of June 2020 the Pacic Labour Scheme listed just 984 workers present in Australia.
This highlights a major opportunity. Demand for vocational occupations will rise greatly in 21 Australia in the decades to come, particularly in 'fundamental' occupations. At the same time, the supply of natives with training appropriate for those jobs will shrink. The Pacic has sharply rising stocks of workers who are at or near the appropriate level of training, human capital that has gone largely untapped by Australia's skill-focused immigration policy to date. A strategy of focusing on the Pacic to prepare and hire many of those workers could spur economic growth and the creation of additional higher-order employment in Australia, help oset the eects of demographic change, and serve Australia's foreign policy goals in the region.
Regional preferences go against the political pressure for nondiscriminatory immigration policy that has sought to place all foreign countries on equal footing as a reaction to the legacy of the White Australia Policy (Rivett 1988, 276). That reaction to the burdens of history is understandable. But we note several reasons why the active facilitation of vocationally-skilled labour mobility from the Pacic region could serve Australia's interests.

Why the Pacific?
The Pacic Islands are particularly suited to meet Australian demand for workers with vocational skills for ve reasons that we detail below: (i) their physical and cultural proximity to Australia; (ii) the rising supply of skill in the Islands' workforce; (iii) the opportunities for Australia to further upskill the requisite workforce at source; (iv) the potential for circulation of skilled workers between Australia and the islands; and (v) the potential for regional labour mobility to complement Australia's foreign policy, commercial policy, and geopolitical objectives in the region.
Physical and cultural proximity: The suspension of international travel during the -19 pandemic saw a sharp decline in labour supply to Australian agriculture, with Working Holiday Maker visas collapsing by an order of magnitude from 2019 to 2020 (ABARES 2021). Some of the worker shortages were lled from the Pacic Islands, due to their physical proximity to Australia and the ready pool of workers available there. This was a period when Australian farmers resorted to the use of charted ights with specialised quarantine facilities to bring in workers from the neighbouring Pacic islands to harvest and tender crops, securing supplies of food for the public (see ABC News). Cultural connectedness to Australia is just as important, with Pacic islanders having close trading and sporting links that extend for more than a century.
The shortage of workers during lockdowns has highlighted the need for a reliable supply and a sustainable source for workers for the Australian economy rather than heavy reliance on working holiday makers from outside the Pacic (Howe et al. 2019). Access to a stable workforce, as evidenced by the regional Recognised Seasonal Employer (RSE) scheme of New Zealand, boosts labour productivity and raises income (Gibson and Bailey 2021) and in times of crisis becomes a safeguard of national security (Ciccone 2020). The circulation of workers between rural and remote regions, and in healthcare and social assistance where most jobs are projected to be created (LMIP 2020), could be critical to the survival of regional communities and services.
Rising skill prevalence in the region: As large as the worker shortages in Table 2 may seem, the human capital to meet that demand will exist in the Pacic. Figure 7 shows forecasts of the working-age population with secondary education in Australia and the Pacic region. The num-ber of working-age people with secondary education in the Pacic island countries will rise by over four million by 2050. Workers with completed secondary education would either already have the skills needed to perform many of the new vocational tasks that will be demanded in Australia, or could acquire those skills in a relatively short training program such as the Certicate III programs oered by the Australia-Pacic Training Coalition.
The overall supply of workers from the Pacic islands and Timor-Leste is projected to increase by 88 percent, or to a total of 12. Opportunities for circular migration: Finally, the Pacic islands oer an opportunity for circulation of workers between Australia, New Zealand, and the neighbouring islands. The large wage gaps between the Pacic islands and Australia provide the economic incentives for international mobility (Howe et al. 2019), with the potential for work experience abroad sucient to secure long-term employment at home and income for retirement. Workers from the Pacic have experience in subsistence agriculture and hospitality, thus are familiar with the seasonal nature and manual demands of both these industries. They can also pick up skills from being formally employed in a commercial operation requiring accredited qualications-those that may be gained through a combination of TVET and on-the-job experience. Many of the islands lack the scale to oer the requisite work experience for TVET graduates, a void that could be lled through temporary work experience abroad. Proximity to Australia together with family links at home means that pacic islanders are likely to return home. Indeed, Gibson and McKenzie (2011) provide evidence of the 'best and the brightest' from the islands having a high propensity to return home after having worked for short periods abroad-a fact largely attributed to the cultural connectedness to the kin left behind.
