NHH Brage (Norges Handelshøyskole)
Not a member yet
8813 research outputs found
Sort by
Ansvarlig bruk av KI: En kvalitativ studie av implementering og retningslinjer
Get PDFKunstig intelligens (KI) har utviklet seg mye de siste årene, hvor spesielt generativ KI har endret flere arbeidspraksiser og bransjer. Et økende antall forskningsartikler peker mot at KI kan skape merverdi både internt i et selskap og eksternt ut mot kunder. Samtidig er det lite forskning på hvordan man implementerer KI i virksomheten, og samtidig sikrer ansvarlig bruk av den nye teknologien. Denne masteroppgaven bidrar derfor til å dekke et forskningsgap, hvor vi utforsker prosessen med å integrere retningslinjer og bruk av KI i to norske bedrifter. Dette er en kvalitativ studie, med en induktiv forskningstilnærming og et utforskende design. Vi har intervjuet tolv informanter om deres erfaringer og opplevelser rundt implementeringsprosessen av KI i sine selskaper.
Basert på våre funn har selskapene gode argumenter for å implementere KI. Samtidig dukker det opp tre dilemmaer i implementeringsprosessen av KI. Det første er mulighetene for raske gevinster KI-teknologi kan gi, men hvor dette også medfører risiko for datasikkerheten til selskapene. Begge selskapene valgte en løsning hvor de utviklet interne KI-verktøy, som sikret ansvarlig bruk av verktøyene. Det andre dilemmaet gjelder spørsmål om det egentlig er behov for egne retningslinjer for bruk av KI, da det allerede dekkes av eksisterende rammeverk for datasikkerhet og personvern hos selskapene. Dette er tett knyttet til det siste dilemmaet som dreier seg om forventinger fra samfunnet opp mot de behovene som de ansatte har. Eksternt press gjør at selskaper utvikler overordnete retningslinjer for KI, mens mangel på digital modenhet blant de ansatte stiller større krav til mer konkrete retningslinjer for KI. Våre funn tyder på at selskapene har lagt for stor vekt på teoretisk opplæring knyttet til retningslinjer og ansvarlig bruk av KI-verktøy. Dette har ført til at ansatte i begge selskaper etterspør en mer praktisk tilnærming til bruk av KI, og bedre dialog mellom ansatte og ledelse for å møte mangel på modenhet.
Formålet med denne masterstudien er å gi en dypere forståelse av hvordan KI kan integreres i organisasjoner. Det er viktig at virksomheter som ønsker å ta i bruk KI, er klar over risikoen for datasikkerheten dette medfører. Det må derfor utarbeides gode og tydelige retningslinjer, for å sikre at organisasjonen følger de offentlige kravene for KI. I tillegg bør det foreligge en strategi for praktisk opplæring og dialog, slik at man ivaretar de ansatte i implementeringsprosessen og sikrer ansvarlig bruk av KI.Artificial intelligence (AI) has evolved significantly in recent years, with generative AI in particular transforming various work practices and industries. An increasing number of research articles suggest that AI can create added value both internally within companies and externally for customers. At the same time, there is little research on how to implement AI in businesses while ensuring the responsible use of this new technology. This master’s thesis aims to address this research gap by exploring the process of integrating AI guidelines and usage in two Norwegian companies. This is a qualitative study with an inductive research approach and an exploratory design. We conducted interviews with twelve participants about their experiences and perceptions of the AI implementation process in their companies.
Based on our findings, the companies have strong arguments for implementing AI. However, three dilemmas emerge during the AI implementation process. The first is the potential for rapid gains offered by AI technology, which also brings risks to the companies' data security. Both companies chose to develop internal AI tools to ensure the responsible use of the technology. The second dilemma concerns whether there is a need for specific guidelines for AI usage, as existing frameworks for data security and privacy already address this to some extent. This is closely tied to the third dilemma, which revolves around societal expectations versus the needs of employees. External pressures lead companies to develop overarching AI guidelines, while a lack of digital maturity among employees necessitates more specific AI guidelines. Our findings indicate that the companies placed too much emphasis on theoretical training related to AI guidelines and responsible use of AI tools. This has resulted in employees in both companies requesting a more practical approach to using AI and better dialogue between employees and management to address the lack of maturity.
