We estimate the causal effects of a large unanticipated natural disaster on high schoolers’ university enrolment decisions and subsequent medium-term labour market outcomes. Using national administrative data after a destructive earthquake in New Zealand, we estimate that the disaster raises tertiary education enrolment of recent high school graduates by 6.1 percentage points. The effects are most pronounced for males, students who are academically weak relative to their peers, and students from schools directly damaged by the disaster. As relatively low ability males are overrepresented in sectors of the labour market helped by the earthquake, greater demand for university may stem from permanent changes in deeper behavioural parameters such as risk aversion or time preference, rather than as a coping response to poor economic opportunities.
We aim to investigate the role of insurance in business recovery following the devastating Christchurch earthquake in February, 22nd, 2011. We analyze data from two business surveys conducted after the earthquake to examine how insurance affected business operation in the aftermath of the earthquake both in the short-term and longer-term. For the short-term analysis, we use a combination of propensity score matching (PSM) and linear probability model (LPM) to analyze the data. We first estimate the propensity scores for insurance take-up of each firm conditional on the firm’s individual characteristics. Stratification based on the estimated propensity scores is used to match the treated (insured) and the control (uninsured) firms. We then estimate the probability of firms’ continuing operations with a set of control variables to account for the level of damage and disruption caused by the quake in each stratum. We find little evidence of any beneficial effect of insurance coverage on business continuity in the short-run. For the longer-term analysis, we analyze the available survey data using logistic regression. The result suggests that business interruption insurance significantly promotes increased level of long-term productivity for surviving firms following the earthquake.
The Mѡ=7.1 Darfield (Canterbury) earthquake struck on 4 September 2010, approximately 45 km west of Christchurch, New Zealand. It revealed a previously unknown fault (the Greendale fault) and caused billions of dollars of damage due to high peak ground velocities and extensive liquefaction. It also triggered the Mw=6.3 Christchurch earthquake on 22 February 2011, which caused further damage and the loss of 185 lives. The objective of this research was to determine the relationship between stress and seismic properties in a seismically active region using manually-picked P and S wave arrival times from the aftershock sequence between 8 September 2010-13 January 2011 to estimate shear-wave splitting (SWS) parameters, VP =VS-ratios, anisotropy (delay-time tomography), focal mechanisms, and tectonic stress on the Canterbury plains. The maximum horizontal stress direction was highly consistent in the plains, with an average value of SHmax=116 18 . However, the estimates showed variation in SHmax near the fault, with one estimate rotating by as much as 30° counter-clockwise. This suggests heterogeneity of stress at the fault, though the cause remains unclear. Orientations of the principal stresses predominantly indicate a strike-slip regime, but there are possible thrust regimes to the west and north/east of the fault. The SWS fast directions (ø) on the plains show alignment with SHmax at the majority of stations, indicating stress controlled anisotropy. However, structural effects appear more dominant in the neighbouring regions of the Southern Alps and Banks Peninsula.
This dissertation contains three essays on the impact of unexpected adverse events on student outcomes. All three attempt to identify causal inference using plausibly exogenous shocks and econometric tools, applied to rich administrative data. In Chapter 2, I present evidence of the causal effects of the 2011 Christchurch earthquake on tertiary enrolment and completion. Using the shock of the 2011 earthquake on high school students in the Canterbury region, I estimate the effect of the earthquake on a range of outcomes including tertiary enrolment, degree completion and wages. I find the earthquake causes a substantial increase in tertiary enrolment, particularly for low ability high school leavers from damaged schools. However, I find no evidence that low ability students induced by the earthquake complete a degree on time. In Chapter 3, I identify the impact of repeat disaster exposure on university performance, by comparing outcomes for students who experience their first earthquake while in university, to outcomes for students with prior earthquake exposure. Using a triple-differences estimation strategy with individual-by-year fixed effects, I identify a precise null effect, suggesting that previous experience of earthquakes is not predictive of response to an additional shock two years later. The final chapter investigates the impact of injuries sustained in university on academic performance and wages, using administrative data including no-fault insurance claims, emergency department attendance and hospital admissions, linked with tertiary enrolment. I find injuries, including minor injuries, have a negative effect on re-enrolment, degree completion and grades in university.