Complementing Australia's other regional objectives: Greater and more benecial regional labour mobility is consistent with the Australian Government's policy of a 'Pacic Step-up' that aims to strengthen the people-to-people links with the aim of greater strategic security, economic stability, and political sovereignty for regional governments. The Australian government has acknowledged the long-standing and enduring nature of its relationship with the Pacic islands, and has set the goal of "promoting economic cooperation and greater integration within the Pacic and also with the Australian and New Zealand economies, including through labour mobility" (DFAT 2017, 99).
In other words, labour mobility within the region serves Australia's development, foreign policy, public diplomacy, and trade objectives (cf. Barker 2010). It is a high-impact form of overseas assistance. "The A$99.4 million in net income that Tonga has gained through the [Seasonal Worker] programme since its inception, " notes the World Bank (2018, xi), "is more than double the annual bilateral aid budget from the Australian Government". Migration for vocationally-skilled work generates important knowledge transfer to Pacic countries of origin, such as technologies to improve horticultural yields (Dun et al. 2018). Beyond creating overall employment in the region, labour mobility especially creates formal employment where it is extremely scarce. Just 16 percent of the labour force of Papua New Guinea holds a formal sector job (ILO 2017, 4). Finally, regional labour mobility strengthens Australia's geopolitical interests in the Pacic in a historical era of growing inuence by China (White 2011;Beeson 2017). Enhanced labour mobility has long been a priority for regional governments (e.g. PIF 2006, 2), but de-facto expan-sion of access to the Australian labour market remains highly limited, especially for Melanesia.
The Australia-Pacic Training Coalition, for example, though created expressly to extend labour market access to the region, according to the World Bank (2017, 37) "has failed to deliver on its labor-mobility mandate". All of the motives underlying Australia's New Comprehensive Strategic and Economic Partnership with Papua New Guinea, announced in 2019, are well served by enhanced labour market access.

Conclusion
The evidence we have discussed yields two broad lessons. The rst is that a policy to actively recruit workers with less than college education is in the medium-term economic interest of native Australian workers. The exact level of that migration cannot be forecast with precision.
But there is strong evidence that roughly half of the Australian employment share for less-thancollege workers will be fundamental workers, resistant to ongoing trends of automation and oshoring.
For Australian native workers, the likely alternative to a national policy for recruiting vocationallyskilled labour will be either reduced economic welfare or occupational downgrading. Australia's increased focus on the admission of highly trained immigrants in the 1990s did not aect indicators of the usage of their training in the Longitudinal Survey of Immigrants to Australia (Tani 2020). A shift in immigration policy toward actively seeking vocationally-skilled labour is needed as the economy recovers from the 2020 contraction and will be more acutely needed in the decades to follow.
In principle the labour to supply the rising demand for vocationally-skilled workers could come from anywhere on earth. But the second lesson of this analysis is that a substantial portion of it could and should come from the Pacic region. Education levels in the region have been rising sharply and the supply of high-school-educated workers in the region is far more than sucient to supply any realistic share of Australia's demand for foreign vocationally-skilled workers. Beyond this, developing pathways for those workers from the Pacic will serve Australia's self-interested development policy, trade policy, public diplomacy, and geopolitical interests in the region. Current initiatives such as the Pacic Labour Facility and Australia-Pacic Training Coalition, while precisely the seed of such pathways that is needed, remain extremely limited in magnitude. Chand et al. (2021) propose one regional mobility scheme-the 'Pacic Skills Visa'that could contribute to realising these gains.