The purpose of this master’s study is to provide a deeper understanding of how AI can be integrated into organizations. It is crucial for businesses looking to adopt AI to be aware of the data security risks it entails. Therefore, well-crafted and clear guidelines must be established to ensure the organization complies with public AI regulations. Additionally, there should be a strategy for practical training and dialogue to support employees during the implementation process and ensure the responsible use of AI
From Bill Shock to Solar Pop: How have soaring energy costs changed Norwegian rooftop solar capacity?
Get PDFIn February 2022, political turmoil, supply chain issues, and the COVID-19 pandemic resulted in an European energy crisis. The substantial increase in electricity prices directly impacted the energy bills of millions of households in Norway. The awareness around energy use has increased interest in renewable energy solutions, especially solar photovoltaic (PV) systems. In this paper, we study the relationship between rising electricity prices and the adoption of solar panels among Norwegian households. One key factor that is central to our analysis is the impact of governmental measures, including the electricity subsidy
scheme and ENOVA’s renewable energy support. These measures aim to alleviate the financial burden of high electricity prices and encourage the adoption of green technology. Using a Least Squares Dummy Variable model (LSDV), we explore how the adoption of solar PV technology has been affected by fluctuating electricity prices among the
different bidding areas in Norway in the period 2018-2024. We found that a one percent increase in electricity prices over the last five months correlates with a 0.524 percent rise in solar PV installations in NO1, with similar results in the remaining Norwegian electricity market areas. This result provides evidence that higher energy costs encourage Norwegian end-users to adopt residential solar PV. We propose one main explanation for this: consumers tend to have a myopic relationship towards PV-adoption and electricity prices.
An important implication of this result is the effect of the Norwegian electricity subsidy scheme. While this scheme provides immediate financial relief, it also reduces the incentive for households to invest in energy efficiency technology. This effect raises critical policy implications, highlighting a potential need to balance the short-term goals of the subsidy with strategies that encourage the long-term goals of green technology adoption.In February 2022, political turmoil, supply chain issues, and the COVID-19 pandemic resulted in an European energy crisis. The substantial increase in electricity prices directly impacted the energy bills of millions of households in Norway. The awareness around energy use has increased interest in renewable energy solutions, especially solar photovoltaic (PV) systems. In this paper, we study the relationship between rising electricity prices and the adoption of solar panels among Norwegian households. One key factor that is central to our analysis is the impact of governmental measures, including the electricity subsidy
scheme and ENOVA’s renewable energy support. These measures aim to alleviate the financial burden of high electricity prices and encourage the adoption of green technology. Using a Least Squares Dummy Variable model (LSDV), we explore how the adoption of solar PV technology has been affected by fluctuating electricity prices among the
different bidding areas in Norway in the period 2018-2024. We found that a one percent increase in electricity prices over the last five months correlates with a 0.524 percent rise in solar PV installations in NO1, with similar results in the remaining Norwegian electricity market areas. This result provides evidence that higher energy costs encourage Norwegian end-users to adopt residential solar PV. We propose one main explanation for this: consumers tend to have a myopic relationship towards PV-adoption and electricity prices.
An important implication of this result is the effect of the Norwegian electricity subsidy scheme. While this scheme provides immediate financial relief, it also reduces the incentive for households to invest in energy efficiency technology. This effect raises critical policy implications, highlighting a potential need to balance the short-term goals of the subsidy with strategies that encourage the long-term goals of green technology adoption
An analytical framework to assess the impact of hull fouling in shipping: Insights from predictive modeling and residual analysis
Get PDFThe global shipping industry transports over 80% of the world’s goods by volume. This makes
even small improvements in operational efficiency impactful, both economically and environmentally. One approach to enhancing operational efficiency is management of hull fouling.