The last seven years have seen southern New Zealand a ected by several large and damaging earthquakes: the moment magnitude (MW) 7.8 Dusky Sound earthquake on 15 July 2009, the MW 7.1 Dar eld (Canterbury) earthquake on 4 September 2010, and most notably the MW 6.2 Christchurch earthquake on 22 February 2011 and the protracted aftershock sequence. In this thesis, we address the postseismic displacement produced by these earthquakes using methods of satellite-based geodetic measurement, known as Interferometric Synthetic Aperture Radar (InSAR) and Global Positioning System (GPS), and computational modelling. We observe several ground displacement features in the Canterbury and Fiordland regions during three periods: 1) Following the Dusky Sound earthquake; 2) Following the Dar eld earthquake and prior to the Christchurch earthquake; and 3) Following the Christchurch earthquake until February 2015. The ground displacement associated with postseismic motion following the Dusky Sound earthquake has been measured by continuous and campaign GPS data acquired in August 2009, in conjunction with Di erential Interferometric Synthetic Aperture Radar (DInSAR) observations. We use an afterslip model, estimated by temporal inversion of geodetic data, with combined viscoelastic rebound model to account for the observed spatio-temporal patterns of displacement. The two postseismic processes together induce a signi cant displacement corresponding to principal extensional and contractual strain rates of the order of 10⁻⁷ and 10⁻⁸ yr⁻¹ respectively, across most of the southern South Island. We also analyse observed postseismic displacement following the Dusky Sound earthquake using a new inversion approach in order to describe afterslip in an elasticviscoelastic medium. We develop a mathematical framework, namely the "Iterative Decoupling of Afterslip and Viscoelastic rebound (IDAV)" method, with which to invert temporally dense and spatially sparse geodetic observations. We examine the IDAV method using both numerical and analytical simulations of Green's functions. For the post-Dar eld time interval, postseismic signals are measured within approximately one month of the mainshock. The dataset used for the post-Dar eld displacement spans the region surrounding previously unrecognised faults that ruptured during the mainshock. Poroelastic rebound in a multi-layered half-space and dilatancy recovery at shallow depths provide a satisfactory t with the observations. For the post-Christchurch interval, campaign GPS data acquired in February 2012 to February 2015 in four successive epochs and 66 TerraSAR-X (TSX) SAR acquisitions in descending orbits between March 2011 and May 2014 reveal approximately three years of postseismic displacement. We detect movement away from the satellite of ~ 3 mm/yr in Christchurch and a gradient of displacement of ~ 4 mm/yr across a lineament extending from the westernmost end of the Western Christchurch Fault towards the eastern end of the Greendale East Fault. The postseismic signals following the Christchurch earthquake are mainly accounted for by afterslip models on the subsurface lineament and nearby faults.
<strong>Sea level rise is one consequence of Earth’s changing climate. Century-long tide gauge records show that global-mean sea-level rise reached 11-16 cm during the twentieth century at a mean rate of 1.2 mm/y. Today, the average rate of global-mean sea-level rise is higher at 3-4 mm/y and is expected to increase in the future. This represents a hazard to low elevation coastal zones worldwide. Yet, before global sea level projections can be used to characterise future coastal flood hazard at a local scale, the effects of tectonics (and other processes) that drive vertical land motion (VLM) must be considered. VLM is defined as the vertical velocity (uplift or subsidence) of the solid surface with respect to the centre of Earth. In this study, new VLM maps are generated over coastal strips in New Zealand, using Sentinel-1 InSAR and GNSS data.</strong>In New Zealand, measuring VLM using InSAR on naturally vegetated or agricultural land is difficult due to signal decorrelation. Along the rural Bay of Plenty coastal strip, I use a persistent-scatterer approach to generate a VLM map from both east-looking ascending and west-looking descending Sentinel-1 data between 2015-2021. Using time-series data over the same time period from a dense network of 20 GNSS sensors, I tie InSAR-derived line-of-sight velocity to the 2014 ITRF reference frame. I test two different methods for measuring VLM and compare the results against GNSS vertical velocity along the Bay of Plenty coast. Best results are achieved by first removing the interpolated horizontal GNSS velocity field from each of the InSAR datasets, before averaging the two VLM estimates. Measured VLM is between -3 and 3 mm/y, with negative values (subsidence) occurring within the low-lying Rangitāiki Plain and Ōpōtiki valley, and uplift across the elevated region west of Matatā.This thesis integrates geomorphological, geological, and historical levelling VLM records with modern satellite datasets to assess VLM across timescales ranging from 10 to 100,000 years at Matatā. Uplift rate has been variable through time, with average uplift over the last 300,000 years of 1 mm/y, 4.5 mm/y since 1720 years, 2 mm/y between 1950-1978, and 10 mm/y between 2004-2011. Previous modelling has shown that the best fit to the 2004-2011 rapid uplift rates is an inflating magmatic source at ~10 km depth beneath Matatā. To reconcile all data, I present a VLM model that consists of short-lived periods (7 years) of rapid uplift (10 mm/y), separated by longer periods (30 years) of lower background uplift (3 mm/y). The episodic nature of VLM at Matatā likely reflects short-lived periods of magmatic intrusion. Episodic VLM characterised by large rates of uplift (10 mm/y) has been seen at Taupō volcano, and other volcanic centers globally. It has been 12 years since the end of the last intrusion episode; this modelling suggest one may expect to observe increased uplift rates at Matatā in the coming decades. Densely populated urban coastal strips are most at risk from the effects of relative sea-level rise. At the same time, anthropogenic activities associated with urbanization, such as groundwater withdrawal, and land reclamation can lead to local land subsidence (LLS), further exacerbating the risk to urban infrastructure. LLS refers to subsidence relative to nearby land area assumed to be stable. In this thesis, I create the first high-resolution (10 m) maps of LLS at six urban coastal strips in New Zealand, with a combined length of 285 km, using Sentinel-1 InSAR data between 2018-2021. This analysis reveals 89% of urban coastal strips are subsiding at rates of -0.5 mm/y or greater, and 11% is subsiding at higher rates of -3.0 mm/y or greater. On average, subsidence is -0.6 to -2.9 mm/y higher at the coastal strip, compared to inland areas occupied by GNSS stations. This analysis also documents highly-localised hotspots of LLS, with subsidence rates of up to -15 mm/y. In Christchurch, rapid and localised subsidence (-8 mm/y) is observed within coastal suburbs New Brighton and Southshore. In most cities, the highest subsidence rates occur on land reclaimed in the early-late twentieth century, and in areas built on Holocene sediment. Time-series analysis of LLS at sites of reclaimed land shows both linear and non-linear rates of deformation over time periods of up to 6-8 years. This thesis highlights the variable exposure to relative sea-level rise of New Zealand coastal strips, and demonstrates that in many cases current rates of VLM should be expected to continue for the next few decades.