Hull fouling is the build-up of marine organisms that increases underwater resistance, which
increases fuel consumption, and consequently drives up costs and carbon emissions. Despite
the importance of hull cleaning, absence of data on hull fouling conditions makes it harder to
demonstrate its impact on ship performance. Therefore, a clear, standardized methodology for
assessing it’s statistical significance on vessel performance is useful, yet currently lacking.
This thesis addresses that gap by developing a methodological framework grounded in predictive modeling and residual analysis to evaluate the impact of hull cleaning, which happens
during dry-docking, on ship performance. As direct measurements of hull fouling conditions
are typically unavailable, the framework relies on indirect indicators and available operational
data. Leveraging pre-dry-dock operational data, predictive models of ship performance are
constructed, incorporating factors such as engine power, weather conditions, and sea state. The
models are then applied to post-dry-dock data, generating predicted performance and enabling
the calculation of residuals. If the post-dry-docking residuals show a statistically significant
positive shift compared to the expected distribution of the residuals, it provides evidence that
hull cleaning has effectively improved vessel performance.
To validate this framework, the methodology is applied to a real life case study vessel. The
results demonstrate the framework’s practicality and robustness, highlighting its potential for
industry stakeholders to apply it to their own vessels and assess whether hull cleaning has a
positive impact on performance. While the study’s findings are based on one vessel, the framework itself is designed to be adaptable and generalizable.
By providing a systematic, data-driven approach to determine whether hull cleaning has a positive impact on a specific vessel, this thesis aims to establish a foundational framework that can
be further developed and ultimately utilized to make informed decisions to optimize cleaning
schedules. These informed decisions may further lead to improved efficiency, cost savings, and
reduced emissions, aligning the maritime industry with broader environmental and regulatory
targets.The global shipping industry transports over 80% of the world’s goods by volume. This makes
even small improvements in operational efficiency impactful, both economically and environmentally. One approach to enhancing operational efficiency is management of hull fouling.
Hull fouling is the build-up of marine organisms that increases underwater resistance, which
increases fuel consumption, and consequently drives up costs and carbon emissions. Despite
the importance of hull cleaning, absence of data on hull fouling conditions makes it harder to
demonstrate its impact on ship performance. Therefore, a clear, standardized methodology for
assessing it’s statistical significance on vessel performance is useful, yet currently lacking.
This thesis addresses that gap by developing a methodological framework grounded in predictive modeling and residual analysis to evaluate the impact of hull cleaning, which happens
during dry-docking, on ship performance. As direct measurements of hull fouling conditions
are typically unavailable, the framework relies on indirect indicators and available operational
data. Leveraging pre-dry-dock operational data, predictive models of ship performance are
constructed, incorporating factors such as engine power, weather conditions, and sea state. The
models are then applied to post-dry-dock data, generating predicted performance and enabling
the calculation of residuals. If the post-dry-docking residuals show a statistically significant
positive shift compared to the expected distribution of the residuals, it provides evidence that
hull cleaning has effectively improved vessel performance.
To validate this framework, the methodology is applied to a real life case study vessel. The
results demonstrate the framework’s practicality and robustness, highlighting its potential for
industry stakeholders to apply it to their own vessels and assess whether hull cleaning has a
positive impact on performance. While the study’s findings are based on one vessel, the framework itself is designed to be adaptable and generalizable.
By providing a systematic, data-driven approach to determine whether hull cleaning has a positive impact on a specific vessel, this thesis aims to establish a foundational framework that can
be further developed and ultimately utilized to make informed decisions to optimize cleaning
schedules. These informed decisions may further lead to improved efficiency, cost savings, and
reduced emissions, aligning the maritime industry with broader environmental and regulatory
targets
Recurrent Neural Networks in Diverse Market Conditions
Get PDFThis thesis investigates the predictive performance of Recurrent Neural Networks (RNNs) in forecasting excess return in zero-coupon bonds. We evaluate their performance using data from the U.S. and German bond markets. The study assesses predictive accuracy and the economic value in different market conditions.
Technically, we implement various models, including linear regressions, Random Forest Regressors, Principal Component Regression (PCR), Partial Least Squares (PLS), and Recurrent Neural Networks (RNNs). Forward rates and macroeconomic variables are integrated to enhance predictive accuracy, and their impact is analyzed across different market conditions, including the COVID-19 pandemic.
Our analysis shows that RNNs achieved statistically significant improvements R-Squared out-of-sample over the benchmark. For longer maturities, we found improvements of up to 35%, much of it as a result of out-performance in 2020 and 2021. These forecasting accuracy gains translated into significant economic value for the U.S. market. Although promising prediction results were also observed for German bonds(bunds), they did not yield the same economic utility.
This study highlights the promise of RNNs for financial forecasting, but also emphasizes the challenge of making models that generalize to the characteristics of multiple markets.This thesis investigates the predictive performance of Recurrent Neural Networks (RNNs) in forecasting excess return in zero-coupon bonds. We evaluate their performance using data from the U.S. and German bond markets. The study assesses predictive accuracy and the economic value in different market conditions.
Technically, we implement various models, including linear regressions, Random Forest Regressors, Principal Component Regression (PCR), Partial Least Squares (PLS), and Recurrent Neural Networks (RNNs). Forward rates and macroeconomic variables are integrated to enhance predictive accuracy, and their impact is analyzed across different market conditions, including the COVID-19 pandemic.
Our analysis shows that RNNs achieved statistically significant improvements R-Squared out-of-sample over the benchmark. For longer maturities, we found improvements of up to 35%, much of it as a result of out-performance in 2020 and 2021. These forecasting accuracy gains translated into significant economic value for the U.S. market. Although promising prediction results were also observed for German bonds(bunds), they did not yield the same economic utility.
This study highlights the promise of RNNs for financial forecasting, but also emphasizes the challenge of making models that generalize to the characteristics of multiple markets
Verdien av ekspertise
Get PDFTidligere studier har vist at kvaliteten på menneskelig input kan forbedre interaksjon og feedback fra store språkmodeller (LLM). Dette indikerer at språkmodellene er mottagelige for menneskelig input. Med dette som utgangspunkt undersøkte denne studien om opplevd ekspertise i input påvirket LLMs verdivurdering av startups. I tillegg undersøkte studien om det var systematiske forskjeller mellom de ledende språkmodellene ChatGPT, Claude, Llama og Gemini. Ved å gjøre et utvalg av 50 norske startups, presenterte vi selskapenes prospekter ved kapitalinnhenting til språkmodellene og spurte om en investeringsanbefaling og en verdivurdering. Språkmodellene ble deretter gitt input som gikk imot deres opprinnelige vurdering. Denne inputen ble presentert som om den kom enten fra en ekspert eller en ikke-ekspert på feltet, med fokus på verdivurdering av oppstartselskaper. Resultatene indikerte at alle modellene endret sine opprinnelige vurderinger basert på menneskelig input. Selv om Gemini var mer skeptisk mot å endre mening uten ytterligere opplysning. Det var ingen signifikant forskjell på hvordan modellene ble påvirket av ekspert- eller ikke-ekspertinput når modellene ble undersøkt samlet. Dette viser at store språkmodeller kan generelt være sensitiv til uenighet, uavhengig av kilden. Studien setter søkelys på behovet for ytterligere forskning på hvordan forbedre robuste og pålitelige språkmodeller ved integrasjon av menneskelig input. Spesielt som beslutningsstøtte i situasjoner med høy usikkerhet og risiko som i Venture Capital (VC)-bransjen
Fertility, Partner Choice, and Human Capital
Get PDFThis paper generates new insights into the effect of education on fertility and partner choice across multiple generations. Using an intensity-of-treatment design, we leverage population-wide panel data for Norway in combination with a school reform in the 1930s, changing the instruction time during the school year in rural municipalities. The reform was binding for most of the rural population and allows us to estimate the effect of education on fertility behavior across the life-cycle, partner choice, and spillover effects on the next generation’s fertility. We present robust evidence of reduced total fertility and an increase in the age at first birth driven by increased years of education, better labor market outcomes, and mating with better-educated partners. In addition, the reform also affected the fertility behavior of the children and decreased fertility rates across multiple generations
«Vi skyt litt i alle retningar» Rektor si skildring av handlingsrommet i arbeid med elevar sitt skulenærvær
Get PDFMasteroppgåva vår er basert på ein kvalitativ studie som undersøkjer rektorar sitt handlingsrom i arbeid med elevar sitt skulenærvær. Vi har sett på korleis rektorar sjølve skildrar sitt arbeid med saker knytt til skulefråvær, og korleis dei jobbar for elevar sitt skulenærvær. For å svare på problemstillinga har vi nytta intervju som metode, og vi har intervjua sju rektorar i Noreg.
Som eit teoretisk rammeverk i oppgåva har vi støtta oss til forsking knytt til leiing, samt sett på korleis kontekst påverkar rektorar sitt handlingsrom. Gjennom studien har vi utvikla ein modell for å skissere rektor sitt handlingsrom. Vi har lagt vekt på korleis rektorar snakkar om tematikken, og i den samanheng har det vore naudsynt å drøfte språk og omgrep knytt til temaet.
Vi er sjølve skuleleiarar, og opplever i praksisfeltet at elevar har eit større omfang av skulefråvær no enn tidlegare. Samstundes ser vi at rektorane er usikre på korleis dei skal handtere desse sakene, og som ein av informantane våre sa: «Vi skyt litt i alle retningar». Både Utdanningsdirektoratet og Kunnskapsdepartementet har peika på at det er behov for betre oversikt over omfanget av saker ute i skulane. 1.august 2024 trer den nye opplæringslova i kraft, og frå denne dagen er kommunane pliktige til å setje inn tiltak dersom elevar har skulefråvær.
Vi har analysert funna våre opp mot relevant teori. På bakgrunn av analysen vil vi peike på at det er behov for vidare forsking på området, samt tydelegare rammer frå styresmaktene. Det er behov for eit felles språk innan denne tematikken, og det må verte avklart korleis vi skal forstå desse sakene. Tilsette i skulen treng klare føringar for å lukkast i arbeidet med elevar sitt skulenærvær.nhhma
A Distributional Robust Analysis of Buyback and Cap-and-Trade Policies
Get PDFThis study delves into a dynamic Stackelberg game comprised of a manufacturer and a retailer, operating in an environment with fluctuating demand and price-dependent consumer behavior. The multi-period optimization challenges the manufacturer to strategically set wholesale and buyback prices, while the retailer determines the retail price and order quantities within a single contract. In this dynamic framework, the players operate under the constraints of a cap-and-trade policy, with limited knowledge of demand distributions, characterized only by mean and standard deviation parameters. To address this inherent uncertainty, we employ a distributionally robust approach. Additionally, we explore the enduring effects of historical decisions on present-day demand, reflecting a memory-like market behavior. Through numerical examples, we illuminate the influence of buyback contracts and cap-and-trade policies on decision-making processes within this setting
The Effects of Labour Migration and Interventions on Tax Compliance
Get PDFFirst of all, I would like to thank my supervisors at the Norwegian School of Economics (NHH),
Evelina Gavrilova-Zoutman and Floris Zoutman for excellent guidance throughout my PhD.
Following my many years outside academia, prior to my PhD journey, their effort was timely
and appreciated. I would also like to thank Jarle Møen for facilitating the admission to the PhD
program at the NHH. Thanks also to Lars Jonas Andersson at NHH for providing good
guidance on machine learning, and support for my notion on its relevance to this project.
This PhD could not have been realised without the funding and support from the Norwegian
Research Council and the Norwegian Tax Administration (NTA). But equally important, the
many brilliant colleagues in the latter institution. I would like to thank Marcus Zackrisson who
paved the way for this project at management level, and Terje Nordli and Monica Bredesen for
anchoring the interest among the colleagues engaged in the important work of preventing tax
related labour market crime. Research coordinator at the NTA, Torhild Henriksen, deserves a
special thanks for facilitating and maintaining research interaction between the NTA and the
NHH, and for giving professional advice during the whole period.
I am very thankful for the great talks and substantial contributions to machine learning
provided by Nils Gaute Voll. Øystein Olsen and Tore Sjøstedt helped a lot with specifying the
data extraction. I would like to thank Hanne Beate Næringsrud for the many comments on
specific issues, including interpretation, message, and narratives. Nina Serdarevic, Julia
Tropina Bakke, Knut Løyland, Inge Sandstad Skrondal, and Arnstein Øvrum have provided
useful comments on earlier drafts as well. I would also like to thank Kari Djupdal, Anders
Berset, Andreas Olden, Joakim Døving Dalen, and Terje Dalen for the many fruitful
discussions I have had over the years at the NTA. What a great knowledge pool you all are!
A special thanks goes to Anne May Melsom who was appointed my co-supervisor at the NTA
and co-authored two of the papers. Not only do I owe you for improving my Stata knowledge
to an adequate level, but also for your impeccable understanding of the data, your substantial
inquiries, and finally for the considerable effort to the very end. Your help has been invaluable.
I would also like to thank academic staff at other institutions, for their valuable insights and
comments. Those are Joel Slemrod at the University of Michigan, Ann Arbor, Steinar Strøm,
Andreas Kotsadam and Thor Olav Thoresen at the University of Oslo, and Hamed Saiedi at
the Norwegian Business School.
I am grateful for the many conversations and colloquial preparations with peer students at the
Stockholm Doctoral Course Program in Economics, Econometrics and Finance (SDPE) jointly
by Stockholm University and Stockholm School of Economics, where I undertook most of my
4
course components. I am equally thankful for the talks and encounters at various conferences
with my peer PhD students at the NHH.
A particular gratitude goes to my dear friends at the NTA, namely Hanne Beate Næringsrud,
Julia Tropina Bakke, Øystein Olsen, Anders Berset, and Ivana Haakens for being there in
challenging times. Friends like You last a lifetime.
I am forever thankful to Rebecka Maria Norman for the continuous, but nevertheless (at least
for me) useful discussions on so many topics on statistical inference over the years. I hope our
kids, Felicia, Gabriel, and August, were not permanently damaged by nitty gritty talks on
standard error clustering or heteroscedasticity. Their patience has been remarkable.
Finally, my heartfelt gratitude to Ann-Kristin Midtskog for new perspectives on compliance,
text clarifications, strategic choices, and for your unconditional love and support (“…Og jeg
kan ikke miste det uansett hva som skjer”).
Oslo, December 2023
Thomas Lang
Cross-border shopping of alcohol – What is the effect on tax revenue and sales and which products are most affected?
Get PDFWe use COVID-19 border closings and comprehensive store-level data on Norwegian alcohol sales to quantify the effect cross-border shopping of alcohol on sales volume and commodity tax revenue. Effects are large, for instance we estimate that commodity tax revenue for wine is about 20% lower because of cross-border shopping. Using product level data we establish that effects come from across all products rather than just a few, but effects are especially marked for bag-in-box wines. Neither availability of the exact same product in Sweden nor idiosyncratic product-level price difference with respect to Sweden has any marked effect on the impact of cross-border shopping on sales